‫ﻧﺎﻡ ﭘﺮﻭﺭﺩﮔﺎﺭ ﻳﻜﺘﺎ‬

‫ﺏ‬

‫‪1‬‬
 ‫ﺟﺰﻭﻩ ﺩﺭﺳﻲ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﭘﻴﺸﺮﻓﺘﻪ‬

‫ﺑﺮﺍﻱ ﺩﺍﻧﺸﺠﻮﻳﺎﻥ ﻛﺎﺭﺷﻨﺎﺳﻲ ﺍﺭﺷﺪ ﻣﻬﻨﺪﺳﻲ ﻛﺎﻣﭙﻴﻮﺗﺮ‬

‫ﮔﺮﺍﻳﺶ ﻫﺎﻱ ﻥﺭﻡ ﺍﻓﺰﺍﺭ ﻭ ﻫﻮﺵ ﻣﺼﻨﻮﻋﻲ‬

‫ﺗﺎﻟﻴﻒ ‪ :‬ﺩﻛﺘﺮ ﺭﺿﺎ ﻗﺎﺋﻤﻲ‬

‫‪2‬‬
 ‫ﻓﺼﻞ ‪:1‬‬
‫ﻣﺮﻭﺭﻱ ﺑﺮ ﻣﺪﻳﺮﻳﺖ‬
‫ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ‬

‫‪3‬‬
 ‫‪ 1-1‬ﻣﻘﺪﻣﻪ ‪ :‬ﺳﻴﺴﺘﻢ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﭼﻴﺴﺖ؟‬
‫ﺳﻴﺴﺘﻢ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﻳﺎ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ )‪ (Data Base‬ﺑﻪ ﻃﻮﺭ ﻋﻤﻮﻣﻲ ﻳﻚ ﺳﻴﺴﺘﻢ ﻛﺎﻣﭙﻴﻮﺗﺮﻱ‬
‫ﻧﮕﻬﺪﺍﺭﻱ ﺭﻛﻮﺭﺩﻫﺎ ﻣﻲ ﺑﺎﺷﺪ‪ .‬ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺭﺍ ﻣﻲ ﺗﻮﺍﻥ ﺑﻪ ﻋﻨﻮﺍﻥ ﻗﻔﺴﻪ ﺑﺎﻳﮕﺎﻧﻲ ﺍﻟﻜﺘﺮﻭﻧﻴﻜﻲ ﺩﺭ ﻧﻈﺮ‬
‫ﮔﺮﻓﺖ ﻛﻪ ﻣﺨﺰﻧﻲ ﺑﺮﺍﻱ ﻓﺎﻳﻞ ﺩﺍﺩﻩ ﻛﺎﻣﭙﻴﻮﺗﺮﻱ ﺍﺳﺖ ‪ .‬ﻛﺎﺭﺑﺮﺍﻥ ﺳﻴﺴﺘﻢ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﻋﻤﻠﻴﺎﺕ ﮔﻮﻧﺎﮔﻮﻧﻲ ﺭﺍ‬
‫ﺑﺮ ﺭﻭﻱ ﺍﻳﻦ ﻓﺎﻳﻞ ﻫﺎ ﺍﻧﺠﺎﻡ ﺩﻫﻨﺪ‪ ،‬ﻣﺎﻧﻨﺪ ‪:‬‬
‫• ﺍﻓﺰﻭﺩﻥ ﻓﺎﻳﻞ ﻫﺎﻱ ﺟﺪﻳﺪ ﺑﻪ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ‬
‫• ﺍﻓﺰﻭﺩﻥ ﺩﺍﺩﻩ ﻫﺎ ﺑﻪ ﻓﺎﻳﻞ ﻫﺎﻱ ﻣﻮﺟﻮﺩ‬
‫• ﺑﺎﺯﻳﺎﺑﻲ ﺩﺍﺩﻩ ﻫﺎ ﺍﺯ ﻓﺎﻳﻞ ﻫﺎﻱ ﻣﻮﺟﻮﺩ‬
‫• ﺗﻐﻴﻴﺮ ﺩﺍﺩﻩ ﻫﺎﻱ ﻓﺎﻳﻞ ﻣﻮﺟﻮﺩ‬
‫• ﺣﺬﻑ ﺩﺍﺩﻩ ﻫﺎ ﺍﺯ ﻓﺎﻳﻞ ﻣﻮﺟﻮﺩ‬
‫• ﺣﺬﻑ ﻓﺎﻳﻞ ﻫﺎﻳﻲ ﺍﺯ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ‬
‫ﻫﻤﺎﻧﻄﻮﺭ ﻛﻪ ﮔﻔﺘﻴﻢ ﺳﻴﺴﺘﻢ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ‪ ،‬ﻳﻚ ﺳﻴﺴﺘﻢ ﻛﺎﻣﭙﻴﻮﺗﺮﻱ ﺑﺮﺍﻱ ﻧﮕﻬﺪﺍﺭﻱ ﺭﻛﻮﺭﺩﻫﺎ ﻣﻲ‬
‫ﺑﺎﺷﺪ ﺑﺪﻳﻦ ﻣﻌﻨﻲ ﻛﻪ ﻳﻚ ﺳﻴﺴﺘﻢ ﻛﺎﻣﭙﻴﻮﺗﺮﻱ ﻛﻪ ﻫﺪﻑ ﺁﻥ ﺫﺧﻴﺮﻩ ﺍﻃﻼﻋﺎﺕ ﺍﺳﺖ ﻭ ﻛﺎﺭﺑﺮﺍﻥ ﻣﻲ‬
‫ﺗﻮﺍﻧﻨﺪ ﺁﻥ ﺍﻃﻼﻋﺎﺕ ﺭﺍ ﺑﺎﺯﻳﺎﺑﻲ ﻳﺎ ﺑﻪ ﻫﻨﮕﺎﻡ ﻛﻨﻨﺪ ‪ .‬ﻫﺮ ﭼﻴﺰﻱ ﻛﻪ ﺑﺮﺍﻱ ﻳﻚ ﻓﺮﺩ ﺑﺎ ﻳﻚ ﺳﺎﺯﻣﺎﻥ ﺑﺎ ﺍﺭﺯﺵ‬
‫ﺑﺎﺷﺪ‪ ،‬ﺍﻃﻼﻋﺎﺕ ﻧﺎﻡ ﺩﺍﺭﺩ ‪ .‬ﻳﻚ ﺳﻴﺴﺘﻢ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺍﺯ ﭼﻬﺎﺭ ﻗﻄﻌﻪ ﺗﺸﻜﻴﻞ ﺷﺪﻩ ﺍﺳﺖ ‪ :‬ﺩﺍﺩﻩ ﻫﺎ ‪،‬‬
‫ﺳﺨﺖ ﺍﻓﺰﺍﺭ‪ ،‬ﻧﺮﻡ ﺍﻓﺰﺍﺭ ﻭ ﻛﺎﺭﺑﺮﺍﻥ‪.‬‬

‫‪4‬‬
 ‫ﺩﺍﺩﻩ ﻫﺎ )‪(Data‬‬
‫ﻫﺮﭼﻴﺰﻱ ﻛﻪ ﺑﺮﺍﻱ ﻳﻚ ﻓﺮﺩ ﻳﺎ ﺳﺎﺯﻣﺎﻥ ﺑﺎ ﺍﺭﺯﺵ ﺑﺎﺷﺪ ‪ ،‬ﺩﺍﺩﻩ ﺩﺭ ﻧﻈﺮ ﮔﺮﻓﺘﻪ ﻣﻲ ﺷﻮﺩ‪.‬‬
‫ﻗﺎﺑﻠﻴﺖ ﻫﺎﻳﻲ ﻛﻪ ﻳﻚ ﺳﻴﺴﺘﻢ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺍﺭﺍﺋﻪ ﻣﻲ ﻛﻨﺪ‪ ،‬ﺗﺎ ﺣﺪ ﺯﻳﺎﺩﻱ ﺑﻪ ﻛﺎﻣﭙﻴﻮﺗﺮﻱ ﺑﺴﺘﮕﻲ ﺩﺍﺭﺩ‬
‫‪ .‬ﺳﻴﺴﺘﻢ ﻫﺎ ﺑﺮ ﺭﻭﻱ ﻛﺎﻣﭙﻴﻮﺗﺮﻫﺎﻱ ﺑﺰﺭگ‪ ،‬ﻣﻌﻤﻮﻻ‬ ‫ﻛﻪ ﺍﻳﻦ ﺳﻴﺴﺘﻢ ﺑﺮ ﺭﻭﻱ ﺁﻥ ﻛﺎﺭ ﻣﻲ ﻛﻨﺪ‬
‫ﭼﻨﺪﻛﺎﺭﺑﺮﻩ ﻫﺴﺘﻨﺪ‪ ،‬ﺩﺭ ﺣﺎﻟﻴﻜﻪ ﺳﻴﺴﺘﻢ ﻫﺎ ﺩﺭ ﻛﺎﻣﭙﻴﻮﺗﺮﻫﺎﻱ ﻛﻮﭼﻚ‪ ،‬ﺗﻚ ﻛﺎﺭﺑﺮﻩ ﻫﺴﺘﻨﺪ ‪ .‬ﺳﻴﺴﺘﻢ‬
‫ﺗﻚ ﻛﺎﺭﺑﺮﻩ ﺳﻴﺴﺘﻤﻲ ﺍﺳﺖ ﻛﻪ ﺩﺭ ﺁﻥ‪ ،‬ﺩﺭ ﺁﻥِ ﻭﺍﺣﺪ ﻓﻘﻂ ﻳﻚ ﻧﻔﺮ ﻣﻲ ﺗﻮﺍﻧﺪ ﺍﺯ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ‬
‫ﺍﺳﺘﻔﺎﺩﻩ ﻛﻨﺪ ﻭ ﺳﻴﺴﺘﻢ ﭼﻨﺪﻛﺎﺭﺑﺮﻩ ﺳﻴﺴﺘﻤﻲ ﺍﺳﺖ ﻛﻪ ﺩﺭ ﺁﻥِ ﻭﺍﺣﺪ ﭼﻨﺪ ﻧﻔﺮ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﺍﺯ ﻳﻚ‬
‫ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺍﺳﺘﻔﺎﺩﻩ ﻛﻨﻦﺩ ‪‍.‬‬
‫*ﻫﺪﻑ ﺍﺻﻠﻲ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﭼﻨﺪﻛﺎﺭﺑﺮﻩ ﺍﻳﻦ ﺍﺳﺖ ﻛﻪ ﺑﻪ ﻛﺎﺭﺑﺮ ﺍﺟﺎﺯﻩ ﺩﻫﻨﺪ‪ ،‬ﺍﺯ ﺁﻥ ﺑﻪ ﻋﻨﻮﺍﻥ ﻳﻚ‬
‫ﺳﻴﺴﺘﻢ ﺗﻚ ﻛﺎﺭﺑﺮﻩ ﺍﺳﺘﻔﺎﺩﻩ ﻧﻤﺎﻳﺪ‪.‬‬
‫ﺑﻪ ﻃﻮﺭ ﻛﻞ ﺩﺍﺩﻩ ﻫﺎ ﺩﺭ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ‪ ،‬ﺑﺨﺼﻮﺹ ﺩﺭ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﺑﺰﺭگ‪ ،‬ﻫﻢ ﺑﻪ ﺻﻮﺭﺕ ﻣﺠﺘﻤﻊ ﻭ‬
‫ﻫﻢ ﺑﻪ ﺻﻮﺭﺕ ﺍﺷﺘﺮﺍﻛﻲ ﻫﺴﺘﻨﺪ )‪ (Integration & Sharing‬ﻛﻪ ﺩﺭ ﺍﺩﺍﻣﻪ ﺗﻮﺿﻴﺢ ﺩﺍﺩﻩ ﻣﻲ ﺷﻮﻧﺪ‪:‬‬
‫ﺟﺎﻣﻌﻴﺖ )‪ :(Integration‬ﻣﻔﻬﻮﻡ ﺟﺎﻣﻌﻴﺖ ﻳﺎ ﻣﺠﺘﻤﻊ ﺑﻮﺩﻥ ﺍﻳﻦ ﺍﺳﺖ ﻛﻪ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﻣﺠﻤﻮﻋﻪ‬
‫ﺍﻱ ﺍﺯ ﻓﺎﻳﻞ ﻫﺎﺳﺖ ﻛﻪ ﺑﺨﺸﻲ ﺍﺯ ﺍﻃﻼﻋﺎﺕ ﺍﺿﺎﻓﻲ ﺍﺯ ﺑﻴﻦ ﺁﻥ ﻫﺎ ﺣﺬﻑ ﺷﺪﻩ ﺍﺳﺖ ‪ .‬ﺍﺯﺩﻳﺪﮔﺎﻩ ﻫﺎﻱ ﺩﻳﮕﺮ‪،‬‬
‫ﺟﺎﻣﻌﻴﺖ ﺑﻪ ﻣﻌﻨﺎﻱ ﺻﺤﺖ ﺩﺍﺩﻩ ﻫﺎ‪ ،‬ﭘﺮﺩﺍﺯﺵ ﻫﺎ ﻭ ﭘﻴﺮﻭﻱ ﺍﺯ ﻣﻘﺮﺭﺍﺕ ﺳﻴﺴﺘﻢ ﻣﻲ ﺑﺎﺷﺪ‪ .‬ﻩﺏ ﻋﻨﻮﺍﻥ‬
‫ﻣﺜﺎﻝ‪ ،‬ﻣﻮﺟﻮﺩﻱ ﻭﺍﻗﻌﻲ ﺣﺴﺎﺏ ﻫﺎﻱ ﺑﺎﻧﻜﻲ ﻧﺒﺎﻳﺪ ﻣﻨﻔﻲ ﺑﺎﺷﺪ‪.‬‬
‫ﺍﺷﺘﺮﺍﻙ )‪ :(Sharing‬ﻣﻔﻬﻮﻡ ﻣﺸﺘﺮﻙ ﺑﻮﺩﻥ ﺍﻳﻦ ﺍﺳﺖ ﻛﻪ ﺩﺍﺩﻩ ﻫﺎﻱ ﻣﻮﺟﻮﺩ ﺩﺭ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﻣﻲ‬
‫ﺗﻮﺍﻧﺪ ﺑﻴﻦ ﻛﺎﺭﺑﺮﺍﻥ ﻡ ﺧﺘﻠﻒ ﺑﻪ ﺍﺷﺘﺮﺍﻙ ﮔﺬﺍﺷﺘﻪ ﺷﻮﻧﺪ‪ ،‬ﻳﻌﻨﻲ ﻫﺮ ﻳﻚ ﺍﺯ ﻛﺎﺭﺑﺮﺍﻥ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﺍﺯ ﻳﻚ‬
‫ﺑﺨﺶ ﺍﺯ ﺩﺍﺩﻩ ﻫﺎ ﺑﺮﺍﻱ ﺍﻫﺪﺍﻑ ﺧﺎﺻﻲ ﺍﺳﺘﻔﺎﺩﻩ ﻛﻨﻨﺪ ‪ .‬ﻫﻤﻴﻨﻄﻮﺭ ﻣﻴﺘﻮﺍﻥ ﮔﻔﺖ ﻛﺎﺭﺑﺮﺍﻥ ﺑﻪ ﻃﻮﺭ ﻫﻤﺰﻣﺎﻥ‬
‫ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﺑﻪ ﻳﻚ ﺑﺨﺶ ﺍﺯ ﺩﺍﺩﻩ ﻫﺎ ﺩﺳﺘﻴﺎﺑﻲ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻨﺪ ‪ .‬ﻫﺮ ﻛﺎﺭﺑﺮ ﻓﻘﻂ ﺑﺨﺸﻲ ﺍﺯ ﺩﺍﺩﻩ ﻫﺎﻱ ﺑﺎﻧﻚ‬
‫ﺍﻃﻼﻋﺎﺗﻲ ﺭﺍ ﻣﻮﺭﺩ ﺍﺳﺘﻔﺎﺩﻩ ﻗﺮﺍﺭ ﻣﻲ ﺩﻫﺪ ﻭ ﺗﻮﺳﻂ ﺩﻳﺪﮔﺎﻩ ﻫﺎﻱ ﻣﺨﺘﻠﻔﻲ ﺑﺎ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺩﺭ ﺗﻌﺎﻣﻞ‬
‫ﺍﺳﺖ‪ .‬ﺍﮔﺮ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﻣﺸﺘﺮﻙ ﻧﺒﺎﺷﺪ‪ ،‬ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺷﺨﺼﻲ ﻳﺎ ﻛﺎﺭﺑﺮ ﻭﻳﮋﻩ ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪5‬‬
 ‫ﺳﺨﺖ ﺍﻓﺰﺍﺭ )‪(Hardware‬‬
‫ﻗﻄﻌﺎﺕ ﺳﺨﺖ ﺍﻓﺰﺍﺭﻱ ﺳﻴﺴﺘﻢ ﺷﺎﻣﻞ ﻣﻮﺍﺭﺩ ﺯﻳﺮ ﺍﺳﺖ‪:‬‬
‫‪ ‬ﺣﺎﻓﻈﻪ ﻫﺎﻱ ﺟﺎﻧﺒﻲ ‪ :‬ﻛﻪ ﻣﻌﻤﻮﻻ ﺩﻳﺴﻚ ﻫﺎ ﻫﺴﺘﻨﺪ‪ ،‬ﺑﺮﺍﻱ ﺫﺧﻴﺮﻩ ﺩﺍﺩﻩ ﻫﺎ ﺑﻪ ﻛﺎﺭ ﻣﻲ ﺭﻭﻧﺪ ﻭ‬
‫ﺩﺳﺘﮕﺎﻩ ﻫﺎﻱ ﻭﺭﻭﺩﻱ‪/‬ﺧﺮﻭﺟﻲ ﻣﺎﻧﻨﺪ ﮔﺮﺩﺍﻧﻨﺪﻩ ﻫﺎ‪ ،‬ﻛﻨﺘﺮﻟﮕﺮﻫﺎﻱ ﺩﺳﺘﮕﺎﻩ ﻫﺎ ﻭ ‪...‬‬
‫‪ ‬ﭘﺮﺩﺍﺯﻧﺪﻩ ﻫﺎﻱ ﺳﺨﺖ ﺍﻓﺰﺍﺭﻱ ﻭ ﺣﺎﻓﻈﻪ ﻫﺎﻱ ﺍﺻﻠﻲ ﻛﻪ ﺑﺮﺍﻱ ﭘﺸﺘﻴﺒﺎﻧﻲ ﺍﺯ ﺍﺟﺮﺍﻱ ﻧﺮﻡ ﺍﻓﺰﺍﺭ‬
‫ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺑﻪ ﻛﺎﺭ ﻣﻲ ﺭﻭﻧﺪ‪.‬‬

‫ﻧﺮﻡ ﺍﻓﺰﺍﺭ)‪(Software‬‬
‫ﺳﻪ ﻧﻮﻉ ﻧﺮﻡ ﺍﻓﺰﺍﺭ ﺭﺍﻳﺞ ﺩﺭ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﻭﺟﻮﺩ ﺩﺍﺭﺩ ‪ :‬ﺳﻴﺴﺘﻢ ﻣﺪﻳﺮﻳﺖ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ )‪،(DBMS‬‬
‫ﺳﻴﺴﺘﻢ ﻋﺎﻣﻞ ﻭ ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﻛﺎﺭﺑﺮﺩﻱ‪ .‬ﺳﻴﺴﺘﻢ ﻣﺪﻳﺮﻳﺖ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﻻﻳﻪ ﺍﻱ ﺍﺯ ﻧﺮﻡ ﺍﻓﺰﺍﺭ ﺑﻴﻦ ﺩﺍﺩﻩ‬
‫ﻫﺎﻱ ﻓﻴﺰﻳﻜﻲ ﺫﺧﻴﺮﻩ ﺷﺪﻩ ﻭﻛﺎﺭﺑﺮﺍﻥ ﺳﻴﺴﺘﻢ ﻣﻲ ﺑﺎﺷﺪ ﻛﻪ ﺑﻪ ﺁﻥ ﻛﺎﺭﮔﺰﺍﺭ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ‬
‫)‪(Database Server‬ﮔﻔﺘﻪ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﻛﺎﺭﺑﺮﺍﻥ)‪(Users‬‬
‫ﻛﺎﺭﺑﺮﺍﻥ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺭﺍ ﻣﻲ ﺗﻮﺍﻥ ﺳﻪ ﺩﺳﺘﻪ ﺩﺭ ﻧﻈﺮ ﮔﺮﻓﺖ ‪:‬‬
‫‪ ‬ﺑﺮﻧﺎﻣﻪ ﻧﻮﻳﺴﺎﻥ ﻛﺎﺭﺑﺮﺩﻱ ﻛﻪ ﻣﺴﺌﻮﻝ ﻧﻮﺷﺘﻦ ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﻛﺎﺭﺑﺮﺩﻱ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺍﻧﺪ ‪.‬‬
‫‪ ‬ﻛﺎﺭﺑﺮﺍﻥ ﻧﻬﺎﻳﻲ ﻛﻪ ﺍﺯ ﻃﺮﻳﻖ ﺍﻳﺴﺘﮕﺎﻩ ﻫﺎﻱ ﻛﺎﺭﻱ ﻳﺎ ﭘﺎﻳﺎﻧﻪ ﻫﺎ ﺑﻪ ﺳﻴﺴﺘﻢ ﺩﺳﺘﺮﺳﻲ ﺩﺍﺭﻧﺪ‪.‬‬
‫‪ ‬ﺍﻳﻦ ﺩﺳﺘﻪ ﺍﺯ ﻛﺎﺭﺑﺮﺍﻥ ﻛﻪ ﺩﻳﺪﻩ ﻧﻤﻴﺸﻮﺩ‪ ،‬ﻣﺪﻳﺮ ﻳﺎ ﺳﺮﭘﺮﺳﺖ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ )‪ (DBA‬ﺍﺳﺖ ﻛﻪ‬
‫ﻭﻇﻴﻔﻪ ﺳﺮﭘﺮﺳﺖ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﻭ ﺳﺮﭘﺮﺳﺖ ﺩﺍﺩﻩ ﻫﺎﺳﺖ ‪.‬‬

‫‪6‬‬
 ‫ﺍﺟﺰﺍﻱ ﺳﻴﺴﺘﻢ ﻣﺪﻳﺮﻳﺖ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ‪:‬‬

‫ﻭﻇﺎﻳﻒ ‪ DBMS‬ﺗﻮﺳﻂ ﺗﻌﺪﺍﺩﻱ ﻣﻮﻟﻔﻪ ﻧﺮﻡ ﺍﻓﺰﺍﺭﻱ ﺍﻧﺠﺎﻡ ﻣﻲ ﺷﻮﺩ ‪ .‬ﻫﺮ ﻛﺪﺍﻡ ﺍﺯ ﺍﻳﻦ ﻣﻮﻟﻔﻪ ﻫﺎ ﻣﻤﻜﻦ‬
‫ﺍﺳﺖ ﻣﺮﻛﺐ ﺍﺯ ﭼﻨﺪ ﻭﺍﺣﺪ ﻛﻮﭼﻜﺘﺮ ﺑﺎﺷﻨﺪ ‪ .‬ﺗﻌﺪﺍﺩﻱ ﺍﺯ ﺳﺮﻭﻳﺲ ﻫﺎﻱ ﻛﻪ ﺗﻮﺳﻂ ﻣﻮﻟﻔﻪ ﻫﺎﻱ ‪DBMS‬‬
‫ﺩﺍﺩﻩ ﻣﻲ ﺷﻮﻧﺪ ﺩﺭ ﺯﻳﺮ ﻟﻴﺴﺖ ﺷﺪﻩ ﺍﺳﺖ‪:‬‬

‫‪ .1‬ﭘﺮﺩﺍﺯﺵ ﺗﺮﺍﻛﻨﺶ )‪(Transaction Processing‬‬

‫ﭘﺮﺩﺍﺯﺵ ﺗﺮﺍﻛﻨﺶ ﻋﻤﻠﻴﺎﺗﻲ ﻛﻪ ﺍﺯ ﻣﻨﺎﺑﻊ ﻣﺨﺘﻠﻒ ﻣﻲ ﺭﺳﺪ ﺭﺍ ﺭﻭﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺍﺟﺮﺍ ﻣﻲ ﻛﻨﺪ ﺑﻪ ﻧﺤﻮﻱ‬
‫ﻛﻪ ﺧﻮﺍﺹ ﻣﻄﻠﻮﺏ ﺗﺮﺍﻛﻨﺶ ﺧﺪﺷﻪ ﺩﺍﺭ ﻧﺸﻮﺩ ‪ .‬ﺳﺮﻭﻳﺲ ﻫﺎﻱ ﻛﻨﺘﺮﻝ ﻫﻤﺮﻭﻧﺪﻱ ﻭ ﺗﺮﻣﻴﻢ ﺑﻪ ﺍﻳﻦ ﻣﻮﻟﻔﻪ‬
‫ﺑﺮﺍﻱ ﺑﺮﻗﺮﺍﺭﻱ ﺧﻮﺍﺹ ‪ ACID‬ﻛﻤﻚ ﻣﻲ ﻛﻨﻨﺪ ‪ .‬ﺑﻪ ﺍﻳﻦ ﺗﺮﺗﻴﺐ ﺍﺟﺮﺍﻱ ﻫﻤﺮﻭﻧﺪ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﻭ‬
‫ﺳﺎﺯﮔﺎﺭﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺣﺘﻲ ﺩﺭ ﺻﻮﺭﺕ ﻭﻗﻮﻉ ﺷﻜﺴﺘﻲ ﺩﺭ ﺳﻴﺴﺘﻢ ﺗﻀﻤﻴﻦ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫‪ .2‬ﻛﻨﺘﺮﻝ ﻫﻤﺮﻭﻧﺪﻱ )‪(Concurrency Control‬‬

‫ﻣﺪﻳﺮﻳﺖ ﺍﺟﺮﺍﻱ ﻫﻤﺮﻭﻧﺪ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺭﻭﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺩﺭ ﺣﻴﻦ ﺑﺮﻗﺮﺍﺭﻱ ﺳﺎﺯﮔﺎﺭﻱ ﺭﺍ ﺑﻪ ﻋﻬﺪﻩ ﺩﺍﺭﺩ ‪.‬‬

‫‪ .3‬ﺗﺮﻣﻴﻢ )‪(Recovery‬‬

‫ﺗﺮﻣﻴﻢ ﺗﻀﻤﻴﻦ ﻣﻲ ﻛﻨﺪ ﻛﻪ ﺍﮔﺮ ﺍﺟﺮﺍﻱ ﺗﺮﺍﻛﻨﺶ ﺑﺎ ﻋﺪﻡ ﻣﻮﻓﻘﻴﺖ ﻳﺎ ﺳﻖ ﻁ ﺭﻭﺑﺮﻭ ﺷﺪ‪ ،‬ﺗﺎﺛﻴﺮ ﻧﺎﻣﻄﻠﻮﺑﻲ‬
‫ﺭﻭﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻳﺎ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﺩﻳﮕﺮ ﻧﮕﺬﺍﺭﺩ ﻭ ﺣﺎﻟﺖ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺭﺍ ﻫﻤﻴﺸﻪ ﺳﺎﺯﮔﺎﺭ ﻧﮕﻪ ﺩﺍﺭﺩ ‪.‬‬

‫‪ .4‬ﻣﺪﻳﺮﻳﺖ ﺛﺒﺖ ﺍﺣﻮﺍﻝ)‪(Log Management‬‬

‫ﻫﺮ ﺍﺗﻔﺎﻗﻲ ﺩﺭ ﺳﻴﺴﺘﻢ ﺩﺭ ﻳﻚ ﻓﺎﻳﻞ ﺫﺧﻴﺮﻩ ﻣﻲ ﺷﻮﺩ ﻭ ﺗﻮﺳﻂ ﻣﺪﻳﺮﻳﺖ ﺗﺮﻣﻴﻢ ﺑﺮﺍﻱ ﺣﻔﻆ ﺻﺤﺖ ﻭ‬
‫ﺍﻋﺘﺒﺎﺭ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻫﻨﮕﺎﻡ ﺧﺮﺍﺑﻲ ﻳﺎ ﺳﻘﻂ ﺳﻴﺴﺘﻢ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪7‬‬
 ‫‪ .5‬ﻭﺍﺳﻄﻪ ﺯﺑﺎﻧﻲ )‪(Language Interface‬‬

‫ﺩﺳﺘﻮﺭﺍﺗﻲ ﺭﺍ ﺑﺮﺍﻱ ﺗﻌﺮﻳﻒ ﺩﺍﺩﻩ‪ ،‬ﻛﺎﺭ ﻛﺮﺩﻥ ﺑﺎ ﺁﻥ ﺩﺭﺍﺧﺘﻴﺎﺭ ﻛﺎﺭﺑﺮﺍﻥ ﻭ ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﻛﺎﺭﺑﺮﺩﻱ ﻗﺮﺍﺭ ﻣﻲ‬
‫ﺩﻫﺪ‪.‬‬

‫‪ .6‬ﺗﺤﻤﻞ ﭘﺬﻳﺮﻱ ﺧﻄﺎ)‪(Fault Tolerance‬‬

‫ﺗﻮﺍﻧﺎﺋﻲ ﺍﺭﺍﺋﻪ ﺳﺮﻭﻳﺲ ﻫﺎﻱ ﻗﺎﺑﻞ ﺍﻃﻤﻴﻨﺎﻥ ﺗﻮﺳﻂ ‪ DBMS‬ﺣﺘﻲ ﺩﺭ ﺻﻮﺭﺕ ﺑﺮﻭﺯ ﻧﻘﺺ ﺭﺍ ﺗﺤﻤﻞ‬
‫ﭘﺬﻳﺮﻱ ﺧﻄﺎ ﻣﻲ ﮔﻮﻳﻨﺪ‪ .‬ﺍﻧﻮﺍﻉ ﺧﻄﺎﻫﺎﺋﻲ ﻛﻪ ﻣﻤﻜﻦ ﺍﺳﺖ ﭘﻴﺶ ﺑﻴﺎﻳﺪ ﻋﺒﺎﺭﺗﻨﺪ ﺍﺯ ‪:‬‬

‫• ﺧﻄﺎﻱ ﻣﻨﻄﻘﻲ‪ :‬ﺗﺮﺍﻛﻨﺶ ﻣﻮﻓﻖ ﻧﻤﻲ ﺷﻮﺩ ﻣﺜﻼ ﺑﻪ ﺩﻟﻴﻞ ﻭﺭﻭﺩﻱ ﺑﺪ‪ ،‬ﺳﺮﺭﻳﺰﻱ‬
‫• ﺧﻄﺎﻱ ﺳﻴﺴﺘﻤﻲ‪ :‬ﺗﺮﺍﻛﻨﺶ ﻣﻮﻓﻖ ﻧﻤﻲ ﺷﻮﺩ ﻣﺜﻼ ﺑﻪ ﺩﻟﻴﻞ ﺑﻦ ﺑﺴﺖ‬
‫• ﺳﻘﻂ ﺳﻴﺴﺘﻢ‪ :‬ﻗﻄﻊ ﺑﺮﻕ‪ ،‬ﭘﺎﻙ ﺷﺪﻥ ﺣﺎﻓﻈﻪ ﺍﺻﻞﻱ‪ ،‬ﭘﺮ ﺷﺪﻥ ﺩﻳﺴﻚ‬
‫• ﻧﺎﺗﻮﺍﻧﻲ ﺩﻳﺴﻚ ‪ :‬ﺧﺮﺍﺑﻲ ﻫﺪ‪ ،‬ﺧﺮﺍﺑﻜﺎﺭﻱ ﻋﻤﺪﻱ‪ ،‬ﺁﺗﺶ ﺳﻮﺯﻱ‪،‬‬

‫‪ .7‬ﻛﺎﺗﺎﻟﻮگ ﺩﺍﺩﻩ )‪(Data Catalog‬‬

‫ﻳﺎ ﺩﻳﻜﺸﻨﺮﻱ ﺩﺍﺩﻩ ﻳﻚ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺳﻴﺴﺘﻤﻲ ﺷﺎﻣﻞ ﺍﻃﻼﻋﺎﺗﻲ ﺩﺭﺑﺎﺭﻩ ﺩﺍﺩﻩ‪ ،‬ﺍﺭﺗﺒﺎﻃﺎﺕ ﻭ ﻗﻴﺪﻫﺎ ﺩﺭ‬
‫ﭘﺎﻳﮕﺎﻩ ﺍﺻﻠﻲ ﺍﺳﺖ‪ .‬ﮔﺎﻫﻲ ﺑﻪ ﺁﻥ ﻣﺘﺎﺩﺍﺩﻩ ﻫﻢ ﮔﻔﺘﻪ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫‪ .8‬ﺍﻣﻨﻴﺖ )‪(Security‬‬

‫ﺍﻣﻨﻴﺖ ﺑﻪ ﻣﺤﺎﻓﻈﺖ ﺩﺍﺩﻩ ﺩﺭ ﻣﻘﺎﺑﻞ ﺍﻓﺸﺎﺷﺪﻥ‪ ،‬ﺗﻐﻴﻴﺮ ﻭ ﺧﺮﺍﺑﻲ ﺍﺷﺎﺭﻩ ﺩﺍﺭﺩ ‪ .‬ﻫﺮ ﻛﺎﺑﺮ ﻭ ﺑﺮﻧﺎﻣﻪ ﻛﺎﺭﺑﺮﺩﻱ‬
‫ﺍﻣﺘﻴﺎﺯ ﻭﻳﮋﻩ ﺍﻱ ﺑﺮﺍﻱ ﺩﺳﺘﺮﺳﻲ ﺑﻪ ﺩﺍﺩﻩ ﺩﺍﺭﺩ ‪ .‬ﻛﺎﺭﺑﺮﺍﻥ ﻣﻤﻜﻦ ﺍﺳﺖ ﺩﻳﺪﮔﺎﻩ ﻫﺎﻱ ﻣﺨﺘﻠﻔﻲ ﻧﺴﺒﺖ ﺑﻪ‬
‫ﺩﺍﺩﻩ ﻫﺎﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺑﺎ ﺗﻮﺟﻪ ﺍﻣﺘﻴﺎﺯﺍﺕ ﻭﻳﮋﻩ ﺧﻮﺩ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻨﺪ ‪ .‬ﺳﻴﺴﺘﻢ ﺍﻣﻨﻴﺘﻲ ﻫﻤﭽﻨﻴﻦ‪ ،‬ﺗﻮﺳﻂ‬
‫ﺭﻭﻳﻪ ﻫﺎﻱ ﺷﻨﺎﺳﺎﺋﻲ ﻭ ﻣﺠﻮﺯ‪ ،‬ﺩﺳﺘﺮﺳﻲ ﺑﻪ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺭﺍ ﻣﺤﺪﻭﺩ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫‪8‬‬
 ‫‪ .9‬ﻣﺪﻳﺮﻳﺖ ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ )‪(Storage Management‬‬

‫‪ DBMS‬ﻣﻜﺎﻧﻴﺴﻢ ﻫﺎﻱ ﺧﺎﺻﻲ ﺑﺮﺍﻱ ﺫﺧﻴﺮﻩ ﺩﺍﺋﻤﻲ ﺩﺍﺩﻩ ﻭ ﺩﺳﺘﺮﺳﻲ ﺑﻪ ﻣﻨﺒﻊ ﻓﻴﺰﻳﻜﻲ ﻭ ﺑﺎﺯﻳﺎﺑﻲ ﺩﺍﺩﻩ‬
‫ﺩﺍﺭﺩ‪ .‬ﻣﺪﻳﺮ ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ ﺑﻴﻦ ﺩﺍﺩﻩ ﺫﺧﻴﺮﻩ ﺷﺪﻩ ﺩﺭ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻭ ﺑﺮﻥ ﺍﻣﻪ ﻛﺎﺭﺑﺮﺩﻱ ﻭ ﭘﺮﺱ ﻭ ﺟﻮﻫﺎﻱ‬
‫ﺍﺭﺳﺎﻝ ﺷﺪﻩ ﺑﻪ ﺳﻴﺴﺘﻢ ﻭﺍﺳﻄﻪ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪ .10‬ﻣﺪﻳﺮﻳﺖ ﻗﻔﻞ )‪(Lock Management‬‬

‫)ﻣﺜﻞ ‪Read‬‬ ‫ﻫﻨﮕﺎﻡ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺷﺘﺮﺍﻛﻲ ﺍﺯ ﺩﺍﺩﻩ ﺍﻧﻮﺍﻉ ﻣﺨﺘﻠﻔﻲ ﺍﺯ ﻗﻔﻞ ﺭﻭﻱ ﺩﺍﺩﻩ ﮔﺬﺍﺷﺘﻪ ﻣﻲ ﺷﻮﺩ‬
‫‪ Lock‬ﻭ ‪.(Write Lock‬‬

‫‪ .11‬ﻣﺪﻳﺮﻳﺖ ﺑﻦ ﺑﺴﺖ )‪(Deadlock Management‬‬

‫ﺑﻦ ﺑﺴﺖ ﻭﻗﺘﻲ ﺍﺗﻔﺎﻕ ﻣﻲ ﺍﻓﺘﺪ ﻛﻪ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺑﺮﺍﻱ ﺑﺪﺳﺖ ﺁﻭﺭﺩﻥ ﻣﻨﺎﺑﻊ ﺩﺭ ﻳﻚ ﺩﺍﻳﺮﻩ ﺑﺴﺘﻪ ﻗﺮﺍﺭ‬
‫ﮔﻴﺮﻧﺪ ﻳﻌﻨﻲ ﻫﺮ ﻳﻚ ﻣﻨﺒﻌﻲ ﺩﺭ ﺍﺧﺘﻴﺎﺭ ﺩﺍﺭﺩ ﻛﻪ ﻣﻮﺭﺩ ﺗﻘﺎﺿﺎﻱ ﺩﻳﮕﺮﻱ ﺍﺳﺖ ﻭ ﺩﺭﺧﻮﺍﺳﺖ ﻣﻨﺒﻌﻲ ﺭﺍ‬
‫ﻣﻲ ﻛﻨﺪ ﻛﻪ ﺩﺭ ﺍﺧﺘﻴﺎﺭ ﺗﺮﺍﻛﻨﺶ ﻣﻨﺘﻈﺮ ﻣﻨﺒﻊ ﺍﺳﺖ ‪ .‬ﺩﺭ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻣﻨﺎﺑﻊ ﺭﻛﻮﺭﺩﻫﺎ ﻫﺴﺘﻨﺪ ‪ .‬ﻣﺪﻳﺮﻳﺖ‬
‫ﻣﻨﺒﻊ ﻣﺴﺌﻮﻝ ﺭﻓﻊ ﺍﻳﻦ ﻣﺸﻜﻞ ﻫﺴﺘﻨﺪ‪.‬‬

‫ﻣﻌﻤﺎﺭﻱ ‪:ANSI-SPARC‬‬
‫ﻫﺮ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺑﺎ ﺗﺒﻌﻴﺖ ﺍﺯ ﻳﻚ ﻧﻮﻉ ﻣﻌﻤﺎﺭﻱ ﻃﺮﺍﺣﻲ ﻣﻲ ﮔﺮﺩﺩ‪ .‬ﻃﺒﻖ ﺍﻳﻦ ﻧﻮﻉ ﻣﻌﻤﺎﺭﻱ ﺑﺎﻧﻚ‬
‫ﺍﻃﻼﻋﺎﺗﻲ ﺩﺍﺭﺍﻱ ﺳﻪ ﺳﻄﺢ ﺍﺳﺖ‪.‬‬
‫ﺳﻪ ﺳﻄﺢ ﻣﻌﻤﺎﺭﻱ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ‪:‬‬
‫‪ ‬ﺳﻄﺢ ﺩﺍﺧﻠﻲ ‪ :‬ﻛﻪ ﺳﻄﺢ ﻓﻴﺰﻳﻜﻲ ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﺩ ﻭ ﺑﻪ ﺣﺎﻓﻈﻪ ﻓﻴﺰﻳﻜﻲ ﻧﺰﺩﻳﻚ ﺍﺳﺖ‬
‫ﻳﻌﻨﻲ ﻣﺮﺑﻮﻁ ﺑﻪ ﺫﺧﻴﺮﻩ ﻓﻴﺰﻳﻜﻲ ﻣﻴﺸﻮﺩ‪.‬‬
‫‪ ‬ﺳﻄﺢ ﺧﺎﺭﺟﻲ ‪ :‬ﻛﻪ ﺳﻄﺢ ﻣﻨﻄﻘﻲ ﻛﺎﺭﺑﺮ ﻧﻴﺰ ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﺩ ﻭ ﺑﻪ ﻛﺎﺭﺑﺮ ﻧﺰﺩﻳﻚ ﺍﺳﺖ ﻭ‬
‫ﺑﺎ ﻣﺸﺎﻫﺪﻩ ﺩﺍﺩﻩ ﻫﺎ ﺗﻮﺳﻂ ﻛﺎﺭﺑﺮﺍﻥ ﺳﺮﻭﻛﺎﺭ ﺩﺍﺭﺩ‪.‬‬

‫‪9‬‬
‫‪ ‬ﺳﻄﺢ ﺍﺩﺭﺍﻛﻲ‪ :‬ﻛﻪ ﺳﻄﺢ ﻣﻨﻄﻘﻲ ﺍﺟﺘﻤﺎﻉ ﺍﺳﺖ ﻳﻌﻨﻲ ﺳﻄﺢ ﻏﻴﺮﻣﺴﺘﻘﻴﻢ ﺑﻴﻦ ﺩﻭ ﺳﻄﺢ‬
‫ﺩﻳﮕﺮ ﺍﺳﺖ‪.‬‬

‫ﺑﺮﺭﺳﻲ ﺩﻳﺪﮔﺎﻩ ﺧﺎﺭﺟﻲ ‪:‬‬

‫ﺳﻄﺢ ﺧﺎﺭﺟﻲ ‪ ،‬ﺳﻂﺡ ﻫﺮﻛﺎﺭﺑﺮ ﺍﺯﺟﻤﻠﻪ ﻣﺤﺼﻮﻻﺕ ﺧﺎﺹ ‪ SQL‬ﻣﺜﻞ‪ Oracle‬ﺍﺳﺖ ﻛﺎﺭﺑﺮ ﻣﻲ ﺗﻮﺍﻧﺪ‬
‫ﻳﻚ ﺑﺮﻧﺎﻣﻪ ﻧﻮﻳﺲ ﻛﺎﺭﺑﺮﺩﻱ ﻳﺎ ﻳﻚ ﻛﺎﺭﺑﺮ ﻧﻬﺎﻳﻲ ﺑﺎ ﻫﺮ ﺩﺭﺟﻪ ﺍﻱ ﺍﺯ ﺗﺨﺼﺺ ﺑﺎﺷﺪ )ﻣﺪﻳﺮ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ‬
‫‪ ،‬ﻓﺮﺩ ﺧﺎﺻﻲ ﺍﺳﺖ ﻭ ﺑﺎﻳﺪ ﺑﻪ ﺳﻄﻮﺡ ﺩﺍﺧﻠﻲ ﻭ ﺍﺩﺭﺍﻛﻲ ﻋﻼﻗﻪ ﻣﻨﺪ ﺑﺎﺷﺪ(‪.‬‬
‫ﻫﺮﻛﺎﺭﺑﺮ ﺩﺍﺭﺍﻱ ﻳﻚ ﺯﺑﺎﻥ ﺍﺳﺖ‪:‬‬
‫ﺑﺮﺍﻱ ﺑﺮﻧﺎﻣﻪ ﻧﻮﻳﺲﺍﻥ ﻛﺎﺭﺑﺮﺩﻱ ‪ ،‬ﺍﻳﻦ ﺯﺑﺎﻥ ﻣﻤﻜﻦ ﺍﺳﺖ ﻳﻚ ﺯﺑﺎﻥ ﺑﺮﻧﺎﻣﻪ ﻧﻮﻳﺴﻲ )ﻣﺜﻞ ‪ ،C++‬ﺟﺎﻭﺍ ﻳﺎ‬
‫‪ ( PL/I‬ﺑﺎﺷﺪ ﻳﺎ ﻳﻚ ﺯﺑﺎﻥ ﺍﺧﺘﺼﺎﺻﻲ ﺑﺎ ﺷﺪ ﻛﻪ ﻣﺮﺑﻮﻁ ﺑﻪ ﺧﻮﺩ ﺁﻥ ﺳﻴﺴﺘﻢ ﺍﺳﺖ ‪ .‬ﺍﻳﻦ ﺯﺑﺎﻥ ﻫﺎﻱ‬
‫ﺍﺧﺘﺼﺎﺻﻲ ﺭﺍ ﻣﻌﻢ ﻭﻻ ﺯﺑﺎﻥ ﻫﺎﻱ ﻧﺴﻞ ﭼﻬﺎﺭﻡ ﻣﻲ ﮔﻮﻳﻨﺪ ‪ .‬ﺯﺑﺎﻥ ﻣﺎﺷﻴﻦ ‪ ،‬ﺯﺑﺎﻥ ﺍﺳﻤﺒﻠﻲ ﻭ ﺯﺑﺎﻥ ﻫﺎﻳﻲ‬
‫ﻣﺜﻞ ‪ C++‬ﺭﺍ ﻣﻲ ﺗﻮﺍﻥ ﺟﺰء ﺯﺑﺎﻥ ﻫﺎﻱ ﺳﻪ ﻧﺴﻞ ﻗﺒﻠﻲ ﺩﺍﻧﺴﺖ ‪ .‬ﺯﺑﺎﻥ ﻫﺎﻱ ﻧﺴﻞ ﭼﻬﺎﺭﻡ ﺑﺎ ﺗﻐﻴﻴﺮﺍﺗﻲ ﺑﺮ‬
‫ﺭﻭﻱ ﺯﺑﺎﻥ ﻫﺎﻱ ‪ 3‬ﻧﺴﻞ ﻗﺒﻠﻲ ﺍﻳﺠﺎﺩ ﻣﻲ ﺷﻮﻧﺪ ‪.‬‬
‫ﺯﺑﺎﻥ ﻛﺎﺭﺑﺮ ﻧﻬﺎﻳﻲ ‪ ،‬ﺯﺑﺎﻥ ﺗﻘﺎﺿﺎ ﻳﺎ ﺯﺑﺎﻥ ﺧﺎﺻﻲ ﻣﺜﻞ ﻣﻨﻮﮔﺮﺍ ﻭ ﻓﺮﻡ ﮔﺮﺍ ﺍﺳﺖ ﻛﻪ ﺧﻮﺍﺳﺘﻪ ﻫﺎﻱ ﻛﺎﺭﺑﺮﺍﻥ‬
‫ﺭﺍ ﺑﺮﺁﻭﺭﺩﻩ ﻣﻲ ﻛﻨﺪ ﻭ ﺗﻮﺳﻂ ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﻛﺎﺭﺑﺮﺩﻱ ﺁﻧﻼﻳﻦ ﭘﺸﺘﻴﺒﺎﻧﻲ ﻣﻲ ﺷﻮﺩ ‪.‬‬
‫ﺑﺨﺸﻲ ﺍﺯ ﺍﻳﻦ ﺯﺑﺎﻥ ﻫﺎ ﻛﻪ ﺑﺎ ﺍﺷﻴﺎ ﻭ ﻋﻤﻠﻴﺎﺕ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺳﺮﻭﻛﺎﺭ ﺩﺍﺭﺩ‪ ،‬ﻣﻬﻢ ﺍﺳﺖ ‪ .‬ﺍﻳﻦ ﺑﺨﺶ ﺭﺍ‬
‫‪ Data Sub Language‬ﻳﺎ ‪ DSL‬ﺩﺭ ﺯﺑﺎﻥ ﻣﻴﺰﺑﺎﻥ ﻣﺘﻨﺎﻇﺮ ﺗﻌﺒﻴﻪ ﺷﺪﻩ ﺍﺳﺖ ‪ .‬ﺯﺑﺎﻥ ﻣﻴﺰﺑﺎﻥ ﻣﺴﺌﻮﻝ‬
‫ﺗﻬﻴﻪ ﺍﻣﻜﺎﻧﺎﺕ ﮔﻮﻧﺎﮔﻮﻥ ﻏﻴﺮ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺍﺳﺖ‪ ،‬ﻣﺜﻞ ﻣﺘﻐﻴﺮﻫﺎﻱ ﻣﺤﻠﻲ‪ ،‬ﺍﻋﻤﺎﻝ ﻣﺤﺎﺳﺒﺎﺗﻲ‪ ،‬ﻣﻨﻄﻖ‬
‫ﺗﺼﻤﻴﻢ ﮔﻴﺮﻱ ﻭ ﻏﻴﺮﻩ ‪ .‬ﻳﻚ ﺳﻴﺴﺘﻢ ﻣﻤﻜﻦ ﺍﺳﺖ ﻫﺮﺗﻌﺪﺍﺩﻱ ﺍﺯ ﺯﺑﺎﻥ ﻫﺎﻱ ﻣﻴﺰﺑﺎﻥ ﻭ ﺗﻌﺪﺍﺩﻱ ﺍﺯ ‪DSL‬‬
‫ﺭﺍ ﭘﺸﺘﻴﺒﺎﻧﻲ ﻛﻨﺪ ‪ .‬ﺍﻣﺎ ﻳﻚ ‪ DSL‬ﺧﺎﺹ ﻛﻪ ﺗﻮﺳﻂ ﺍﻏﻠﺐ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﻓﻌﻠﻲ ﭘﺸﺘﻴﺒﺎﻧﻲ ﻣﻲ ﺷﻮﺩ‬
‫‪ SQL‬ﺍﺳﺖ‪ .‬ﺩﺭ ﺍﻏﻠﺐ ﺍﻳﻦ ﺳﻴﺴﺘﻢ ﻫﺎ ﻣﻲ ﺗﻮﺍﻥ ﺍﺯ ‪ SQL‬ﻫﻢ ﺑﻪ ﺻﻮﺭﺕ ﻣﺤﺎﻭﺭﻩ ﺍﻱ ﻭ ﺑﻪ ﻋﻨﻮﺍﻥ ﻳﻚ‬
‫ﺯﺑﺎﻥ ﻣﺴﺘﻘﻞ ﺗﻘﺎﺿﺎ ﺍﺳﺘﻔﺎﺩﻩ ﻛﺮﺩ ﻭ ﻫﻢ ﺑﻪ ﻋﻨﻮﺍﻥ ﺯﺑﺎﻥ ﺗﻌﺒﻴﻪ ﺷﺪﻩ ﺩﺭ ﺯﺑﺎﻥ ﻫﺎﻳﻲ ﻣﺜﻞ ﺟﺎﻭﺍ ﺍﺳﺘﻔﺎﺩﻩ‬
‫ﻧﻤﻮﺩ‪.‬‬
‫ﻫﺮ ‪ DSL‬ﺗﺮﻛﻴﺒﻲ ﺍﺯ ﺩﻭ ﺯﺑﺎﻥ ﺩﻳﮕﺮ ﺍﺳﺖ ‪ :‬ﺯﺑﺎﻥ ﺗﻌﺮﻳﻒ ﺩﺍﺩﻩ ﻫﺎ )‪ (DDL‬ﻛﻪ ﺗﻌﺮﻳﻒ ﻳﺎ ﺍﻋﻼﻥ ﺍﺷﻴﺎﻱ‬
‫ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺭﺍ ﭘﺸﺘﻴﺒﺎﻧﻲ ﻣﻲ ﻛﻨﺪ ﻭ ﺯﺑﺎﻥ ﺩﺳﺘﻜﺎﺭﻱ ﺩﺍﺩﻩ ﻫﺎ )‪ (DML‬ﻛﻪ ﭘﺮﺩﺍﺯﺵ ﺍﻳﻦ ﺍﺷﻴﺎ ﺭﺍ‬
‫ﺍﻣﻜﺎﻥ ﭘﺬﻳﺮ ﻣﻲ ﺳﺎﺯﺩ ‪ DSL .‬ﺍﻳﻦ ﻛﺎﺭﺑﺮ ﻣﺘﺸﻜﻞ ﺍﺯ ﻭﻳﮋﮔﻲ ﻫﺎﻳﻲ ﺍﺯ ‪ PL/I‬ﺍﺳﺖ ﻛﻪ ﺑﺮﺍﻱ ﺑﺮﻗﺮﺍﺭﻱ‬
‫ﺍﺭﺗﺒﺎﻁ ﺑﺎ ﺳﻴﺴﺘﻢ ﻣﺪﻳﺮﻳﺖ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺑﻪ ﻛﺎﺭ ﻣﻲ ﺭﻭﺩ‪:‬‬

‫‪10‬‬
 ‫ﺑﺨﺶ‪ DDL‬ﺣﺎﻭﻱ ﺳﺎﺧﺘﺎﺭﻫﺎﻱ ﺍﻋﻼﻧﻲ ‪ PL/I‬ﺍﺳﺖ ﻛﻪ ﺑﺮﺍﻱ ﺗﻌﺮﻳﻒ ﺍﺷﻴﺎﻱ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺿﺮﻭﺭﻱ‬
‫ﺍﺳﺖ – ﺩﺳﺘﻮﺭ ‪ ) DECLARE‬ﻳﺎ ‪ ،( DCL‬ﺑﻌﻀﻲ ﺍﺯ ﺍﻧﻮﺍﻉ ﺩﺍﺩﻩ ‪ PL/I‬ﻳﺎ ﺗﻮﺳﻌﻪ ﺍﻱ ﺍﺯ ‪ PL/I‬ﻛﻪ‬
‫ﺩﺭ ‪ PL/I‬ﻓﻌﻠﻲ ﻣﻮﺟﻮﺩ ﻧﻴﺴﺖ‪.‬‬
‫ﺑﺨﺶ ‪ DML‬ﺷﺎﻣﻞ ﺩﺳﺘﻮﺭﺍﺕ ﺍﺟﺮﺍﻳﻲ ‪ PL/I‬ﺍﺳﺖ ﻛﻪ ﺍﻃﻼﻋﺎﺕ ﺭﺍ ﺑﻪ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺍﻥﺗﻘﺎﻝ ﺩﺍﺩﻩ‪ ،‬ﺍﺯ‬
‫ﺁﻥ ﺑﺎﺯﻳﺎﺑﻲ ﻣﻲ ﻛﻨﺪ‪ .‬ﻣﻤﻜﻦ ﺍﺳﺖ ﺩﺳﺘﻮﺭﺍﺕ ﺧﺎﺻﻲ ﮔﻨﺠﺎﻧﺪﻩ ﺷﻮﻧﺪ ‪.‬‬
‫‪ PL/I‬ﺩﺭﺣﺎﻝ ﺣﺎﺿﺮ ﻓﺎﻗﺪ ﻭﻳﮋﮔﻲ ﻫﺎﻱ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺍﺳﺖ ‪ .‬ﺩﺳﺘﻮﺭﺍﺕ‪ DML‬ﻫﻤﺎﻥ ﺩﺳﺘﻮﺭ‬
‫‪ CALL‬ﺍﺳﺖ ﻛﻪ ﺳﻴﺴﺘﻢ ﻣﺪﻳﺮﻳﺖ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺭﺍ ﻓﺮﺍﺧﻮﺍﻧﻲ ﻣﻲ ﻛﻨﺪ )ﮔﺮﭼﻪ ﻣﻤﻜﻦ ﺍﺳﺖ ﺍﻳﻦ‬
‫ﺩﺳﺘﻮﺭ ﻓﺮﺍﺧﻮﺍﻧﻲ ﻃﻮﺭﻱ ﺗﻐﻴﻴﺮ ﻛﻨﺪ ﻛﻪ ﺍﺯ ﻧﻈﺮ ﻛﺎﺭﺑﺮ ﻗﺎﺑﻞ ﻗﺒﻮﻝ ﻭ ﺭﺍﺣﺖ ﺑﺎﺷﺪ (‪.‬‬
‫ﺗﺎ ﻛﻨﻮﻥ ﻣﺘﻮﺟﻪ ﺷﺪﻳﻢ ﻛﻪ ﻫﺮ ﻛﺎﺭﺑﺮ ﻓﻘﻂ ﺑﻪ ﺑﺨﺸﻲ ﺍﺯ ﻛﻞ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﻋﻼﻗﻤﻨﺪ ﺍﺳﺖ ‪ .‬ﺩﻳﺪﮔﺎﻩ‬
‫ﻛﺎﺭﺑﺮ ﺑﻪ ﺁﻥ ﺑﺨﺶ‪ ،‬ﺩﺭ ﻣﻘﺎﻳﺴﻪ ﺑﺎ ﺭﻭﺵ ﺫﺧﻴﺮﻩ ﻓﻴﺰﻳﻜﻲ ﺩﺍﺩﻩ ﻫﺎ‪ ،‬ﺗﺎ ﺣﺪﻱ ﺍﻧﺘﺰﺍﻋﻲ ﺍﺳﺖ ‪ .‬ﺩﺭ ﺍﺻﻄﻼﺡ‬
‫‪ ،ANSI/SPARC‬ﺩﻱﺩﮔﺎﻩ ﻛﺎﺭﺑﺮ‪ ،‬ﺩﻳﺪﮔﺎﻩ ﺧﺎﺭﺟﻲ ﻧﺎﻡ ﺩﺍﺭﺩ‪ .‬ﻟﺬﺍ ﺩﻳﺪﮔﺎﻩ ﺧﺎﺭﺟﻲ ﻫﻤﺎﻥ ﻣﺤﺘﻮﻳﺎﺕ‬
‫) ﻳﻌﻨﻲ ‪ ،‬ﺍﺯ ﻧﻈﺮ ﺁﻥ ﻛﺎﺭﺑﺮ‪ ،‬ﺩﻳﺪﮔﺎﻩ‬ ‫ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺍﺳﺖ ﻛﻪ ﺗﻮﺳﻂ ﻛﺎﺭﺑﺮﺍﻥ ﻗﺎﺑﻞ ﻣﺸﺎﻫﺪﻩ ﺍﺳﺖ‬
‫ﺧﺎﺭﺟﻲ ‪ ،‬ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺍﺳﺖ (‪ .‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ ‪ ،‬ﻛﺎﺭﺑﺮﻱ ﺍﺯ ﺑﺨﺶ ﭘﺮﺳﻨﻠﻲ ﻣﻤﻜﻦ ﺍﺳﺖ ﺑﺎﻧﻚ‬
‫ﺍﻃﻼﻋﺎﺗﻲ ﺭﺍ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺭﻭﻱﺩﺍﺩﻫﺎﻱ ﺭﻛﻮﺭﺩ ﻛﺎﺭﻛﻨﺎﻥ ﻭ ﻣﺤﻞ ﻛﺎﺭ ﺁﻧﺎﻥ ﺑﺪﺍﻧﺪ ﻭ ﺍﺯ ﺭﻭﻳﺪﺍﺩﻫﺎﻱ ﺭﻛﻮﺭﺩ‬
‫ﻗﻄﻌﺎﺕ ﻭ ﻋﺮﺿﻪ ﻛﻨﻨﺪﻩ ﻫﺎ ﻛﻪ ﺗﻮﺳﻂ ﻛﺎﺭﺑﺮﺍﻥ ﻓﺮﻭﺵ ﺩﺳﺘﻴﺎﺑﻲ ﻣﻲ ﺷﻮﺩ‪ ،‬ﺑﻲ ﺧﺒﺮ ﺑﺎﺷﺪ ‪.‬‬
‫ﺑﻪ ﻃﻮﺭ ﻛﻠﻲ‪ ،‬ﺩﻳﺪﮔﺎﻩ ﺧﺎﺭﺟﻲ ﺷﺎﻣﻞ ﺭﻭﻳﺪﺍﺩﻫﺎﻱ ﻣﺘﻌﺪﺩﻱ ﺍﺯ ﻫﺮﻳﻚ ﺍﺯ ﺍﻧﻮﺍﻉ ﺭﻛﻮﺭﺩ ﺧﺎﺭﺟﻲ ﺍﺳﺖ )ﻛﻪ‬
‫ﺍﻟﺰﺍﻣﺎ ﻫﻤﺎﻥ ﺭﻛﻮﺭﺩ ﺫﺧﻴﺮﻩ ﺷﺪﻩ ﻧﻴﺴﺖ (‪ .‬ﻓﺮﺽ ﻣﻴﻜﻨﻴﻢ ﺗﻤﺎﻡ ﺍﻃﻼﻋﺎﺗﻲ ﻛﻪ ﺩﺭ ﺳﻄﺢ ﺧﺎﺭﺟﻲ ﻧﻤﺎﻳﺶ‬
‫ﺩﺍﺩﻩ ﻣﻲ ﺷﻮﻧﺪ ﺑﻪ ﺷﻜﻞ ﺭﻛﻮﺭﺩﻫﺎﺳﺖ ‪ .‬ﺩﺭ ﺣﺎﻟﻴﻜﻪ ﺩﺭ ﺑﻌﻀﻲ ﺍﺯ ﺳﻴﺴﺘﻢ ﻫﺎ ﻣﻤﻜﻦ ﺍﺳﺖ ﺍﻃﻼﻋﺎﺕ ﺑﻪ‬
‫ﺷﻜﻞ ﺭﻛﻮﺭﺩ ﺫﺧﻴﺮﻩ ﻧﺸﻮﺩ ‪ DSL .‬ﻛﺎﺭﺑﺮ‪ ،‬ﺑﺮﺍﺳﺎﺱ ﺭﻛﻮﺭﺩﻫﺎﻱ ﺧﺎﺭﺟﻲ ﺗﻌﺮﻳﻒ ﺷﺪﻩ ﺍﺳﺖ ‪ :‬ﺑﻪ ﻋﻨﻮﺍﻥ‬
‫ﻣﺜﺎﻝ‪ ،‬ﻋﻤﻞ ﺑﺎﺯﻳﺎﺑﻲ ﻣﺮﺑﻮﻁ ﺑﻪ ‪ DML‬ﺭﻭﻳﺪﺍﺩﻫﺎﻱ ﺭﻛﻮﺭﺩ ﺧﺎﺭﺟﻲ ﺭﺍ ﺑﺎﺯﻳﺎﺑﻲ ﻣﻲ ﻛﻨﺪ‪ ،‬ﻧﻪ ﺭﻭﻳﺪﺍﺩ‬
‫ﺭﻛﻮﺭﺩ ﺫﺧﻴﺮﻩ ﺷﺪﻩ ﺭﺍ‪.‬‬
‫‪ .‬ﺍﺯ ﺍﻳﻨﺠﺎ ﺑﻪ ﺑﻌﺪ ﺍﺯ‬ ‫ﺗﻮﺟﻪ ‪ :‬ﻣﻲ ﺑﻴﻨﻴﻢ ﻛﻪ ﺍﺻﻄﻼﺡ ﺭﻛﻮﺭﺩ ﻣﻨﻄﻘﻲ ﺑﻪ ﺭﻛﻮﺭﺩ ﺧﺎﺭﺟﻲ ﺍﺷﺎﺭﻩ ﺩﺍﺭﺩ‬
‫ﺍﺻﻄﻼﺡ ﺭﻛﻮﺭﺩ ﻣﻨﻄﻘﻲ ﺍﺳﺘﻔﺎﺩﻩ ﻧﻤﻲ ﻛﻨﻴﻢ‪.‬‬
‫ﻫﺮ ﺩﻳﺪﮔﺎﻩ ﺧﺎﺭﺟﻲ ﺑﻪ ﻭﺳﻴﻠﻪ ﺷﻤﺎﻱ ﺧﺎﺭﺟﻲ ﺗﻌﺮﻳﻒ ﻣﻲ ﺷﻮﺩ ﻛﻪ ﺣﺎﻭﻱ ﺗﻌﺎﺭ ﻳﻒ ﺍﻧﻮﺍﻉ ﻫﺮ ﻳﻚ ﺍﺯ‬
‫ﺭﻛﻮﺭﺩﻫﺎﻱ ﺧﺎﺭﺟﻲ ﺩﺭ ﺁﻥ ﺩﻳﺪﮔﺎﻩ ﺧﺎﺭﺟﻲ ﺍﺳﺖ ‪ .‬ﺷﻤﺎﻱ ﺧﺎﺭﺟﻲ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺑﺨﺶ ‪ DDL‬ﻣﺮﺑﻮﻁ‬
‫ﺑﻪ ‪ DSL‬ﻧﻮﺷﺘﻪ ﻣﻲ ﺷﻮﺩ ‪ .‬ﻟﺬﺍ ‪ DDL‬ﺭﺍ ﮔﺎﻫﻲ ‪ DDL‬ﺧﺎﺭﺟﻲ ﻧﻴﺰ ﻣﻲ ﮔﻮﻳﻨﺪ ‪ .‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ‪ ،‬ﻧﻮﻉ‬

‫‪11‬‬
‫ﺭﻛﻮﺭﺩ ﺧﺎﺭﺟﻲ ﻛﺎﺭﻣﻨﺪ ﻣﻤﻜﻦ ﺍﺳﺖ ﺷﺎﻣﻞ ﺍﻳﻦ ﻓﻴﻠﺪﻫﺎ ﺑﺎﺷﺪ ‪ :‬ﺷﻤﺎﺭﻩ ﻛﺎﺭﻣﻨﺪﻱ ‪ 6‬ﻛﺎﺭﺍﻛﺘﺮﻱ ‪ ،‬ﻓﻴﻞﺩ‬
‫ﺣﻘﻮﻕ ‪ 5‬ﺭﻗﻤﻲ ﻭ ﻏﻴﺮﻩ‪.‬‬

‫ﺑﺮﺭﺳﻲ ﺳﻄﺢ ﺍﺩﺭﺍﻛﻲ ‪:‬‬

‫ﺩﻳﺪﮔﺎﻩ ﺍﺩﺭﺍﻛﻲ ﻧﻤﺎﻳﺸﻲ ﺍﺯ ﻛﻞ ﺍﻃﻼﻋﺎﺕ ﻣﻮﺟﻮﺩ ﺩﺭ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺍﺳﺖ )ﻫﻤﭽﻮﻥ ﺩﻳﺪﮔﺎﻩ ﺧﺎﺭﺟﻲ ( ﻭ‬
‫ﺩﺭ ﻣﻘﺎﻳﺴﻪ ﺑﺎ ﺭﻭﺵ ﺫﺧﻴﺮﻩ ﻓﻴﺰﻳﻜﻲ ﺩﺍﺩﻩ ﻫﺎ‪ ،‬ﺗﺎ ﺣﺪﻱ ﺍﻧﺘﺰﺍﻋﻲ ﺍﺳﺖ ‪ .‬ﺑﻄﻮﺭ ﻛﻠﻲ‪ ،‬ﺑﺎ ﻧﮕﺮﺵ ﻛﺎﺭﺑﺮ ﺑﻪ‬
‫ﺩﺍﺩﻩ ﻫﺎ ﻣﺘﻔﺎﻭﺕ ﺍﺳﺖ‪ .‬ﺩﺭ ﺩﻳﺪﮔﺎﻩ ﺍﺩﺭﺍﻛﻲ‪ ،‬ﺑﻪ ﺩ ﺍﺩﻩ ﻫﺎ ﺑﻪ ﻫﻤﺎﻥ ﺷﻜﻠﻲ ﻛﻪ ﻫﺴﺘﻨﺪ ﻧﮕﺎﻩ ﻣﻲ ﺷﻮﺩ‪ ،‬ﻧﻪ ﺑﻪ‬
‫ﺷﻜﻠﻲ ﻛﻪ ﻛﺎﺭﺑﺮﺍﻥ ﺑﻪ ﺩﻟﻴﻞ ﻣﺤﺪﻭﺩﻳﺖ ﻫﺎﻱ ﻣﻮﺟﻮﺩ ﺩﺭ ﻳﻚ ﺯﺑﺎﻥ ﻳﺎ ﺳﺨﺖ ﺍﻓﺰﺍﺭ‪ ،‬ﻣﺠﺒﻮﺭ ﺑﻪ ﺩﻳﺪﻥ ﺁﻥ‬
‫ﻫﺎ ﻣﻲ ﺷﻮﻧﺪ ‪.‬‬
‫ﺩﻳﺪﮔﺎﻩ ﺍﺩﺭﺍﻛﻲ ﺷﺎﻣﻞ ﺭﻭﻳﺪﺍﺩﻫﺎﻱ ﻣﺘﻌﺪﺩﻱ ﺍﺯ ﻫﺮﻳﻚ ﺍﺯ ﺍﻧﻮﺍﻉ ﺭﻛﻮﺭﺩ ﺍﺩﺭﺍﻛﻲ ﺍﺳﺖ ‪ .‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ‪،‬‬
‫ﻣﻤﻜﻦ ﺍﺳﺖ ﺷﺎﻣﻞ ﺭﻭﻳﺪﺍﺩﻫﺎﻱ ﺭﻛﻮﺭﺩ ﺳﺎﺯﻣﺎﻥ‪ ،‬ﺑﻪ ﻋﻼﻭﻩ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺭﻭﻳﺪﺍﺩﻫﺎﻱ ﺭﻛﻮﺭﺩ ﻛﺎﺭﻣﻨﺪ‪ ،‬ﺑﻪ‬
‫ﻋﻼﻭﻩ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺭﻭﻳﺪﺍﺩﻫﺎﻱ ﺭﻛﻮﺭﺩ ﻋﺮﺿﻪ ﻛﻨﻨﺪﻩ‪ ،‬ﺑﻪ ﻋﻼﻭﻩ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺭﻭﻳﺪﺍﺩﻫﺎﻱ ﺭﻛﻮﺭﺩ‬
‫ﻗﻄﻌﺎﺕ ﻭ ﻏﻴﺮﻩ ﺑﺎﺷﺪ‪ .‬ﺭﻛﻮﺭﺩ ﺍﺩﺭﺍﻛﻲ‪ ،‬ﺍﻟﺰﺍﻣﺎ ﺑﺎ ﺭﻛﻮﺭﺩ ﺧﺎﺭﺟﻲ ﻳﺎ ﺭﻛﻮﺭﺩ ﺫﺧﻴﺮﻩ ﺷﺪﻩ ﻳﻜﺴﺎﻥ ﻧﻴﺴﺖ ‪.‬‬
‫ﺩﻳﺪﮔﺎﻩ ﺍﺩﺭﺍﻛﻲ ﺑﻪ ﻭﺳﻲﻟﻪ ﺷﻤﺎﻱ ﺍﺩﺭﺍﻛﻲ ﺗﻌﺮﻳﻒ ﻣﻲ ﺷﻮﺩ ﻛﻪ ﺣﺎﻭﻱ ﺗﻌﺎﺭﻳﻒ ﻫﺮﻳﻚ ﺍﺯ ﺍﻧﻮﺍﻉ ﻣﺨﺘﻠﻒ‬
‫ﺭﻛﻮﺭﺩ ﺍﺩﺭﺍﻛﻲ ﺍﺳﺖ ‪ .‬ﺷﻤﺎﻱ ﺍﺩﺭﺍﻛﻲ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻳﻚ ﺯﺑﺎﻥ ﺗﻌﺮﻳﻒ ﺩﺍﺩﻩ ﺑﻪ ﻧﺎﻡ ‪ DDL‬ﺍﺩﺭﺍﻛﻲ ﻧﻮﺷﺘﻪ‬
‫‪ DDL‬ﺍﺩﺭﺍﻛﻲ ﻧﺒﺎﻳﺪ ﺷﺎﻣﻞ‬ ‫ﻣﻲ ﺷﻮﺩ ‪ .‬ﺍﮔﺮ ﺍﺳﺘﻘﻼﻝ ﻓﻴﺰﻳﻜﻲ ﺩﺍﺩﻩ ﻫﺎ ﺣﻔﻆ ﺷﻮﺩ‪ ،‬ﺍﻳﻦ ﺗﻌﺮﻳﻒ‬
‫ﻣﻼﺣﻈﺎﺗﻲ ﺩﺭ ﺧﺼﻮﺹ ﻧﻤﺎﻳﺶ ﻓﻴﺰﻳﻜﻲ ﻳﺎ ﺗﻜﻨﻴﻚ ﻫﺎﻱ ﺩﺳﺘﻴﺎﺑﻲ ﺑﺎﺷﺪ ‪ .‬ﻓﻘﻂ ﺑﺎﻳﺪ ﻣﺤﺘﻮﻳﺎﺕ ﺍﻃﻼﻋﺎﺕ‬
‫ﺭﺍ ﺗﻌﺮﻳﻒ ﻛﻨﻨﺪ ‪ .‬ﺑﻨﺎﺑﺮﺍﻳﻦ‪ ،‬ﺩﺭ ﺷﻤﺎﻱ ﺍﺩﺭﺍﻛﻲ ﻧﺒﺎﻳﺪ ﺑﻪ ﻧﻤﺎﻳﺶ ﻓﻴﻠﺪ ﺫﺧﻴﺮﻩ ﺷﺪﻩ‪ ،‬ﺩﻧﺒﺎﻟﻪ ﺭﻛﻮﺭﺩ ﺫﺧﻴﺮﻩ‬
‫ﺷﺪﻩ‪ ،‬ﺷﺎﺧﺺ ﻫﺎ‪ ،‬ﺭﻭﺵ ﻫﺎﻱ ﺩﺭﻫﻢ ﺳﺎﺯﻱ‪ ،‬ﺍﺷﺎﺭﻩ ﮔﺮﻫﺎ ﻭ ﺟﺰﺋﻴﺎﺕ ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ ﻭ ﺩﺳﺘﻴﺎﺑﻲ‪ ،‬ﺍﺭﺟﺎﻉ‬
‫ﺷﻮﺩ‪ .‬ﺍﮔﺮ ﺷﻤﺎﻱ ﺍﺩﺭﺍﻛﻲ ﻛﺎﻣﻼ ﻣﺴﺘﻘﻞ ﺍﺯ ﺩﺍﺩﻩ ﻫﺎ ﺳﺎﺧﺘﻪ ﺷﻮﺩ‪ ،‬ﺁﻧﮕﺎﻩ ﺷﻤﺎﻱ ﺧﺎﺭﺟﻲ‪ ،‬ﻛﻪ ﺑﺮﺣﺴﺐ‬
‫ﺷﻤﺎﻱ ﺍﺩﺭﺍﻛﻲ ﺗﻌﺮﻳﻒ ﺷﺪﻩ ﺍﺳﺖ‪ ،‬ﻣﺴﺘﻘﻞ ﺍﺯ ﺩﺍﺩﻩ ﻫﺎ ﺧﻮﺍﻫﺪ ﺑﻮﺩ ‪.‬‬
‫ﭘﺲ‪ ،‬ﺩﻳﺪﮔﺎﻩ ﺍﺩﺭﺍﻛﻲ‪ ،‬ﺩﻳﺪﮔﺎﻫﻲ ﻧﺴﺒﺖ ﺑﻪ ﻛﻞ ﻣﺤﺘﻮﻳﺎﺕ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺍﺳﺖ ﻭ ﺷﻤﺎﻱ ﺍﺩﺭﺍﻛﻲ‪ ،‬ﺁﻥ‬
‫ﺩﻳﺪﮔﺎﻩ ﺭﺍ ﺗﻌﺮﻳﻒ ﻣﻲ ﻛﻨﺪ ‪ .‬ﻧﻤﻴﺘﻮﺍﻥ ﮔﻔﺖ ﺷﻤﺎﻱ ﺍﺩﺭﺍﻛﻲ ﻫﻤﺎﻥ ﺗﻌﺎﺭﻳﻒ ﺭﻛﻮﺭﺩ ﺩﺭ ﺯﺑﺎﻥ ﻛﻮﺑﻮﻝ ﺍﺳﺖ ‪.‬‬
‫ﺗﻌﺎﺭﻳﻒ ﺩﺭ ﺷﻤﺎﻱ ﺍﺩﺭﺍﻛﻲ ﺷﺎﻣﻞ ﻭﻳﮋﮔﻲ ﻫﺎﻱ ﻣﺘﻌﺪﺩﻱ ﺍﺳﺖ ‪ .‬ﺑﻌﻀﻲ ﺍﺯ ﻣﺆﻟﻔﻴﻦ ﭘﻴﺸﻨﻬﺎﺩ ﻣﻲ ﻛﻨﻨﺪ ﻛﻪ‬
‫ﻫﺪﻑ ﻧﻬﺎﻳﻲ ﺷﻤﺎﻱ ﺍﺩﺭﺍﻛﻲ‪ ،‬ﺗﻮﺻﻴﻒ ﻛﻞ ﻣﺆﺳﺴﻪ ﺍﺳﺖ – ﻧﻪ ﻓﻘﻂ ﺧﻮﺩ ﺩﺍﺩﻩ ﻫﺎ‪ ،‬ﺑﻠﻜﻪ ﭼﮕﻮﻧﮕﻲ‬
‫ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺩﺍﺩﻩ ﻫﺎ ﻧﻴﺰ ﻫﺴﺖ ‪ :‬ﺩﺍﺩﻩ ﻫﺎ ﺩﺭ ﻣﺆﺳﺴﻪ‪ ،‬ﭼﮕﻮﻧﻪ ﺍﺯ ﻧﻘﻄﻪ ﺍﻱ ﺑﻪ ﻧﻘﻄﻪ ﺩﻳﮕﺮ ﻣﻲ ﺭﻭ ﻧﺪ ﻭ ﻏﻴﺮﻩ‪.‬‬

‫‪12‬‬
‫ﻳﺎﺩﺁﻭﺭﻱ ﻣﻲ ﺷﻮﺩ ﻛﻪ ﻫﻴﭽﻜﺪﺍﻡ ﺍﺯ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﺍﻣﺮﻭﺯﻱ ﺍﺯ ﺷﻤﺎﻱ ﺍﺩﺭﺍﻛﻲ ﻛﻪ ﺍﻳﻦ ﺩﺭﺟﻪ ﺍﺯ ﺗﻤﺎﻣﻴﺖ ﺭﺍ‬
‫ﺷﺎﻣﻞ ﺷﻮﺩ‪ ،‬ﭘﺸﺘﻴﺒﺎﻧﻲ ﻧﻤﻲ ﻛﻨﻨﺪ ‪ .‬ﺩﺭ ﺍﻏﻠﺐ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﻣﻮﺟﻮﺩ ‪ ،‬ﺷﻤﺎﻱ ﺍﺩﺭﺍﻛﻲ ﻛﻤﻲ ﺑﻴﺸﺘﺮ ﺍﺯ‬
‫ﺍﺟﺘﻤﺎﻉ ﺗﻤﺎﻡ ﺷﻤﺎﻫﺎﻱ ﺧﺎﺭﺟﻲ‪ ،‬ﺑﻪ ﺍﺿﺎﻓﻪ ﻗﻴﺪﻫﺎﻱ ﺍﻣﻨﻴﺘﻲ ﻭ ﺟﺎﻣﻌﻴﺖ ﺍﺳﺖ ‪ .‬ﺍﻣﺎ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﺁﻱﻧﺪﻩ‬
‫ﻣﻤﻜﻦ ﺍﺳﺖ ﺩﺭ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺷﻤﺎﻱ ﺍﺩﺭﺍﻛﻲ ﺗﺨﺼﺺ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻨﺪ ‪.‬‬

‫ﺑﺮﺭﺳﻲ ﺳﻄﺢ ﺩﺍﺧﻠﻲ ‪:‬‬

‫ﺩﻳﺪﮔﺎﻩ ﺍﺩﺭﺍﻛﻲ‪ ،‬ﻧﻤﺎﻳﺶ ﺳﻄﺢ ﭘﺎﻳﻴﻨﻲ ﺍﺯ ﻛﻞ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺍﺳﺖ ﻭ ﺷﺎﻣﻞ ﺭﻭﻳﺪﺍﺩﻫﺎﻱ ﻣﺘﻌﺪﺩﻱ ﺍﺯ‬
‫ﺍﻧﻮﺍﻉ ﻣﺨﺘﻠﻒ ﺭﻛﻮﺭﺩ ﺩﺍﺧﻠﻲ ﺍﺳﺖ ‪ .‬ﺭﻛﻮﺭﺩ ﺩﺍﺧﻠﻲ ﺍﺻﻄﻼﺡ ‪ ANSI/SPARC‬ﺑﺮﺍﻱ ﺳﺎﺧﺘﺎﺭﻱ ﺍﺳﺖ‬
‫ﻛﻪ ﺁﻥ ﺭﺍ ﺭﻛﻮﺭﺩ ﺫﺧﻴﺮﻩ ﺷﺪﻩ ﻧﺎﻣﻴﺪﻳﻢ ‪ .‬ﺩﻳﺪﮔﺎﻩ ﺩﺍﺧﻠﻲ ﺑﺎﻻﺗﺮ ﺍﺯ ﺳﻄﺢ ﻓﻴﺰﻳﻜﻲ ﺍﺳﺖ‪ ،‬ﺯﻳﺮﺍ ﺑﺎ ﺍﺻﻄﻼﺡ‬
‫‪ .‬ﻫﻤﭽﻨﻴﻦ ﺑﺎ‬ ‫ﺭﻛﻮﺭﺩﻫﺎﻱ ﻓﻴﺰﻳﻜﻲ ﻛﻪ ﺑﻠﻮﻙ ﻫﺎ ﻳﺎ ﺻﻔﺤﺎﺕ ﻧﻴﺰ ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﻧﺪ‪ ،‬ﺳﺮﻭﻛﺎﺭ ﻧﺪﺍﺭﺩ‬
‫‪ .‬ﺑﻌﺒﺎﺭﺕ ﺩﻳﮕﺮ‪ ،‬ﺩﻳﺪﮔﺎﻩ‬ ‫ﻣﻼﺣﻈﺎﺕ ﺩﺳﺘﮕﺎﻩ ﻫﺎ ﻣﺜﻞ ﺍﻧﺪﺍﺯﻩ ﺳﻴﻠﻨﺪﺭﻫﺎ ﻭ ﺷﻴﺎﺭﻫﺎ ﻧﻴﺰ ﺳﺮﻭﻛﺎﺭ ﻧﺪﺍﺭﺩ‬
‫ﺩﺍﺧﻠﻲ‪ ،‬ﻓﻀﺎﻱ ﺧﻄﻲ ﻧﺎﻣﺤﺪﻭﺩﻱ ﺭﺍ ﻓﺮﺽ ﻣﻲ ﻛﻨﺪ‪ ،‬ﺟﺰﺋﻴﺎﺕ ﭼﮕﻮﻧﮕﻲ ﻧﮕﺎﺷﺖ ﻓﻀﺎﻱ ﺁﺩﺭﺱ ﺑﻪ ﺣﺎﻓﻈﻪ‬
‫ﻓﻴﺰﻳﻜﻲ ﻛﺎﻣﻼ ﻭﺍﺑﺴﺘﻪ ﺑﻪ ﺳﻴﺴﺘﻢ ﺍﺳﺖ ﻭ ﺍﺯ ﻣﻌﻤﺎﺭﻱ ﻛﻠﻲ ﺣﺬﻑ ﺷﺪﻩ ﺍﺳﺖ ‪.‬‬
‫‪ I/O‬ﺑﻴﻦ‬ ‫ﺗﻮﺟﻪ ‪ :‬ﺑﻠﻮﻙ ﻳﺎ ﺻﻔﺤﻪ ‪ ،‬ﻭﺍﺣﺪ ‪ I/O‬ﺍﺳﺖ ‪ ،‬ﻳﻌﻨﻲ ﻣﻴﺰﺍﻥ ﺍﻃﻼﻋﺎﺗﻲ ﻛﻪ ﺩﺭ ﻳﻚ ﻋﻤﻞ‬
‫ﺣﺎﻓﻈﻪ ﺍﺻﻠﻲ ﻭ ﺣﺎﻓﻈﻪ ﺛﺎ ﻧﻮﻳﻪ ﺗﺒﺎﺩﻝ ﻣﻲ ﺷﻮﺩ‪ .‬ﺻﻔﺤﺎﺕ ﻣﻤﻜﻦ ﺍﺳﺖ ﺍﺯ ‪ 1k‬ﺗﺎ ‪ 64k‬ﺑﺎﺷﻨﺪ )‪k=1024‬‬
‫ﺑﺎﻳﺖ(‪.‬‬
‫ﺩﻳﺪﮔﺎﻩ ﺩﺍﺧﻠﻲ ﺑﻪ ﻭﺳﻴﻠﻪ ﺷﻤﺎﻱ ﺩﺍﺧﻠﻲ ﺗﻮﺻﻴﻒ ﻣﻲ ﺷﻮﺩ ﻛﻪ ﻧﻪ ﺗﻨﻬﺎ ﺍﻧﻮﺍﻉ ﺭﻛﻮﺭﺩ ﺫﺧﻴﺮﻩ ﺷﺪﻩ ﺭﺍ‬
‫ﺗﻌﺮﻳﻒ ﻣﻲ ﻛﻨﺪ ﺑﻠﻜﻪ ﻣﺸﺨﺺ ﻣﻲ ﻛﻨﺪ ﭼﻪ ﺷﺎﺧﺺ ﻫﺎﻳﻲ ﻭﺟﻮﺩ ﺩﺍﺭﺩ‪ ،‬ﻓﻴﻠﺪﻫﺎﻱ ﺫﺧﻴﺮﻩ ﺷﺪﻩ ﭼﮕﻮﻧﻪ‬
‫ﻧﻤﺎﻳﺶ ﺩﺍﺩﻩ ﻣﻲ ﺷﻮﻧﺪ‪ ،‬ﺗﺮﺗﻴﺐ ﻑ ﻳﺰﻳﻜﻲ ﺫﺧﻴﺮﻩ ﺭﻛﻮﺭﺩﻫﺎ ﭼﮕﻮﻧﻪ ﺍﺳﺖ ﻭ ﻏﻴﺮﻩ ‪ .‬ﺷﻤﺎﻱ ﺩﺍﺧﻠﻲ ﺑﺎ‬
‫ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻳﻚ ﺯﺑﺎﻥ ﺗﻌﺮﻳﻒ ﺩﺍﺩﻩ ﻫﺎ ﺑﻪ ﻧﺎﻡ ‪ DDL‬ﺩﺍﺧﻠﻲ ﻧﻮﺷﺘﻪ ﻣﻲ ﺷﻮﺩ‪.‬‬
‫ﺗﻮﺟﻪ ‪ :‬ﺩﺭﺍﻳﻨﺠﺎ ﺍﺯ ﺍﺻﻄﻼﺡ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺫﺧﻴﺮﻩ ﺷﺪﻩ ﺑﺠﺎﻱ ﺩﻳﺪﮔﺎﻩ ﺩﺍﺧﻠﻲ ﻭ ﺍﺯ ﺍﺻﻄﻼﺡ ﺗﻌﺮﻳﻒ‬
‫ﺳﺎﺧﺘﺎﺭ ﺣﺎﻓﻈﻪ ﺑﻪ ﺟﺎﻱ ﺷﻤﺎﻱ ﺩﺍﺧﻠﻲ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﻛﻨﻴﻢ ‪.‬‬
‫ﺩﺭ ﺣﺎ ﻟﺖ ﻫﺎﻳﻲ‪ ،‬ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﻛﺎﺭﺑﺮﺩﻱ )ﺑﻪ ﺧﺼﻮﺹ ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﺳﻮﺩﻣﻨﺪ (‪ ،‬ﻣﻤﻜﻦ ﺍﺳﺖ ﺍﻳﻦ ﺍﻣﻜﺎﻥ ﺭﺍ‬
‫ﻓﺮﺍﻫﻢ ﻛﻨﻨﺪ ﻛﻪ ﺑﺠﺎﻱ ﻛﺎﺭﻛﺮﺩﻥ ﺩﺭ ﺳﻄﺢ ﺧﺎﺭﺟﻲ‪ ،‬ﺑﺘﻮﺍﻥ ﻣﺴﺘﻘﻴﻤﺎ ﺩﺭ ﺳﻄﺢ ﺩﺍﺧﻠﻲ ﻛﺎﺭﻛﺮﺩ ‪ .‬ﺍﻳﻦ ﻛﺎﺭ‬
‫ﭘﻴﺸﻨﻬﺎﺩ ﻧﻤﻲ ﺷﻮﺩ ‪ .‬ﺩﺭ ﺍﻳﻦ ﺣﺎﻟﺖ ﺧﻄﺮ ﺍﻣﻨﻴﺘﻲ ﻭ ﺧﻄﺮ ﺟﺎﻣﻌﻴﺖ ﻭﺟﻮﺩ ﺩﺍﺭﺩ ﺯﻳﺮﺍ ﻣﺤﺪﻭﺩﻳﺖ ﻫﺎﻱ‬
‫ﺁﻧﻬﺎ ﻧﺎﺩﻳﺪﻩ ﮔﺮﻓﺘﻪ ﻣﻲ ﺷﻮﺩ ﻭ ﺑﺮﻧﺎﻣﻪ ﻧﻴﺰ ﺑﻪ ﺩﺍﺩﻩ ﻫﺎ ﻭﺍﺑﺴﺘﻪ ﺧﻮﺍﻫﺪ ﺑﻮﺩ ‪ .‬ﺍﻣﺎ ﮔﺎﻫﻲ‪ ،‬ﺍﻳﻦ ﺭﻭﺵ‪ ،‬ﺗﻨﻬﺎ ﺭﺍﻩ‬

‫‪13‬‬
 ‫ﺩﺳﺘﻴﺎﺑﻲ ﺑﻪ ﻋﻤﻠﻜﺮﺩ ﻛﺎﺭﺍﻳﻲ ﻣﻄﻠﻮﺏ ﺍﺳﺖ ‪ .‬ﻣﺜﻞ ﺣﺎﻟﺘﻲ ﻛﻪ ﺑﺮﻧﺎﻣﻪ ﻧﻮﻳﺴﺎﻥ ﺩﺭ ﺯﺑﺎﻥ ﻫﺎﻱ ﺳﻄﺢ ﺑﺎﻻ‬
‫ﻣﺠﺒﻮﺭ ﻣﻲ ﺷﻮﻧﺪ ﺍﺯ ﺯﺑﺎﻥ ﺍﺳﻤﺒﻠﻲ ﺍﺳﺘﻔﺎﺩﻩ ﻛﻨﻨﺪ ﺗﺎ ﻛﺎﺭﺍﻳﻲ ﻣﻄﻠﻮﺏ ﺭﺍ ﺑﻪ ﺩﺳﺖ ﺁﻭﺭﻧﺪ ‪.‬‬

‫ﻧﺮﻣﺎﻝﺳﺎﺯﻱ‬

‫ﻫﺪﻑ ﺍﺯ ﻧﺮﻣﺎﻝ ﺳﺎﺯﻱ ﻛﺎﻫﺶ ﺟﺪﺍﻭﻝ ﭘﻴﭽﻴﺪﻩ ﺑﻪ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﻛﻮﭼﻚ ﻭ ﭘﺎﻳﺪﺍﺭ ﺍﺯ ﺳﺎﺧﺘﺎﺭ ﺩﺍﺩﻩ ﻫﺎ ﻣﻲ ‪-‬‬
‫ﺑﺎﺷﺪ‪ .‬ﺳﺎﺧﺘﺎﺭ ﻧﺮﻣﺎﻝ ﺷﺪﻩ ﺍﻧﻌﻄﺎﻑ ﭘﺬﻳﺮﺗﺮ‪ ،‬ﭘﺎﻳﺪﺍﺭﺗﺮ ﻭ ﺍﺯ ﻟﺤﺎﻅ ﻧﮕﻬﺪﺍﺭﻱ ﺭﺍﺣﺖ ﺗﺮ ﺍﺯ ﺳﺎﺧﺘﺎﺭ ﻏﻴﺮﻧﺮﻣﺎﻝ‬
‫ﺍﺳﺖ‪.‬‬
‫ﺳﻄﻮﺡ ﻧﺮﻣﺎﻝﺳﺎﺯﻱ‪:‬‬
‫• ‪ :1NF‬ﺣﺬﻑ ﮔﺮﻭﻩﻫﺎﻱ ﺗﻜﺮﺍﺭﺷﻮﻧﺪﻩ‬
‫• ‪ :2NF‬ﺣﺬﻑ ﻭﺍﺑﺴﺘﮕﻲﻫﺎﻱ ﺟﺰﺋﻲ‬
‫• ‪ :3NF‬ﺣﺬﻑ ﻭﺍﺑﺴﺘﮕﻲﻫﺎﻱ ﺗﺎﺑﻌﻲ ﺑﺎ ﻭﺍﺳﻄﻪ‬
‫)ﺗﺎ ‪ 3NF‬ﺿﺮﻭﺭﺕ ﺩﺍﺭﺩ(‪.‬‬

‫ﺿﺮﻭﺭﺕ ﻧﺮﻣﺎﻝﺳﺎﺯﻱ‪:‬‬
‫• ﻛﺎﻫﺶ ﺍﻓﺰﻭﻧﮕﻲ ﺩﺍﺩﻩ‬
‫• ﻛﺎﻫﺶ ﺑﻲﻧﻈﻤﻲ‬
‫• ﻛﺎﻫﺶ ﻣﻘﺎﺩﻳﺮ ﺗﻬﻲ‬

‫ﺁﺷﻨﺎﻳﻲ ﺑﺎ ﺩﺍﺩﻩﻛﺎﻭﻱ )‪(Data Mining‬‬

‫ﺩﺍﺩﻩﻛﺎﻭﻱ ﻋﺒﺎﺭﺕ ﺍﺳﺖ ﺍﺯ ﻓﺮﺁﻳﻨﺪ ﺍﺳﺘﺨﺮﺍﺝ ﺩﺍﻧﺶ ﺍﺯ ﺣﺠﻢ ﺍﻧﺒﻮﻩ ﺩﺍﺩﻩﻫﺎ‪.‬‬

‫ﺑﻪ ﻋﻨﻮﺍﻥ ﻳﻚ ﻣﺜﺎﻝ ﺩﻳﮕﺮ ﺩﺭ ﻳﻚ ﺑﻴﻤﺎﺭﺳﺘﺎﻥ‪ ،‬ﺍﻃﻼﻋﺎﺕ ‪ 100‬ﺑﻴﻤﺎﺭ ﻳﻚ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺭﺍ ﺍﻳﺠﺎﺩ ﻣﻲ ﻛﻨﺪ‪.‬‬
‫ﺩﺍﺩﻩﻛﺎﻭﻱ ﺍﺯ ﺍﻳﻦ ﺩﺍﺩﻩ ﻫﺎ ﺩﺍﻧﺶ ﺍﻳﺠﺎﺩ ﻣﻲ ﻛﻨﺪ‪ .‬ﻣﺜﻼ ‪ 95‬ﺩﺭﺻﺪ ﺧﺎﻧﻢ ﻫﺎﻳﻲ ﻛﻪ ﺩﻳﺎﺑﺖ ﺩﺍﺷﺘﻪ ﺍﻧﺪ ﺑﻌﺪ ﺍﺯ‬

‫‪14‬‬
 ‫ﻋﻤﻞ ﻓﻮﺕ ﻧﻤﻮﺩﻩ ﺍﻧﺪ‪ .‬ﺍﻳﻦ ﺩﺍﻧﺶ ﺑﻪ ﻣﺎ ﻧﺸﺎﻥ ﻣﻲ ﺩﻫﺪ ﻛﻪ ﺍﮔﺮ ﭼﻨﻴﻦ ﺑﻴﻤﺎﺭﻱ ﻣﺠﺪﺩﺍ ﺁﻣﺪ ﻛﺎﺩﺭ ﭘﺰﺷﻜﻲ‬
‫ﺑﺪﺍﻧﻨﺪ ﻛﻪ ﭼﮕﻮﻧﻪ ﺑﺎﻳﺪ ﻋﻤﻞ ﻛﻨﻨﺪ‪.‬‬
‫ﺑﻨﺎﺑﺮﺍﻳﻦ ﺩﺍﺩﻩﻛﺎﻭﻱ ﻳﻚ ﺍﺳﺘﻨﺘﺎﺝ ﺍﺯ ﺩﺍﺩﻩﻫﺎ ﻭ ﺑﻪ ﻋﺒﺎﺭﺕ ﺩﻳﮕﺮ ﭘﻴﺪﺍﻛﺮﺩﻥ ﺍﻟﮕﻮ ﺍﺯ ﺩﺍﺩﻩ ﻫﺎ ﻣﻲﺑﺎﺷﺪ‪.‬‬
‫)‪ (task‬ﺍﺻﻠﻲ‬ ‫ﺩﺭ ﺩﺍﺩﻩ ﻛﺎﻭﻱ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻣﺪﻟﻬﺎﻱ ﻫﻮﺵ ﻣﺼﻨﻮﻋﻲ ﻣﺎﻧﻨﺪ ﺗﺌﻮﺭﻱ ﻓﺎﺯﻱ‪ ،‬ﺳﻪ ﻛﺎﺭ‬
‫ﺻﻮﺭﺕ ﻣﻲﮔﻴﺮﺩ‪:‬‬
‫• ‪ :Association‬ﻳﻚ ﺳﺮﻱ ﻗﻮﺍﻧﻴﻦ ﺍﻧﺠﻤﻨﻲ )ﺍﻟﮕﻮ ﻳﺎ ‪ (Pattern‬ﺍﻳﺠﺎﺩ ﻣﻲﻛﻨﺪ ﻭ ﻳﻚ ﺩﺍﻧﺶ ﻭ‬
‫ﺍﻟﮕﻮﻱ ﺧﺎﺹ ﺑﻪ ﻣﺎ ﻣﻲ ﺩﻫﺪ‪ .‬ﻳﻌﻨﻲ ﺍﺯ ﺭﻭﻱ ﺩﺍﺩﻩ ﻫﺎ ﻳﻚ ﻣﺪﻝ ﻫﻮﺷﻤﻨﺪ ﭘﻴﺶ ﺑﻴﻨﻲ ﺍﻳﺠﺎﺩ ﻣﻲ ‪-‬‬
‫ﺷﻮﺩ‪.‬‬
‫• ‪ :Clustering‬ﺑﻪ ﻣﻌﻨﻲ ﺧﻮﺷﻪ ﺑﻨﺪﻱ ﻣﻲ ﺑﺎﺷﺪ‪ .‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻧﻤﻮﻧﻪ ﺍﮔﺮ ﺑﻴﻤﺎﺭﺍﻥ ﺭﺍ ﺑﺮﺍﺳﺎﺱ ﺻﻔﺘﻬﺎ‬
‫ﻳﺎ ﻭﻳﮋﮔﻴﻬﺎ‪ 1‬ﺑﺪﻭﻥ ﻫﻴﭻ ﮔﻮﻧﻪ ﺑﺮﭼﺴﺒﻲ ‪ 2‬ﮔﺮﻭﻩﺑﻨﺪﻱ ﻛﻨﻴﻢ‪ ،‬ﻋﻤﻞ ﺧﻮﺷﻪﺑﻨﺪﻱ ﺍﻧﺞﺍﻡ ﺩﺍﺩﻩ ﺍﻳﻢ‪.‬‬
‫‪F1‬‬ ‫‪F0‬‬

‫• ‪ :Classification‬ﺑﻪ ﻣﻌﻨﻲ ﺩﺳﺘﻪ ﺑﻨﺪﻱ ﻳﺎ ﻛﻼﺱ ﺑﻨﺪﻱ ﺍﺳﺖ ‪ .‬ﻣﺜﻼ ﺍﮔﺮ ﻫﻤﺎﻥ ﺑﻴﻤﺎﺭﺍﻥ ﺭﺍ ﺍﻳﻦ‬
‫ﺑﻨﺪﻱ ﻛﻨﻴﻢ‪ ،‬ﺩﺳﺘﻪ ﺑﻨﺪﻱ‬ ‫ﺑﺎﺭ ﺑﺮﺍﺳﺎﺱ ﻳﻚ ﺳﺮﻱ ﺑﺮﭼﺴﺒﻬﺎﻱ ﺍﺯﭘﻴﺶ ﻣﺸﺨﺺ ﺷﺪﻩ ﮔﺮﻭﻩ‬
‫ﻣﺤﺴﻮﺏ ﻣﻲﺷﻮﺩ‪ .‬ﻳﻜﻲ ﺍﺯ ﺭﻭﺷﻬﺎﻱ ﺩﺳﺘﻪﺑﻨﺪﻱ ﺩﺭﺧﺖ ﺗﺼﻤﻴﻢ ‪ 3‬ﺍﺳﺖ‪.‬‬
‫‪F2‬‬

‫ﺩﺭ ﺣﻘﻴﻘﺖ ﻛﺎﺭ ﭘﻴﺸﮕﻮﻳﻲ ‪ 4‬ﺩﺭ ﻛﻼﺱﺑﻨﺪﻱ ﺍﻧﺠﺎﻡ ﻣﻲﺷﻮﺩ ﻳﻌﻨﻲ ﻓﺮﺁﻳﻨﺪ ‪ classification‬ﺍﻓﺮﺍﺩ ﺟﺪﻳﺪ‬
‫‪F3‬‬

‫ﺭﺍ ﭘﻴﺸﮕﻮﻳﻲ ﻣﻲ ﻛﻨﺪ ﻭ ﻣﺸﺨﺺ ﻣﻲ ﻛﻨﺪ ﻛﻪ ﺁﻳﺎ ﺑﻴﻤﺎﺭ ﺍﺳﺖ ﻳﺎ ﺳﺎﻟﻢ ‪ .‬ﺑﻪ ﻋﺒﺎﺭﺕ ﺩﻳﮕﺮ ﻣﺸﺨﺺ ﻣﻲ ﺷﻮﺩ‬
‫ﻛﻪ ﻓﺮﺩﻱ ﻛﻪ ﺟﺪﻳﺪﺍً ﻭﺍﺭﺩ ﻣﺪﻝ ﺷﺪﻩ ﺍﺳﺖ ﺩﺭ ﻛﺪﺍﻡ ﺩﺳﺘﻪ ﻗﺮﺍﺭ ﻣﻲ ﮔﻴﺮﺩ‪ .‬ﺍﻣﺎ ﺩﺭ ﺧﻮﺷﻪﺑﻨﺪﻱ ﻧﻤﻮﻧﻪﻫﺎﻱ‬
‫ﻗﺪﻳﻤﻲ ﮔﺮﻭﻩﺑﻨﺪﻱ ﻣﻲ ﺷﻮﺩ ﻭ ﺁﻧﻬﺎﻳﻲ ﻛﻪ ﺩﺭ ﻳﻚ ﻛﻼﺳﺘﺮ ﻫﺴﺘﻨﺪ ﺑﻪ ﺍﻳﻦ ﻣﻌﻨﻲ ﺍﺳﺖ ﻛﻪ ﻭﻳﮋﮔﻴﻬﺎﻳﺸﺎﻥ‬
‫ﻧﺰﺩﻳﻚ ﺑﻪ ﻫﻢ ﻣﻲﺑﺎﺷﺪ‪.‬‬
‫ﺩﺍﺩﻩﻛﺎﻭﻱ ﻛﺎﺭﺑﺮﺩﻫﺎﻱ ﺯﻳﺎﺩﻱ ﺩﺍﺭﺩ ﻣﺎﻧﻨﺪ ﭘﺰﺷﻜﻲ‪ ،‬ﻣﺎﻟﻲ ﻭ ‪....‬‬
‫ﻳﺎﺩﮔﻴﺮﻱ ﻣﺎﺷﻴﻦ ‪ :5‬ﺳﻴﺴﺘﻢﻫﺎﻱ ﺧﻮﺩﻳﺎﺩﮔﻴﺮﻧﺪﻩ ﻛﻪ ﺑﻪ ﻃﻮﺭ ﺧﻮﺩﻛﺎﺭ ﻳﺎﺩ ﻣﻲ ﮔﻴﺮﻧﺪ )ﻣﺜﻼ ﻳﻚ ﻣﺎﺷﻴﻦ‬
‫‪F4‬‬

‫ﻛﻪ ﺷﻄﺮﻧﺞ ﺭﺍ ﺑﻪ ﺻﻮﺭﺕ ﺍﺗﻮﻣﺎﺗﻴﻚ ﻳﺎﺩ ﻣﻲ ﮔﻴﺮﺩ(‪.‬‬

‫ﺳﻴﺴﺘﻤﻬﺎﻱ ﭘﻴﺸﻨﻬﺎﺩﺩﻫﻨﺪﻩ ‪ :6‬ﺍﻳﻦ ﺳﻴﺴﺘﻢ ﻫﺎ ﺑﺮﺍﺳﺎﺱ ﻣﻌﻴﺎﺭﻫﺎﻱ ﺷﺒﺎﻫﺖ‪ ،‬ﺁﻳﺘﻤﻬﺎﻱ ﻣﻮﺭﺩ ﻋﻼﻗﻪ ﻛﺎﺭﺑﺮ‬
‫‪F5‬‬

‫ﺭﺍ ﺣﺪﺱ ﻣﻲ ﺯﻧﺪ ﻭ ﺑﻪ ﺍﻭ ﭘﻴﺸﻨﻬﺎﺩ ﻣﻲ ﺩﻫﺪ‪ .‬ﺑﻪ ﻋﺒﺎﺭﺕ ﺩﻳﮕﺮﻱ ﺍﻓﺮﺍﺩ ﺷﺒﻴﻪ ﺑﻪ ﻫﻢ ﺭﺍ ﺷﻨﺎﺳﺎﻳﻲ ﻣﻲ ﻛﻨﺪ ﻭ‬

‫‪1‬‬
‫‪Features‬‬
‫‪2‬‬
‫‪label‬‬
‫‪3‬‬
‫‪Decision Tree‬‬
‫‪4‬‬
‫‪prediction‬‬
‫‪5‬‬
‫‪Machine Learning‬‬
‫‪6‬‬
‫‪Recommender System‬‬

‫‪15‬‬
‫ﺑﻪ ﻳﻚ ﻛﺎﺭﺑﺮ ﺁﻳﺘﻤﻬﺎﻳﻲ ﺭﺍ ﭘﻴﺸﻨﻬﺎﺩ ﻣﻲ ﻛﻨﺪ ﻛﻪ ﻛﺎﺭﺑﺮﺍﻥ ﻣﺸﺎﺑﻪ ﺑﺎ ﻭﻱ ﺁﻧﻬﺎ ﺑﻪ ﺁﻧﻬﺎ ﺍﻣﺘﻴﺎ ﺯ ﺑﺎﻻﻳﻲ ﺩﺍﺩﻩﺍﻧﺪ‪.‬‬
‫ﺑﻪ ﻋﺒﺎﺭﺕ ﺩﻳﮕﺮ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﭘﻴﺸﻨﻬﺎﺩﺩﻫﻨﺪﻩ ﺍﺯ ﻧﻈﺮﻛﺎﻭﻱ ﺑﺮﺍﻱ ﭘﻴﺸﻨﻬﺎﺩ ﺁﻳﺘﻤﻬﺎﻱ ﻣﻨﺎﺳﺐ ﺑﻪ ﻛﺎﺭﺑﺮ‬
‫ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲﻛﻨﻨﺪ‪.‬‬

‫ﭘﻴﺶﺍﺭﺯﻳﺎﺑﻲ ﻛﻼﺱ ﺑﺮﺍﺳﺎﺱ ﻣﻔﺎﻫﻴﻢ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻣﻘﺪﻣﺎﺗﻲ‬

‫‪ - 1‬ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﭼﻴﺴﺖ؟‬

‫ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻳﻚ ﻣﺤﻞ ﻧﮕﻬﺪﺍﺭﻱ ﺭﻛﻮﺭﺩﻫﺎ ﻭ ﺩﺍﺩﻩ ﻫﺎﺳﺖ ﻛﻪ ﻛﺎﺭﺑﺮﺍﻥ ﺑﻪ ﻃﻮﺭ ﻣﺸﺘﺮﻙ ﺍﺯ ﺁﻥ ﺍﺳﺘﻔﺎﺩﻩ‬
‫ﻣﻲﻛﻨﻨﺪ ﻣﺜﻞ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺭﺯﺭﻭﺍﺳﻴﻮﻥ ﻓﺮﻭﺩﮔﺎﻩ ﻭ ﺭﺍﻩﺁﻫﻦ‬
‫ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻓﻘﻂ ﺩﺍﺩﻩﻫﺎ ﺭﺍ ﻧﮕﻬﺪﺍﺭﻱ ﻣﻲ ﻛﻨﺪ ﻭ ﻋﻤﻠﻴﺎﺕ ﺍﺻﻠﻲ ﺭﻭﻱ ﺩﺍﺩﻩ ﻫﺎ ﻋﺒﺎﺭﺗﻨﺪ ﺍﺯ ﺫﺧﻴﺮﻩ ﻭ ﺑﺎﺯﻳﺎﺑﻲ‬
‫ﻛﻪ ﺩﺭ ﻗﺎﻟﺐ ﻳﻚ ﺳﺮﻱ ﭘﺮﺱ ﻭﺟﻮﻫﺎ ﺻﻮﺭﺕ ﻣﻲﮔﻴﺮﺩ‪.‬‬

‫‪ - 2‬ﻣﺤﺒﻮﺑﺘﺮﻳﻦ ﻧﻮﻉ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﭼﻴﺴﺖ؟ ﭼﺮﺍ؟‬

‫ﻧﻮﻉ ﺭﺍﺑﻄﻪﺍﻱ ﻛﻪ ﺍﺯ ﺟﺪﻭﻝ )‪ (Table‬ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲﻛﻨﺪ‪ .‬ﺯﻳﺮﺍ ﺍﺯ ﺳﺎﺩﻩﺗﺮ ﺍﺳﺖ‪.‬‬
‫ﺩﺭ ﺍﻳﻦ ﻧﻮﻉ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ‪:‬‬
‫ﺍﺯ ‪ Join‬ﺑﺎﺭﻱ ﭘﺮﺱﻭﺟﻮ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲﺷﻮﺩ‪.‬‬ ‫•‬
‫ﺍﺯ ﺁﺩﺭﺱ ﻓﻴﺰﻳﻜﻲ ﺍﺳﺘﻔﺎﺩﻩ ﻧﻤﻲﺷﻮﺩ‪.‬‬ ‫•‬
‫ﺟﺪﺍﻭﻝ ﻣﺴﺘﻘﻞ ﺍﺯ ﻫﻢ ﻫﺴﺘﻨﺪ ﻭ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ‪object‬ﻫﺎ ﺳﺎﺧﺘﻤﺎﻥ ﺩﺍﺩﻩ ﻫﺎ ﺭﺍ ﻣﻲ ﺗﻮﺍﻥ ﻣﺪﻝ ‪-‬‬ ‫•‬
‫ﺳﺎﺯﻱ ﻧﻤﻮﺩ ﻭ ﺑﺎ ﻋﻤﻞﮔﺮﻫﺎ ﻭ ‪function‬ﻫﺎ ﺭﻓﺘﺎﺭ ﺳﻴﺴﺘﻢ ﺭﺍ ﻣﺪﻳﺮﻳﺖ ﻛﺮﺩ‪.‬‬

‫‪ - 3‬ﻣﻬﻤﺘﺮﻳﻦ ﻭﻳﮋﮔﻲ ﻳﻚ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﭼﻴﺴﺖ؟‬

‫ﻣﻬﻤﺘﺮﻳﻦ ﻭﻳﮋﮔﻲ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻋﺒﺎﺭﺕ ﺍﺳﺖ ﺍﺯ ﺍﺳﺘﻘﻼﻝ ﺩﺍﺩﻩ ﻫﺎ‪ ،‬ﻣﺼﻮﻧﻴﺖ )ﺍﻳﻤﻨﻲ( ﺑﺮﻧﺎﻣﻪ ﻛﺎﺭﺑﺮﺩﻱ‬
‫ﻧﺴﺒﺖ ﺑﻪ ﺗﻐﻴﻴﺮ ﺭﻭﺵ ﻧﻤﺎﻳﺶ ﻓﻴﺰﻳﻜﻲ ﻭ ﺭﻭﺵ ﻫﺎﻱ ﺩﺳﺘﺮﺳﻲ ﺩﺍﺩﻩ‬

‫‪ - 4‬ﺳﻴﺴﺘﻢ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﭼﻴﺴﺖ؟‬

‫‪16‬‬
‫‪-‬‬ ‫ﻳﻚ ﺳﻴﺴﺘﻢ ﻛﺎﻣﭙﻴﻮﺗﺮﻱ ﺑﺮﺍﻱ ﻧﮕﻬﺪﺍﺭﻱ ﺭﻛﻮﺭﺩﻫﺎﺳﺖ ﻭ ﻫﺪﻑ ﺍﺻﻠﻲ ﺁﻥ ﺫﺧﻴﺮﻩ ﺍﻃﻼﻋﺎﺕ ﻭ ﻓﺮﺍﻫﻢ‬
‫ﻧﻤﻮﺩﻥ ﺍﻣﻜﺎﻧﺎﺗﻲ ﺑﺮﺍﻱ ﺑﺎﺯﻳﺎﺑﻲ ﻭ ﺑﻬﻨﮕﺎﻡ ﺳﺎﺯﻱ ﺍﻃﻼﻋﺎﺕ ﻛﺎﺭﺑﺮﺍﻥ ﺍﺳﺖ ‪ .‬ﺳﻴﺴﺘﻢ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺍﺯ ﭼﻬﺎﺭ‬
‫ﻋﻨﺼﺮ ﺍﺻﻠﻲ ﺗﺸﻜﻴﻞ ﺷﺪﻩ ﺍﺳﺖ‪:‬‬
‫ﺩﺍﺩﻩﻫﺎ‬ ‫•‬
‫ﻧﺮﻡﺍﻓﺰﺍﺭ‬ ‫•‬
‫ﺳﺨﺖﺍﻓﺰﺍﺭ‬ ‫•‬
‫ﻛﺎﺭﺑﺮﺍﻥ‬ ‫•‬

‫‪ - 5‬ﻣﻬﻤﺘﺮﻳﻦ ﻋﻨﺼﺮ ﻳﻚ ﺳﻴﺴﺘﻢ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﭼﻴﺲﺕ؟‬

‫‪) DBMS‬ﺳﻴﺴﺘﻢ ﻣﺪﻳﺮﻳﺖ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻫﺎ(‪ DBMS .‬ﻳﻚ ﻻﻳﻪ ﻧﺮﻡ ﺍﻓﺰﺍﺭﻱ ﺍﺳﺖ ﻛﻪ ﻫﺮ ﻧﻮﻉ ﺩﺭﺧﻮﺍﺳﺖ‬
‫ﻛﻪ ﺍﺯ ﻛﺎﺭﺑﺮﺍﻥ ﺩﺭﻳﺎﻓﺖ ﻣﻲ ﺷﻮﺩ ﻣﺜﻞ ‪ DDL‬ﻳﺎ ‪ DML‬ﻳﺎ ‪ DCL‬ﻭ ‪ ،...‬ﺁﻥ ﺭﺍ ﺟﻮﺍﺏ ﻣﻲ ﺩﻫﺪ ﻭ ﺍﻣﻨﻴﺖ ﺭﺍ‬
‫ﻛﻨﺘﺮﻝ ﻣﻲﻛﻨﺪ‪.‬‬

‫‪ - 6‬ﻧﺮﻣﺎﻝﺳﺎﺯﻱ ﻳﻚ ﺿﺮﻭﺭﺕ ﺍﺳﺖ؟ ﺗﺎ ﭼﻪ ﺳﻄﺤﻲ ﺍﺯ ﻧﺮﻣﺎﻝ ﺳﺎﺯﻱ؟‬

‫ﺗﺎ ﺳﻄﺢ ﺳﻮﻡ )‪.(3NF‬‬

‫‪ - 7‬ﺕﺭﺍﻛﻨﺶ ﭼﻴﺴﺖ؟ ﺁﻳﺎ ﻳﻚ ﺿﺮﻭﺭﺕ ﺍﺳﺖ؟‬

‫ﻳﻚ ﻭﺍﺣﺪ ﺍﺟﺮﺍﻳﻲ ﺑﺮﻧﺎﻣﻪ ﻛﻪ ﺍﻧﻮﺍﻉ ﺁﻳﺘﻤﻬﺎﻱ ﺩﺍﺩﻩ ﺍﻱ ﺭﺍ ﻗﺎﺑﻞ ﺩﺳﺘﺮﺳﻲ ﻭ ﺑﺮﻭﺯﺭﺳﺎﻧﻲ ﻣﻲ ﻛﻨﺪ‪ .‬ﺑﻠﻪ‪،‬‬
‫ﺗﺮﺍﻛﻨﺶ ﻳﻚ ﺿﺮﻭﺭﺕ ﻣﺤﺴﻮﺏ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪ - 8‬ﺧﻮﺍﺹ ﺗﺮﺍﻛﻨﺶ )‪ (ACID‬ﭼﻴﺴﺖ؟‬

‫ﺧﻮﺍﺹ ‪ ACID‬ﺷﺎﻣﻞ ﭼﻬﺎﺭ ﺧﺎﺻﻴﺖ ﻣﻬﻢ ﺩﺭ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺍﺳﺖ ﻛﻪ ﻋﺒﺎﺭﺗﻨﺪ ﺍﺯ ‪:‬‬
‫• ‪ :Atomicity‬ﺍﺗﻤﻴﻚ )ﻏﻴﺮﻗﺎﺑﻞﺗﺠﺰﻳﻪ( ﺑﻮﺩﻥ‪ .‬ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﻳﺎ ﺑﺎﻳﺪ ﺑﻪ ﻃﻮﺭ ﻛﺎﻣﻞ ﺍﺟﺮﺍ ﺷﻮﺩ ﻭ‬
‫ﻳﺎ ﺍﻳﻦ ﻛﻪ ﺍﺻﻼ ﺍﺟﺮﺍ ﻧﺸﻮﺩ )ﻳﻌﻨﻲ ﺍﮔﺮ ﺑﺨﺸﻲ ﺍﺯ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺍﺟﺮﺍ ﺷﺪﻩ ﻭ ﺑﻘﻴﻪ ﺍﺵ ﺑﻪ ﺩﻻﻳﻠﻲ ﻧﻤﻲ ‪-‬‬
‫ﺗﻮﺍﻧﺪ ﻛﺎﻣﻞ ﺷﻮﺩ‪ ،‬ﻻﺯﻡ ﺍﺳﺖ ﻋﻤﻠﻴﺎﺕ ﺍﻧﺠﺎﻡ ﺷﺪﻩ ﺗﺮﺍﻛﻨﺶ ﻫﻢ ﺑﺮﮔﺮﺩﺍﻧﺪﻩ ﺷﻮﺩ(‬
‫• ‪ :Consistency‬ﺳﺎﺯﮔﺎﺭﻱ‪ .‬ﻫﺮ ﺗﺮﺍﻛﻨﺶ ﺑﺎﻳﺪ ﻭﺿﻊﻳﺖ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺭﺍ ﺍﺯ ﻳﻚ ﻭﺿﻌﻴﺖ ﺳﺎﺯﮔﺎﺭ‬
‫ﺑﻪ ﻳﻚ ﻭﺿﻌﻴﺖ ﺳﺎﺯﮔﺎﺭ ﺩﻳﮕﺮ ﻣﻨﺘﻘﻞ ﻛﻨﺪ‪.‬‬
‫• ‪ :Isolation‬ﺟﺎﻣﻌﻴﺖ‪ .‬ﺗﺮﺍﻛﻨﺶ ﺑﺎﻳﺪ ﺻﺤﺖ ﻭ ﺟﺎﻣﻌﻴﺖ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺭﺍ ﺣﻔﻆ ﻛﻨﺪ ‪.‬‬

‫‪17‬‬
 ‫• ‪ :Durabilty‬ﻣﺎﻧﺪﮔﺎﺭﻱ ﻳﺎ ﭘﺎﻳﺪﺍﺭﻱ ‪ .‬ﺩﺭﺻﻮﺭﺗﻲ ﻛﻪ ﺗﺮﺍﻛﻨﺶ ﺑﺎ ﻣﻮﻓﻘﻴﺖ ﭘﺎﻳﺎﻥ ﻳﺎﺑﺪ‪ ،‬ﺍﺛﺮﺍﺕ ﺁﻥ‬
‫ﺭﻭﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺑﺎﻳﺪ ﺑﻪ ﺻﻮﺭﺕ ﺩﺍﺋﻤﻲ ﻣﺎﻧﺪﮔﺎﺭ ﺑﻤﺎﻧﺪ‪.‬‬

‫‪ - 9‬ﻓﺮﻣﺎﻥﻫﺎﻱ ‪ Commit & Rollback‬ﭼﻪ ﻧﻘﺸﻲ ﺩﺭ ﻣﺪﻳﺮﻳﺖ ﺗﺮﺍﻛﻨﺸﻬﺎ ﺩﺍﺭﻧﺪ؟‬

‫ﺯﻣﺎﻧﻲ ﻛﻪ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺑﺎ ﻣﻮﻓﻘﻴﺖ ﺑﻪ ﺍﻧﺠﺎﻡ ﻣﻲ ﺭﺳﺪ ‪ commit‬ﻣﻲﻛﻨﺪ ﺩﺭﻏﻴﺮﺍﻳﻦﺻﻮﺭﺕ ‪Rollback‬‬
‫ﻣﻲﺷﻮﺩ ﺗﺎ ﺍﺛﺮ ﺗﻤﺎﻡ ﺩﺳﺘﻮﺍﺭﺕ ﺭﺍ ﺧﻨﺜﻲ ﻛﻨﺪ ﻳﻌﻨﻲ ﺍﮔﺮ ﻳﻜﻲ ﺍﺯ ﺧﺎﺻﻴﺖ ﻫﺎﻱ ﺗﺮﺍﻛﻨﺶ ﺑﺮﻗﺮﺍﺭ ﻧﺒﺎﺷﺪ‬
‫‪ Rollback‬ﻣﻲﻛﻨﻢ‪ Rollback .‬ﻓﻘﻂ ‪ Abort‬ﻧﻴﺴﺖ ﺑﻠﻜﻪ ﺷﺎﻣﻞ ﻋﻤﻠﻴﺎﺗﻲ ﻣﺎﻧﻨﺪ ‪ Undo‬ﻭ ‪Redo‬‬
‫ﻫﻢ ﻣﻲﺑﺎﺷﺪ‪.‬‬

‫‪ - 10‬ﻣﺪﻳﺮﻳﺖ ﺗﺮﻣﻴﻢ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﭼﻴﺴﺖ؟‬

‫ﻣﻨﻈﻮﺭ ﻋﻤﻠﻴﺎﺗﻲ ﺍﺳﺖ ﻛﻪ ﭘﺎﻳﺪﺍﺭﻱ ﻭ ﺍﺗﻤﻴﻚ ﺑﻮﺩﻥ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺭﺍ ﺗﻀﻤﻴﻦ ﻣﻲﻛﻨﺪ‪.‬‬

‫‪ - 11‬ﻣﺸﻜﻼﺕ ﺍﻧﺠﺎﻡ ﻫﻤﺮﻭﻧﺪ ﺗﺮﺍﻛﻨﺸﻬﺎ )‪ (Concurrency Control‬ﭼﻴﺴﺖ؟‬

‫ﻣﺸﻜﻞ ﺍﺻﻠﻲ ﺩﺭ ﺍﺟﺮﺍﻱ ﻫﻤﺮﻭﻧﺪ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺍﻳﻦ ﺍﺳﺖ ﻛﻪ ﻣﻤﻜﻦ ﺍﺳﺖ ﻧﺎﺳﺎﺯﮔﺎﺭﻱ ﺩﺍﺩﻩ ﻫﺎ ﭘﻴﺶ ﺁﻳﺪ ‪.‬‬
‫ﺭﺍﻩ ﺣﻞ ﻣﻘﺎﺑﻠﻪ ﺑﺎ ﺍﻳﻦ ﻣﺸﻜﻞ ﻧﻈﺎﺭﺕ ﺑﺮ ﺍﺟﺮﺍﻱ ﻫﻤﺮﻭﻧﺪ ﺗﺮﺍﻛﻨﺸﻬﺎ ﺍﺳﺖ ‪.‬‬

‫ﻧﻜﺎﺕ ﺁﺧﺮ ﻓﺼﻞ‪:‬‬

‫‪ ‬ﺳﻪ ﻧﻮﻉ ﻧﺤﻮﻩ ﺩﺳﺘﺮﺳﻲ ﺑﻪ ﺩﺍﺩﻩ ﻫﺎ ﻭﺟﻮﺩ ﺩﺍﺭﺩ‪Public, Protected, Private :‬‬
‫‪ DBMS ‬ﺳﺮﻭﻳﺲ ﻫﺎ ﻱﻱ ﺟﻬﺖ ﺩﺳﺘﺮﺳﻲ ﺩﺍﺩﻩ ﺩﺭ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻓﺮﺍﻫﻢ ﻣﻲ ﻛﻨﺪ ﺑﻨﺤﻮﻱ ﻛﻪ ﺍﺯ‬
‫ﻛﻠﻲ‪‬ﻩ ﺧﻮﺍﺹ ﺩﺍﺩﻩ ﻣﺤﺎﻓﻈﺖ ﺷﻮﺩ ‪.‬‬
‫‪ ‬ﻛﻠﻴﺪﻫﺎ‪:‬‬
‫‪ Primary Key‬ﻳﺎ ﻛﻠﻴﺪ ﺍﺻﻠﻲ‬ ‫‪.1‬‬
‫ﻛﻠﻴﺪ ﺧﺎﺭﺟﻲ‪ :‬ﻛﻪ ﺩﺭ ﺟﺪﺍﻭﻝ ﺩﻳﮕﺮ ﻧﻘﺶ ﻛﻠﻴﺪ ﺍﺻﻠﻲ ﺭﺍ ﺩﺍﺭﺩ ‪.‬‬ ‫‪.2‬‬
‫‪ :Alternative Key‬ﻛﻠﻴﺪﻱ ﺍﺳﺖ ﻛﻪ ﻣﻴﺘﻮﺍﻥ ﺁﻧﺮﺍ ﺑﺮ ﺣﺴﺐ ﻳﻚ ﻣﻌﻴﺎﺭﻱ ﻣﺜﻞ ﻣﻌﺪﻝ‬ ‫‪.3‬‬
‫ﻃﺒﻘﻪ ﺑﻨﺪﻱ ﻛﺮﺩ‪.‬‬
‫‪ :Super Key‬ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺻﻔﺎﺕ ﻛﻪ ﺧﺎﺻﻴ‪‬ﺖ ﻛﻠﻴﺪ ﺭﺍ ﺩﺍﺭﺩ ﻭﻟﻲ ﻛﻤﻴﻨﻪ ﻧﻴﺴﺖ ﻳﻌﻨﻲ‬ ‫‪.4‬‬
‫ﺗﺎ ﺯﻣﺎﻧﻴﻜﻪ ﺻﻔﺎﺕ ﻫﻤﺮﺍﻩ ﺁﻧﺮﺍ ﺣﺬﻑ ﻛﻨﻴﻢ ﻫﻨﻮﺯ ﺻﻔﺖ ﻣﻨﺤﺼﺮ ﺑﻔﺮﺩﻱ ﺭﺍ ﺩﺍﺭﺩ ‪.‬‬

‫‪18‬‬
 ‫‪ ‬ﺑﺮﻗﺮﺍﺭﻱ ﺟﺎﻣﻌ‪‬ﻴﺖ ﻭﻇﻴﻔﻪ ‪ DBMS‬ﻣﻴﺒﺎﺷﺪ ﻛﻪ ﻣﻲ ﺗﻮﺍﻥ ﺁﻧﺮﺍ ﺑﻪ ﻣﻮﺍﺭﺩ ﺯﻳﺮ ﺩﺳﺘﻪ ﺑﻨﺪﻱ ﻛﺮﺩ‪:‬‬
‫‪ .1‬ﺩﺍﻣﻨﻪ ﺍﻱ( ﺩﺭ ﺍﻳﻨﺠﺎ ﻛﻠﻴ‪‬ﻪ ﺻﻔﺎﺕ ﺩﺭ ﺗﻤﺎﻣﻲ ﺭﺍﺑﻄﻪ ﻫﺎ ﻫﻤﻨﻮﻉ ﺩﺍﻣﻨﻪ ﻣﻴﺒﺎﺷﺪ ‪.‬‬
‫‪ .2‬ﺍﺭﺟﺎﻋﻲ( ﺍﻳﻨﺠﺎ ﻛﻠﻴﺪ ﺧﺎﺭﺟﻲ ﺑﺎﻳﺪ ﺩﺭﺳﺖ ﺑﺎﺷﺪ ﻭ ﺗﻬﻲ ﻧﺒﺎﺷﺪ ‪.‬‬
‫‪ .3‬ﺩﺭﻭﻥ ﺭﺍﺑﻄﻪ ﺍﻱ ( ﺍﻳﻨﺠﺎ ﻧﺒﺎﻳﺪ ﻋﻀﻮ ﺗﻜﺮﺍﺭﻱ ﻣﻮﺟﻮﺩ ﺑﺎﺷﺪ‪ ،‬ﻫﻤﭽﻨﻴﻦ ﻛﻠﻴﺪﻫﺎ ‪Null‬‬
‫ﻧﺒﺎﺷﺪ‪.‬‬
‫‪ ‬ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺍﺯ ﺩﻳﺪ ﻛﺎﺭﺑﺮﺍﻥ‪:‬‬
‫‪- 1‬ﻫﺪﻑ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺍﻳﻦ ﺍﺳﺖ ﻛﻪ ﺍﺯ ﺗﻚ ﻛﺎﺭﺑﺮﻩ ﺑﻪ ﭼﻨﺪﻛﺎﺭﺑﺮﻩ ﺑﺮﺳﺪ ‪ .‬ﺑﺪﻳﻦ ﻣﻌﻨﻲ ﻛﻪ‬
‫ﻫﺮ ﻛﺎﺭﺑﺮ ﺣﺲ ﻛﻨﺪ ﻓﻘﻂ ﺧﻮﺩﺵ ﺩﺭ ﺣﺎﻝ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺍﺳﺖ ﻭ ﻭﺟﻮﺩ‬
‫ﺳﺎﻳﺮ ﻛﺎﺭﺑﺮﺍﻥ ﺭﺍ ﺣﺲ ﻧﻜﻨﺪ‪.‬‬
‫‪- 2‬ﻫﻴﭻ ﻛﺪﺍﻡ ﺍﺯ ﻛﺎﺭﺑﺮﺍﻥ ﺣﻴﻦ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺑﺎ ﻳﻜﺪﻳﮕﺮ ﺗﻀّﺎﺩ )‪(Conflict‬‬
‫ﻧﺪﺍﺷﺘﻪ ﺑﺎﺷﻨﺪ‪.‬‬

‫‪19‬‬
 ‫ﻓﺼﻞ ‪:2‬‬
‫ﺗﺮﺍﻛﻨﺶ ﻫﺎ‬

‫‪20‬‬
 ‫ﻣﻔﻬﻮﻡ ﺗﺮﺍﻛﻨﺶ‪:‬‬

‫ﺗﺮﺍﻛﻨﺶ )‪ (Transaction‬ﻳﻚ ﺑﺮﻧﺎﻣﻪ ﻓﻌﺎﻝ ﺍﺳﺖ ﻛﻪ ﺩﻧﺒﺎﻟﻪ ﺍﻱ ﺍﺯ ﺩﺳﺘﻮﺭﺍﺕ ﺭﺍ ﺷﺎﻣﻞ ﻣﻲ ﺷﻮﺩ ﻭ ﺑﻪ‬
‫ﻃﻮﺭ ﺧﺎﺹ ﺑﻌﻀﻲ ﻋﻤﻠﻴﺎﺕ ﺁﻥ ﺭﻭﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺍﺳﺖ ‪.‬‬

‫ﺗﺮﺍﻛﻨﺶ ﻭﺍﺣﺪﻱ ﺍﺯ ﺑﺮﻧﺎﻣﻪ ﺍﺟﺮﺍﻳﻲ ﺍﺳﺖ ﻛﻪ ﺁﻳﺘﻢ ﻫﺎﻱ ﻣﺨﺘﻠﻒ ﺩﺍﺩﻩ ﺭﺍ ﺑﺎﺯﻳﺎﺑﻲ ﻳﺎ ﺑﺮﻭﺯ ﺭﺳﺎﻧﻲ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫• ﺗﺮﺍﻛﻨﺶ ﺑﺎﻳﺪ ﻳﻚ ﺩﺍﺩﻩ ﺭﺍ ﺳﺎﺯﮔﺎﺭ ﺑﺒﻴﻨﺪ‪.‬‬

‫• ﺩﺭ ﻃﻮﻝ ﺍﺟﺮﺍﻱ ﺗﺮﺍﻛﻨﺶ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻣﻤﻜﻦ ﺍﺳﺖ ﺑﻪ ﻃﻮﺭ ﻣﻮﻗﺖ ﻧﺎﺳﺎﺯﮔﺎﺭ ﺑﺎﺷﺪ ‪.‬‬
‫• ﻫﻨﮕﺎﻣﻲ ﻛﻪ ﺗﺮﺍﻛﻨﺶ ﺑﺎ ﻣﻮﻓﻘﻴﺖ ﺗﻜﻤﻴﻞ ﺷﺪ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺑﺎﻳﺪ ﺳﺎﺯﮔﺎﺭ ﺑﺎﺷﺪ ‪.‬‬
‫• ﭘﺲ ﺍﺯ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﻣﻮﻓﻘﻴﺖ ﺁﻣﻴﺰ ﺗﻐﻴﻴﺮﺍﺕ ﺁﻥ ﺩﺭ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺑﺎﻗﻲ ﺑﻤﺎﻧﺪﺣﺘﻲ ﺍﮔﺮ ﺧﻄﺎ ﺩﺭ‬
‫ﺳﻴﺴﺘﻢ ﻭﺟﻮﺩ ﺩﺍﺭﺩ‪.‬‬
‫• ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﭼﻨﺪ ﮔﺎﻧﻪ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﺑﻪ ﻃﻮﺭ ﻣﻮﺍﺯﻱ ﺑﺎ ﻫﻢ ﺍﺟﺮﺍ ﺷﻮﻧﺪ ‪.‬‬

‫ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﻛﺎﺭﺑﺮﺍﻥ ﺑﻪ ﻋﻨﻮﺍﻥ ﺗﺮﺍﻛﻨﺶ ﺑﻪ ﺳﻴﺴﺘﻢ ﻣﺪﻳﺮﻳﺖ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺕ ﺗﺤﻮﻳﻞ ﻣﻲ ﺷﻮﺩ‪ ،‬ﺍﻳﻦ‬
‫ﺳﻴﺴﺘﻢ ﭼﻬﺎﺭ ﻛﻨﺘﺮﻝ ﻣﻮﺳﻮﻡ ﺑﻪ ‪ ASID‬ﺭﺍ ﺭﻭﻱ ﺁﻧﻬﺎ ﺍﻋﻤﺎﻝ ﻭ ﺩﺭ ﻧﻬﺎﻳﺖ ﺍﻳﻦ ﺑﺮﻧﺎﻣﻪ ﻫﺎ ﻳﺎ ﺑﻪ‬
‫ﺧﻮﺑﻲ ﺍﺟﺮﺍ ﺷﺪﻩ ﻭﭘﺎﻳﺎﻥ ﻣﻲ ﻳﺎﺑﻨﺪ ﻛﻪ ﺑﻪ ﺍﻳﻦ ﺣﺎﻟﺖ ﻣﻲ ﺷﻮﻧﺪ ﺍﻧﺠﺎﻡ ﻳﺎ ﺗﺜﺒﻴﺖ )‪ (commit‬ﻣﻲ‬
‫ﮔﻮﻳﻨﺪ ﻭ ﻳﺎ ﺍﻳﻨﻜﻪ ﺳﺎﻗﻂ )‪ (abort‬ﻣﻲ ﺷﻮﻧﺪ‪.‬‬
‫ﭘﺲ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﻋﻤﻠﮕﺮﻫﺎﻱ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺕ ﻛﻪ ﺍﺯ ﺩﻳﺪ ﻛﺎﺭﺑﺮ ﻳﻚ ﻭﺍﺣﺪ ﻣﻨﻄﻘﻲ ﻛﺎﺭ ﺭﺍ ﺗﺸﻜﻴﻞ‬
‫ﻣﻲ ﺩﻫﻨﺪ‪ ،‬ﺗﺮﺍﻛﻨﺶ ﻧﺎﻡ ﺩﺍﺭﺩ‪.‬‬
‫ﻣﺜﻼ ﻭﻗﺘﻲ ﺍﺯ ﻳﻚ ﺣﺴﺎﺏ ‪ A‬ﻣﺒﻠﻐﻲ ﺭﺍ ﺑﻪ ﺣﺴﺎﺏ ‪ B‬ﻣﻨﺘﻘﻞ ﻣﻲﻛﻨﺪ‪ ،‬ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺍﺳﺖ ﻛﻪ ﺷﺎﻣﻞ‬
‫ﺩﻭ ﻛﺎﺭ ﺍﺳﺖ‪:‬‬
‫‪- 1‬ﻣﺒﻠﻎ ﻣﻮﺭﺩ ﻧﻈﺮ ﺍﺯ ﺣﺴﺎﺏ ‪ A‬ﻛﻢ ﺷﻮﺩ‬
‫‪- 2‬ﻫﻤﺎﻥ ﻣﺒﻠﻎ ﺑﻪ ﺣﺴﺎﺏ ‪ B‬ﺍﺿﺎﻓﻪ ﺷﻮﺩ‪.‬‬

‫‪21‬‬
 ‫ﺑﺮﺭﺳﻲ ﺧﻮﺍﺹ ‪:ACID‬‬
‫‪.‬‬ ‫ﺍﮔﺮ ﺍﻳﻦ ﺧﻮﺍﺹ ﺭﺍ ﺭﻭﻱ ﺗﻤﺎﻡ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺍﻋﻤﺎﻝ ﻛﻨﻴﻢ‪ ،‬ﺟﺎﻣﻌﻴﺖ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺣﻔﻆ ﻣﻲ ﺷﻮﺩ‬
‫ﺑﻨﺎﺑﺮﺍﻳﻦ ﻫﺪﻑ ﺍﺻﻠﻲ ﺩﺭ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ‪ ،‬ﺣﻔﻆ ﺟﺎﻣﻌﻴﺖ ﺍﺳﺖ ‪ .‬ﺍﻳﻦ ﺧﻮﺍﺹ ﻋﺒﺎﺭﺗﻨﺪ ﺍﺯ‪:‬‬
‫‪ Atomicity .1‬ﻳﻜﭙﺎﺭﭼﮕﻲ‬
‫‪ Consistency .2‬ﻫﻤﺨﻮﺍﻧﻲ ﻳﺎ ﺳﺎﺯﮔﺎﺭﻱ‬
‫‪ Isolation .3‬ﺍﻳﺰﻭﻟﮕﻲ‬
‫‪ Durability .4‬ﭘﺎﻳﺪﺍﺭﻱ ﻳﺎ ﭘﺎﻳﺎﻳﻲ‬
‫‪ :Atomicity‬ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺍﺗﻤﻴﻚ ﻫﺴﺘﻨﺪ ﻳﺎ ﺍﺻﻼ ﺷﺮﻭﻉ ﻧﻤﻲ ﺷﻮﻧﺪ ﻳﺎ ﻭﻗﺘﻲ ﺁﻏﺎﺯ ﺷﺪﻧﺪ ﺣﺘﻤﺎ ﺑﻪ‬
‫ﭘﺎﻳﺎﻥ ﻣﻲ ﺭﺳﻨﺪ ‪ .‬ﻳﺎ ﻫﻤﻪ ﺩﺳﺘﻮﺭﺍﺕ ﺑﻪ ﻃﻮﺭ ﻛﺎﻣﻞ ﺍﻧﺠﺎﻡ ﻣﻲ ﺷﻮﺩ ﻳﺎ ﻫﻴﭽﻜﺪﺍﻡ ‪ .‬ﻧﮕﻬﺪﺍﺷﺘﻦ ﺧﺎﺻﻴﺖ‬
‫ﺍﺗﻮﻣﻴﻚ ﺑﻮﺩﻥ ﺑﻪ ﻋﻬﺪﻩ ﻛﻨﺘﺮﻝ ﻫﻤﺮﻭﻧﺪﻱ ﻭ ﺗﺮﻣﻴﻢ ﺍﺳﺖ ‪ .‬ﭘﺲ ﺍﺗﻮﻣﻴﻚ ﺑﻮﺩﻥ ﺗﺮﺍﻛﻨﺶ ﻳﻌﻨﻲ ﺍﻳﻨﻜﻪ ﻳﻚ‬
‫ﺗﺮﺍﻛﻨﺶ ﻳﺎ ﺑﻄﻮﺭ ﻛﺎﻣﻞ ﺍﺟﺮﺍ ﺷﻮﺩ ﻳﺎ ﺍﺻﻼ ﺍﻧﺠﺎﻡ ﻧﺸﻮﺩ ‪.‬‬
‫‪ :Consistency‬ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﻳﺎ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺭﺍ ﺑﻪ ﺣﺎﻟﺖ ﺳﺎﺯﮔﺎﺭ ﺟﺪﻳﺪﻱ ﻣﻲ ﺑﺮﺩ ﻳﺎ ﺍﮔﺮ ﺷﻜﺴﺘﻲ ﺭﺥ‬
‫ﺩﺍﺩ ﻛﻠﻴﻪ ﺩﺍﺩﻩ ﻫﺎ ﺑﻪ ﺣﺎﻟﺖ ﻗﺒﻞ ﺍﺯ ﺷﺮﻭﻉ ﺗﺮﺍﻛﻨﺶ ﺑﺮﻣﻲ ﮔﺮﺩﻧﺪ ‪ .‬ﺍﮔﺮ ﺗﺮﺍﻛﻨﺸﻲ ﺗﺜﺒﻴﺖ ﺷﺪ‪ ،‬ﻧﺒﺎﻳﺪ ﺑﺎﻧﻚ‬
‫ﺍﻃﻼﻋﺎﺕ ﺭﺍ ﺧﺮﺍﺏ ﻛﻨﺪ‪ .‬ﻣﺜﻼً ﺍﮔﺮ ﺗﺮﺍﻛﻨﺸﻲ ﻧﻤﺮﻩ ﺍﻱ ﺭﺍ ﺗﻐﻴﻴﺮ ﺩﺍﺩ‪ ،‬ﺑﺎﻳﺪ ﺁﻧﺮﺍ ﺑﻪ ﻳﻚ ﻣﻘﺪﺍﺭ ﻣﺠﺎﺯ ﺗﺒﺪﻳﻞ‬
‫ﻛﻨﺪ‪.‬‬
‫‪ :Isolation‬ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﻫﻤﺮﻭﻧﺪ ﻳﺎ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻳﻲ ﻛﻪ ﻃﻮﻝ ﻋﻤﺮﺷﺎﻥ ﻫﻤﭙﻮﺷﺎﻧﻲ ﺩﺍﺭﺩ‪ ،‬ﻧﺒﺎﻳﺪ ﺗﺄﺛﻴﺮ‬
‫ﻣﺨﺮﺏ ﺭﻭﻱ ﻫﻢ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻨﺪ‪ .‬ﺗﻮﺟﻪ ﻛﻨﻴﺪ ﻛﻪ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﻣﺨﺘﻠﻒ ﺍﺯ ﻭﺟﻮﺩ ﻫﻤﺪﻳﮕﺮ ﺑﻲ ﺧﺒﺮ‬
‫ﻫﺴﺘﻨﺪ‪.‬‬
‫‪ :Durability‬ﺍﮔﺮ ﺗﺮﺍﻛﻨﺸﻲ ﺗﺜﺒﻴﺖ ﺷﺪ‪ ،‬ﺗﺄﺛﻴﺮ ﺁﻥ ﺑﻪ ﻃﻮﺭ ﺍﺗﻔﺎﻗﻲ ﺍﺯ ﺑﻴﻦ ﻧﺨﻮﺍﻫﺪ ﺭﻓﺖ‪ .‬ﻳﻌﻨﻲ‬
‫ﺗﺮﺍﻛﻨﺶ ﺗﺜﺒﻴﺖ ﺷﺪﻩ ﺗﺎﺛﻴﺮﺵ ﺩﺍﺋﻤﻲ ﺍﺳﺖ‪ ،‬ﺣﺘﻲ ﺍﮔﺮ ﺳﻴﺴﺘﻢ ﺧﺮﺍﺏ ﺷﻮﺩ ﺩﺍﺩﻩ ﺩﺭ ﺣﺎﻟﺖ ﺩﺭﺳﺖ ﺑﺎﻗﻲ‬
‫ﻣﻲ ﻣﺎﻧﺪ‪.‬‬

‫‪22‬‬
 ‫ﻭﺿﻌﻴﺖ ﺗﺮﺍﻛﻨﺶ‪:‬‬

‫ﻣﺜﺎﻟﻲ ﺑﺮﺍﻱ ﭼﻬﺎﺭ ﺧﺎﺻﻴﺖ ﺗﺮﺍﻛﻨﺶ‪:‬‬

‫ﻓﺮﺽ ﻛﻨﻴﻢ ﻣﻲ ﺧﻮﺍﻫﻴﻢ ‪50‬ﺩﻻﺭ ﺍﺯ ﺣﺴﺎﺏ‪ A‬ﻛﺎﺳﺘﻪ ﻭ ﺑﻪ ﺣﺴﺎﺏ‪ B‬ﺍﺿﺎﻓﻪ ﻛﻨﻴﻢ‪(A=200,B=100) .‬‬

‫‪ :Atomicity .1‬ﺍﮔﺮ ﺧﻄﺎﻳﻲ ﺑﻌﺪ ﺍﺯ ﺧﻂ‪ 3‬ﻳﺎ ﻗﺒﻞ ﺍﺯ ﺧﻂ‪ 6‬ﺍﻧﺠﺎﻡ ﺷﻮﺩ‪ ،‬ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺍﺯ ﺣﺎﻟﺖ‬
‫ﺳﺎﺯﮔﺎﺭ ﺧﺎﺭﺝ ﻣﻲ ﺷﻮﺩ‪ ،‬ﺯﻳﺮﺍ ﺑﺎﻳﺪ ﺣﺎﺻﻞ‪ A+B‬ﺩﺭ ﻗﺒﻞ ﻭ ﺑﻌﺪ ﺍﺯ ﺗﺮﺍﻛﻨﺶ ﻣﻘﺪﺍﺭﻱ ﻳﻜﺴﺎﻥ‬
‫ﺧﻂ‪ 3‬ﺧﻄﺎ ﺭﺥ ﺩﻫﺪ )ﻳﻌﻨﻲ ‪ A+B=150+100‬ﺣﺎﺻﻞ ﺷﻮﺩ (‬ ‫ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ‪ ،‬ﺍﮔﺮ ﺑﻌﺪ ﺍﺯ‬

‫‪23‬‬
‫ﺩﺳﺘﻮﺭﺍﺕ ﺧﻂ‪ ،4‬ﺧﻂ‪ 5‬ﻭ ﺧﻂ‪ 6‬ﺍﺟﺮﺍ ﻧﺨﻮﺍﻫﻨﺪ ﺷﺪ ﻭ ﻧﺎﺳﺎﺯﮔﺎﺭﻱ ﭘﺪﻳﺪ ﻣﻲ ﺁﻳﺪ ‪) .‬ﺩﺭ ﻭﺍﻗﻊ ﺩﺭ‬
‫ﺧﺎﺻﻴﺖ ﺍﺗﻮﻣﻴﻚ ﺑﻮﺩﻥ ﺑﻴﺎﻥ ﻣﻲ ﺷﻮﺩ ﻛﻪ ﻳﺎ ﻫﻤﻪ ﻱ ﺩﺳﺘﻮﺭﺍﺕ ﺑﺎﻳﺪ ﺍﺟﺮﺍ ﺷﻮﻧﺪ ﻳﺎ ﻫﻴﭻﻛﺪﺍﻡ (‪.‬‬
‫‪ :Consistency .2‬ﭼﻮﻥ ﻣﻘﺪﺍﺭ ‪ A+B‬ﻗﺒﻞ ﺍﺯ ﺗﺮﺍﻛﻨﺶ ‪ 200+100=300‬ﻣﻲ ﺷﻮﺩ ﻭ ﺑﻌﺪ ﺍﺯ‬
‫ﺁﻥ ﻫﻢ ‪ 150+150=300‬ﻣﻲ ﺷﻮﺩ‪ ،‬ﭘﺲ ﺳﺎﺯﮔﺎﺭﻱ ﻭﺟﻮﺩ ﺩﺍﺭﺩ‪.‬‬
‫‪ :Isolation .3‬ﺍﮔﺮ ﺑﻴﻦ ﺧﻂ‪ 3‬ﻭ ﺧﻂ ‪ ،6‬ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺩﻳﮕﺮ ﺑﻴﺎﻳﺪ ﻭ ﺍﺯ ﺩﺍﺩﻩ ﻫﺎﻱ ﺑﻴﻦ ﺍﻳﻦ ﺩﻭ‬
‫ﺧﻂ ﺍﺳﺘﻔﺎﺩﻩ ﻛﻨﺪ‪ ،‬ﺩﺭ ﺍﻳﻦ ﺻﻮﺭﺕ ﻧﺎﺳﺎﺯﮔﺎﺭﻱ ﺩﺭ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻣﺸﺎﻫﺪﻩ ﻣﻲ ﺷﻮﺩ ‪ .‬ﺩﺭ ﺣﻘﻴﻘﺖ‬
‫ﺧﺎﺻﻴﺖ ﺍﻳﺰﻭﻟﮕﻲ ﺍﻳﻦ ﺍﻃﻤﻴﻨﺎﻥ ﺭﺍ ﺣﺎﺻﻞ ﻣﻲ ﻛﻨﺪ ﻛﻪ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺑﻪ ﺻﻮﺭﺕ ﺳﺮﻱ ﺍﺟﺮﺍ ﺷﻮﻧﺪ ‪.‬‬
‫‪ :Durability .4‬ﺑﻌﺪ ﺍﺯ ﺍﻳﻨﻜﻪ ﻫﻤﻪ ﻱ ﺩﺱﺗﻮﺭﺍﺕ ﺗﺮﺍﻛﻨﺶ ﺍﺟﺮﺍ ﺷﺪﻧﺪ‪ ،‬ﺗﻤﺎﻡ ﺗﻐﻴﻴﺮﺍﺕ ﺩﺭ ﭘﺎﻳﮕﺎﻩ‬
‫ﺩﺍﺩﻩ ﺍﻋﻤﺎﻝ ﻣﻲ ﺷﻮﻧﺪ‪) ،‬ﺑﻌﺪ ﺍﺯ ﺍﻳﻨﻜﻪ ﭼﻚ ﺷﺪ ﻛﻪ ﺩﺭ ﺗﺮﺍﻛﻨﺶ ﺧﻄﺎﻳﻲ ﻭﺟﻮﺩ ﻧﺪﺍﺭﺩ‪ ،‬ﺁﻥ‬
‫ﺗﺮﺍﻛﻨﺶ ﺑﺎ ﻣﻮﻓﻘﻴﺖ ﺗﻤﺎﻡ ﺷﺪﻩ ﻭ ﺑﻌﺪ ﺍﺯ ﺁ ﻥ ﺑﺮ ﺭﻭﻱ ﺩﻳﺴﻚ ﺫﺧﻴﺮﻩ ﻣﻲ ﺷﻮﺩ(‪.‬‬

‫ﺍﺯ ﺩﺳﺘﻮﺭﺍﺕ ﻣﻬ‪‬ﻢ ﺩﺭ ﺗﺮﺍﻛﻨﺶ ﻣﻲ ﺗﻮﺍﻥ ﺑﻪ ﻣﻮﺍﺭﺩ ﺯﻳﺮ ﺍﺷﺎﺭﻩ ﻛﺮﺩ‪:‬‬

‫‪ :Commit‬ﺫﺧﻴﺮﻩ ﺗﻤﺎﻡ ﺗﻐﻴﻴﺮﺍﺕ ﺑﺪﺭﺳﺘﻲ ﻭ ﻛﺎﻣﻞ ‪ .‬ﺍﻳﻦ ﺩﺳﺘﻮﺭ ﺍﻧﺘﻬﺎﻱ ﻣﻮﻓّﻘﻴﺖ ﺁﻣﻴﺰ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺭﺍ‬
‫ﺑﻪ ﻣﺪﻳﺮ ﺗﺮﺍﻛﻨﺶ ﺍﻋﻼﻡ ﻣﻲ ﻛﻨﺪ ﻟﺬﺍ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺩﺭ ﺣﺎﻟﺘﻲ ﭘﺎﻳﺪﺍﺭ ﻗﺮﺍﺭ ﺩﺍﺭﺩ ﻭ ﺗﻤﺎﻣﻲ ﺍﻋﻤﺎﻝ ﺍﻧﺠﺎﻡ ﺷﺪﻩ‬
‫ﺗﻮﺳﻂ ﺗﺮﺍﻛﻨﺶ ﻗﺒﻮﻝ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬

‫‪ :Roll back‬ﺑﺎﺯﻳﺎﺑﻲ ﺗﻐﻴﻴﺮﺍﺕ ﺑﻪ ﻋﻠّﺖ ﻋﺪﻡ ﺩﺭﺳﺘﻲ ﻭ ﻧﻘﺺ ‪ .‬ﺍﻳﻦ ﺩﺳﺖﻭﺭ ﺍﻧﺘﻬﺎﻱ ﻧﺎﻣﻮ‪‬ﻓﻖ ﻳﻚ ﺗﺮﺍﻛﻨﺶ‬
‫ﺭﺍ ﺑﻪ ﻣﺪﻳﺮ ﺗﺮﺍﻛﻨﺶ ﺍﻋﻼﻡ ﻣﻲ ﻛﻨﺪ ﻟﺬﺍ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺩﺭ ﺣﺎﻟﺘﻲ ﻧﺎﭘﺎﻳﺪﺍﺭ ﻗﺮﺍﺭ ﺩﺍﺭﺩ ﻭ ﺗﻤﺎﻣﻲ ﺍﻋﻤﺎﻝ ﺍﻧﺠﺎﻡ‬
‫ﺷﺪﻩ ﺗﻮﺳﻂ ﺍﻳﻦ ﺗﺮﺍﻛﻨﺶ ﺑﺎﻳﺴﺘﻲ ﻟﻐﻮ ﮔﺮﺩﺩ‪.‬‬

‫*ﺗﻮﺟﻪ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻴﺪ ﻛﻪ ﺍﮔﺮ ﺗﺮﺍﻛﻨﺸﻲ ﺳﺎﻗﻂ ﺷﻮﺩ‪ ،‬ﻣﻲ ﺗﻮﺍﻥ ﺁﻥ ﺭﺍ ﻣﺠﺪﺩﺍً ﺁﻏﺎﺯ ﻛﺮﺩ ﻭ ﺍﮔﺮ ﺗﺮﺍﻛﻨﺸﻲ ﺁﻣﺎﺩﻩ‬
‫ﺍﻧﺠﺎﻡ ﺷﻮﺩ‪ ،‬ﻣﻤﻜﻦ ﺍﺳﺖ ﺑﺎﺯ ﻫﻢ ﺳﻘﻮﻁ ﻛﻨﺪ‪.‬‬
‫ﺩﻭ ﺭﻭﺵ ﺑﺮﺍﻱ ‪ Abort‬ﻛﺮﺩﻥ ﻭﺟﻮﺩ ﺩﺍﺭﺩ ‪:‬‬
‫‪ Restart .1‬ﻛﺮﺩﻥ‪ :‬ﺯﻣﺎﻧﻲ ﺍﺯ ﺍﻳﻦ ﺭﻭﺵ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺷﻮﺩ ﻛﻪ ﺧﻄﺎﻱ ﻣﻨﻄﻘﻲ ﻭﺟﻮﺩ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ ‪.‬‬

‫‪24‬‬
 ‫‪ Kill .2‬ﻛﺮﺩﻥ‪ :‬ﺯﻣﺎﻧﻲ ﺍﺯ ﺍﻳﻦ ﺭﻭﺵ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺷﻮﺩ ﻛﻪ ﺧﻄﺎﻱ ﺳﻴﺴﺘﻤﻲ ﻭﺟﻮﺩ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ ‪.‬‬

‫ﺷﻤﺎﻱ ‪: shadow data base‬‬

‫ﻓﺮﺽ ﺍﻭﻝ‪ :‬ﻛﻪ ﺩﺭ ﻫﺮ ﻟﺤﻈﻪ ﻓﻘﻂ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺩﺭ ﺣﺎﻟﺖ ﺍﺟﺮﺍ ﺍﺳﺖ ‪.‬‬

‫ﻓﺮﺽ ﺩﻭﻡ ‪ :‬ﻳﻚ ﺍﺷﺎﺭﻩ ﮔﺮ ﺑﻪ ﻧﺎﻡ ‪ db-pointer‬ﻭﺟﻮﺩ ﺩﺍﺭﺩ ﻛﻪ ﺑﻪ ﺁﺧﺮﻳﻦ ﻛﭙﻲ ﺩﺭﺳﺖ ﺍﺯ ﭘﺎﻳﮕﺎﻩ‬
‫ﺩﺍﺩﻩ ﻛﻪ ﺩﺭ ﺣﺎﻟﺖ ﺳﺎﺯﮔﺎﺭﻱ ﺍﺳﺖ‪ ،‬ﺍﺷﺎﺭﻩ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫ﻗﺒﻞ ﺍﺯ ﺍﻳﻨﻜﻪ ﺑﺮﻭﺯ ﺭﺳﺎﻧﻲ ﺍﻧﺠﺎﻡ ﺷﻮﺩ ﻳﻚ ﻛﭙﻲ ﺍﺯ ﺍﻃﻼﻋﺎﺕ ﻣﻮﺟﻮﺩ ﺩﺭ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﮔﺮﻓﺘﻪ ﻣﻲ ﺷﻮﺩ‬
‫ﻭ ‪ db-pointer‬ﺑﻪ ﺍﻳﻦ ﻛﭙﻲ ﺍﺷﺎﺭﻩ ﻣﻲ ﻛﻨﺪ ‪ .‬ﺁﻧﺮﺍ ‪ shadow‬ﻣﻲ ﻧﺎﻣﻴﻢ ‪ .‬ﺑﻌﺪ ﺍﺯ ﺍﺗﻤﺎﻡ ﺗﺮﺍﻛﻨﺶ‬
‫ﺗﻐﻴﻴﺮﺍﺕ ﺭﺍ ﺩﺭ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺍﻋﻤﺎﻝ ﻣﻲ ﺷﻮﺩ ﻭ ﺍﮔﺮ ﺗﺮﺍﻛﻨﺶ ﺩﻳﮕﺮﻱ ﺑﻌﺪ ﺍﺯ ﺁﻥ ﺍﻧﺠﺎﻡ ﺷﻮﺩ‪ ،‬ﻣﺠﺪﺩﺍ ﺍﺯ‬
‫ﺁﺧﺮﻳﻦ ﺣﺎﻟﺖ ﺳﺎﺯﮔﺎﺭ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻳﻚ ﻛﭙﻲ ﺟﺪﻳﺪ ﮔﺮﻓﺘﻪ ﻭ ﺩﺭ‪ shadow‬ﻗﺮﺍﺭ ﻣﻲ ﮔﻴﺮﺩ‪ .‬ﻫﻤﻪ ﻱ‬
‫ﺑﺮﻭﺯ ﺭﺳﺎﻧﻲ ﻫﺎ ﺑﺮﻭﻱ ‪ shadow copy‬ﺍﻧﺠﺎﻡ ﻣﻲ ﺷﻮﻧﺪ ﻭ ﺍﺷﺎﺭﻩ ﮔﺮ ‪ db-pointer‬ﺑﻪ‬
‫‪ shadow copy‬ﺍﻱ ﻛﻪ ﺑﻪ ﻣﺮﺣﻠﻪ ‪ partially committed‬ﺭﺳﻴﺪﻩ ﺍﺳﺖ ﺍﺷﺎﺭﻩ ﻣﻲ ﻛﻨﺪ ‪.‬‬
‫ﻫﻤﻪ ﻱ ﺍﻳﻦ ﺗﻐﻴﻴﺮﺍﺕ ﺩﺭ ﻧﻬﺎﻳﺖ ﭘﺲ ﺍﺯ ﺍﺗﻤﺎﻡ ﺗﺮﺍﻛﻨﺶ ﺭﻭﻱ ﺩﻳﺴﻚ ﺫﺧﻴﺮﻩ ﻣﻲ ﺷﻮﻧﺪ‪.‬‬

‫ﻧﻜﺘﻪ‪ :‬ﺍﻳﻦ ﺭﻭﺵ ﺑﺪﻟﻴﻞ ﺍﻳﻨﻜﻪ ﺗﻌﺪﺍﺩ ﺯﻱﺍﺩﻱ ﻛﭙﻲ ﮔﺮﻓﺘﻪ ﻣﻲ ﺷﻮﺩ‪ ،‬ﺭﻭﺵ ﻛﺎﺭﺁﻣﺪ ﻭ ﻣﻔﻴﺪﻱ ﻧﻴﺴﺖ ‪.‬‬

‫ﺍﺟﺮﺍﻫﺎﻱ ﻫﻤﺮﻭﻧﺪ )ﻫﻤﺰﻣﺎﻥ(‪ :‬ﺩﺭ ﺍﻳﻦ ﻣﺪﻝ ‪ ،‬ﭼﻨﺪ ﺗﺮﺍﻛﻨﺶ ﺩﺭ ﻳﻚ ﺳﻴﺴﺘﻢ ﺍﺟﺎﺯﻩ ﺩﺍﺭﻧﺪ ﺑﻪ‬
‫ﻃﻮﺭ ﻫﻤﺰﻣﺎﻥ )ﺩﺭ ﻳﻚ ﻟﺤﻈﻪ( ﺍﺟﺮﺍ ﺷﻮﻧﺪ‪.‬‬

‫ﻣﺰﺍﻳﺎﻱ ﺍﻳﻦ ﻣﺪﻝ‪:‬‬

‫‪25‬‬
 ‫‪ .1‬ﺍﻓﺰﺍﻳﺶ ﭘﺮﺩﺍﺯﺷﮕﺮ ﻭ ﺑﻬﺮﻩ ﻭﺭﻱ ﺍﺯ ﺩﻳﺴﻚ‪ :‬ﺍﻳﻦ ﺑﻬﺮﻩ ﻭﺭﻱ ﺑﻪ ﻣﻌﻨﺎﻱ ﺑﺎﻻ ﺑﺮﺩﻥ ﺗﻮﺍﻥ‬
‫ﻋﻤﻠﻴﺎﺗﻲ ﺗﺮﺍﻛﻨﺶ ﺍﺳﺖ‪ .‬ﻳﻌﻨﻲ ﺗﺮﺍﻛﻨﺶ ﻣﻲ ﺗﻮﺍﻧﺪ ﺍﺯ ﻭﺍﺣﺪ ﭘﺮﺩﺍﺯﺷﮕﺮ ﻣﺮﻛﺰﻱ ﺩﺭ‬
‫ﺩﻳﺴﻚ ﻭ ﻳﺎ ﻧﻮﺷﺘﻦ ﺑﺮﻭﻱ‬ ‫ﻟﺤﻈﻪ ﺍﻱ ﻛﻪ ﺳﺎﻳﺮ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺩﺭ ﺣﺎﻝ ﺧﻮﺍﻧﺪﻥ ﺍﺯ‬
‫ﺩﻳﺴﻚ ﻣﻲ ﺑﺎﺷﻨﺪ‪ ،‬ﺍﺳﺘﻔﺎﺩﻩ ﻛﻨﺪ‪.‬‬
‫‪ .2‬ﻛﺎﻫﺶ ﺯﻣﺎﻥ ﭘﺎﺳﺦ ‪ :‬ﺑﻪ ﺍﻳﻦ ﻣﻌﻨﺎ ﺳﺖ ﻛﻪ ﻧﻴﺎﺯﻱ ﻧﻴﺴﺖ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﻛﻮﭼﻚ ﺑﺮﺍﻱ‬
‫ﺍﺟﺮﺍ ﺷﺪﻥ ﺯﻣﺎﻥ ﻃﻮﻻﻧﻲ ﺭﺍ ﻣﻨﺘﻈﺮ ﺑﻤﺎﻧﻨﺪ‪.‬‬

‫ﻋﻴﺐ ﺍﻳﻦ ﺭﻭﺵ‪ :‬ﻭﺟﻮﺩ ﺑﺮﺧﻮﺭﺩ ﻫﺎﻱ ﺯﻳﺎﺩ ﺍﻳﺮﺍﺩ ﺍﻳﻦ ﻣﺪﻝ ﺍﺳﺖ ﻛﻪ ﺑﺎﻳﺪ ﻛﻨﺘﺮﻝ ﺷﻮﻧﺪ‪.‬‬

‫ﺯﻣﺎﻧﺒﻨﺪﻫﺎ )‪:(Schedules‬‬
‫ﺯﻣﺎﻧﺒﻨﺪ ﻳﻚ ﺗﻮﺍﻟﻲ ﺍﺯ ﺩﺳﺘﻮﺭﺍﻟﻌﻤﻞ ﻫﺎﺳﺖ ﻛﻪ ﺗﺮﺗﻴﺐ ﻭﻗﻮﻉ )ﺍﺯ ﻧﻈﺮ ﺯﻣﺎﻧﻲ ( ﺩﺳﺘﻮﺭﺍﻟﻌﻤﻞ ﻫﺎﻱ ﺗﺮﺍﻛﻨﺶ‬
‫ﻫﺎﻱ ﻫﻤﺮﻭﻧﺪ ﺍﺟﺮﺍ ﺷﺪﻩ ﺭﺍ ﺗﻌﺮﻳﻒ ﻭ ﻣﺸﺨﺺ ﻣﻲ ﻧﻤﺎﻳﺪ ‪ .‬ﻳﻚ ﺯﻣﺎﻧﺒﻨﺪ ﺑﺮﺍﻱ ﻳﻚ ﻣﺠﻤﻮﻋﻪ ﺍﺯ ﺗﺮﺍﻛﻨﺶ‬
‫ﻫﺎ ﺑﺎﻳﺪ ﺷﺎﻣﻞ ﻫﻤﻪ ﻱ ﺩﺳﺘﻮﺭﺍﻟﻌﻤﻞ ﻫﺎﻱ ﺁﻥ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺑﺎﺷﺪ ‪ .‬ﻫﻤﭽﻨﻴﻦ ﻳﻚ ﺯﻣﺎﻧﺒﻨﺪ ﺑﺎﻳﺪ ﺗﺮﺗﻴﺐ‬
‫ﺯﻣﺎﻧﻲ ﺩﺳﺘﻮﺍﻟﻌﻤﻞ ﻫﺎﻳﻲ ﺭﺍ ﻛﻪ ﺩﺭ ﻫﺮ ﺗﺮﺍﻛﻨﺶ ﻇﺎﻫﺮ ﺷﺪﻩ ﺭﺍ ﺣﻔﻆ ﻭ ﻧﮕﻬﺪﺍﺭﻱ ﻛﻨﺪ ‪.‬‬
‫ﺍﺟﺮﺍﻱ ﻫﻤﺮﻭﻧﺪ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺭﺍ ﺩﺭ ﻗﺎﻟﺐ ﺳﺎﺧﺘﺎﺭﻱ ﺑﻪ ﻧﺎﻡ ﺯﻣﺎﻧﺒﻨﺪ )‪ (Schedule‬ﺑﺮﺭﺳﻲ ﻣﻲ ﻛﻨﻴﻢ‪.‬‬
‫ﺩﺭ ﻳﻚ ﺯﻣﺎﻧﺒﻨﺪ ﻣﻲ ﺗﻮﺍﻥ ﺗﺮﺍﻛﻨﺸﻲ ﺭﺍ ﻣﻨﺘﻈﺮ ﺗﺮﺍﻛﻨﺶ ﺩﻳﮕﺮﻱ ﮔﺬﺍﺷﺖ ﻳﺎ ﻣﻲ ﺗﻮﺍﻥ ﺍﺟﺮﺍﻱ ﻳﻚ‬
‫ﺗﺮﺍﻛﻨﺶ ﺭﺍ ﺑﻪ ﺗﻌﻮﻳﻖ ﺍﻧﺪﺍﺧﺖ‪ .‬ﺍﻣﺎ ﺩﺭﻭﻥ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﻧﻤﻲ ﺗﻮﺍﻥ ﺩﺳﺘﻮﺭﻫﺎ ﺭﺍ ﺟﺎﺑﻪ ﺟﺎ ﻛﺮﺩ ﺑﻪ ﺩﻟﻴﻞ‬
‫ﺍﻳﻨﻜﻪ ﻣﻨﻄﻖ ﺁﻥ ﺗﺮﺍﻛﻨﺶ ﺑﻪ ﻫﻢ ﻣﻲ ﺭﻳﺰﺩ‪.‬‬
‫ﻣﻲ ﺩﺍﻧﻴﻢ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺍﮔﺮ ﺧﺎﺻﻴﺖ ‪ ACID‬ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ ﻧﺘﻴﺠﻪ ﺍﺵ ﺩﺭﺳﺖ ﺍﺳﺖ ﻭ ﻧﻴﺰ ﺍﮔﺮ ﭼﻨﺪ‬
‫ﺗﺮﺍﻛﻨﺶ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻴﻢ ﻛﻪ ﺳﺮﻳﺎﻟﻲ ﺑﺎﺷﻨﺪ‪ ،‬ﺍﻳﻦ ﺯﻣﺎﻧﺒﻨﺪ ﻛﻪ ﺑﻪ ﺁﻥ ﺯﻣﺎﻧﺒﻨﺪ ﺳﺮﻳﺎﻟﻲ ﻣﻲ ﮔﻮﻳﻴﻢ ﻧﻴﺰ‬
‫ﻧﺘﻴﺠﻪ ﺍﺵ ﺩﺭﺳﺖ ﺍﺳﺖ‪.‬‬
‫ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮ ﻳﻌﻨﻲ ﻣﻌﺎﺩﻝ ﺳﺮﻳﺎﻟﻲ‪ ،‬ﻳﻌﻨﻲ ﺍﮔﺮ ﺯﻣﺎﻧﺒﻨﺪﻱ ﻫﺎﻱ ﻫﻤﺮﻭﻧﺪﻱ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻴﻢ ﻛﻪ ﺳﺮﻳﺎﻝ‬
‫ﻧﻴﺴﺘﻨﺪ ﺍﻣﺎ ﻣﻌﺎﺩﻝ ﺳﺮﻳﺎﻟﻲ ﻩ ﺳﺘﻨﺪ‪ ،‬ﺩﺭ ﺍﻳﻦ ﺻﻮﺭﺕ ﻧﺘﻴﺠﻪ ﻛﺎﺭ ﺩﺭﺳﺖ ﺧﻮﺍﻫﺪ ﺑﻮﺩ ‪.‬‬
‫‪:Schedule 1‬‬
‫ﻓﺮﺽ ﻣﻲ ﻛﻨﻴﻢ ﻛﻪ ﻭﻇﻴﻔﻪ ﻱ ﺗﺮﺍﻛﻨﺶ ‪ T1‬ﺍﻳﻦ ﺍﺳﺖ ﻛﻪ ‪ 50‬ﺩﻻﺭ ﺍﺯ ﺣﺴﺎﺏ ‪ A‬ﺑﻪ ﺣﺴﺎﺏ ‪ B‬ﺍﻧﺘﻘﺎﻝ‬
‫ﺩﻫﺪ‪ ،‬ﻭ ﻭﻇﻴﻔﻪ ﻱ ﺗﺮﺍﻛﻨﺶ ‪ T2‬ﺍﻳﻦ ﺍﺳﺖ ﻛﻪ ﺍﺯ ‪10‬ﺩﺭﺻﺪ ﺍﺯ ﺣﺴﺎﺏ ‪ A‬ﺭﺍ ﺑﻪ ﺣﺴﺎﺏ ‪ B‬ﺍﻧﺘﻘﺎﻝ ﺩﻫﺪ ‪.‬‬
‫ﺍﻳﻦ ﺯﻣﺎﻧﺒﻨﺪ ﺳﺮﻳﺎﻝ ﺍﺳﺖ ﺯﻳﺮﺍ ‪ T2‬ﺑﻌﺪ ﺍﺯ ﺍﺗﻤﺎﻡ ﻛﺎﻣﻞ ‪ T1‬ﺍﺟﺮﺍ ﻣﻲ ﺷﻮﺩ‪.‬‬
‫‪26‬‬
 ‫‪:Schedule 2‬‬

‫ﺍﻳﻦ ﺯﻣﺎﻧﺒﻨﺪ ﻧﻴﺰ ﺳﺮﻳﺎﻝ ﺍﺳﺖ ﻭ ﺗﻔﺎﻭﺗﺶ ﺑﺎ ‪ schedule1‬ﺩﺭ ﺗﻮﺍﻟﻲ ﺍﺟﺮﺍ ﺍﺳﺖ ‪ .‬ﻳﻌﻨﻲ ﺍﺑﺘﺪﺍ ‪ T2‬ﻛﺎﻣﻼ‬
‫ﺍﺟﺮﺍ ﻭ ﺑﺎ ﻣﻮﻓﻘﻴﺖ ﻳﻪ ﺍﺗﻤﺎﻡ ﺭﺳﻴﺪﻩ ﺳﭙﺲ ‪ T1‬ﺍﺟﺮﺍ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪:Schedule 3‬‬

‫ﺍﻳﻦ ﺯﻣﺎﻧﺒﻨﺪ ﺳﺮﻳﺎﻝ ﻧﻴﺴﺖ ﻭﻟﻲ ﻧﺘﻴﺠﻪ ﺁﻥ ﻫﻴﭻ ﺗﻔﺎﻭﺗﻲ ﺑﺎ ﻧﺘﻴﺠﻪ ‪ Schedule1‬ﻧﺪﺍﺭﺩ‪ .‬ﺍﮔﺮ ‪ A+B‬ﺩﺭ‬
‫ﺍﺑﺘﺪﺍ ﻭ ﺍﻧﺘﻬﺎ ﺗﻔﺎﻭﺕ ﻛﻨﺪ ﺳﺎﺯﮔﺎﺭﻱ ﺍﺯ ﺑﻴﻦ ﻣﻲ ﺭﻭﺩ )ﺩﺭ ﻭﺍﻗﻊ ﺑﺎ ﺍﻳﻨﻜﻪ ﺯﻣﺎﻧﺒﻨﺪ ﺳﺮﻳﺎﻝ ﻧﻴﺴﺖ ﺍﻣﺎ ﻣﻌﺎﺩﻝ‬
‫ﻳﻚ ﺯﻣﺎﻧﺒﻨﺪ ﺳﺮﻳﺎﻝ ﻋﻤﻞ ﻣﻲ ﻛﻨﺪ(‪.‬‬

‫‪27‬‬
 ‫‪:Schedule 4‬‬
‫ﺩﺭ ﺍﻳﻦ ﺯﻣﺎﻧﺒﻨﺪ ﻣﻘﺪﺍﺭ ‪ A+B‬ﺭﺍ ﺣﻔﻆ ﻧﻤﻲ ﺷﻮﺩ ﻳﻌﻨﻲ ﺍﺑﺘﺪﺍ ﻭ ﺍﻧﺘﻬﺎﻱ ﺯﻣﺎﻧﺒﻨﺪﻫﺎﻱ ‪T1‬ﻭ ‪ ،T2‬ﻣﻘﺪﺍﺭ‬
‫‪ A‬ﺩﺭ‬ ‫‪ A+B‬ﻳﻜﺴﺎﻥ ﻧﻴﺴﺖ ‪ .‬ﭼﻮﻥ ﺩﺳﺘﻮﺭ )‪ write(A‬ﺩﺭ ﺑﺎﻻ ﺯﻣﺎﻧﺒﻨﺪ ‪ T1‬ﻧﻴﺎﻣﺪﻩ ﻭ ﻣﻘﺪﺍﺭ‬
‫‪ Shadow‬ﺛﺒﺖ ﻧﻤﻲ ﺷﻮﺩ ﻳﻌﻨﻲ ﻣﻘﺪﺍﺭ ‪ A‬ﻫﻤﺎﻥ ﻣﻘﺪﺍﺭ ﺍﻭﻟﻴﻪ ﺩﺭ ﻧﻈﺮ ﮔﺮﻓﺘﻪ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫ﻧﻜﺘﻪ‪:‬‬

‫‪28‬‬
 ‫ﺩﻭ ﺭﻭﺵ ﺍﺻﻠﻲ ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮﻱ ﻋﺒﺎﺭﺗﻨﺪ ﺍﺯ‪:‬‬
‫‪ - 1‬ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮ ﺩﺭ ﺑﺮﺧﻮﺭﺩ ﻳﺎ ‪(Conflict Serializability) CSR‬‬
‫‪ - 2‬ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮ ﺩﺭ ﺩﻳﺪ ﻳﺎ ‪(View Serializability) VSR‬‬
‫ﺑﺮﺧﻮﺭﺩ )‪(Conflict‬‬
‫ﭼﻨﺎﻧﭽﻪ ‪ qj ,pi‬ﺑﻪ ﺗﺮﺗﻴﺐ ﺩﺳﺘﻮﺭﺍﺕ ﺗﺮﺍﻛﻨﺸﻬﺎﻱ ‪ tj ,ti‬ﺑﺎﺷﻨﺪ‪ ،‬ﮔﻮﻳﻴﻢ ‪ qj ,pi‬ﺑﺎ ﻫﻢ ﺑﺮﺧﻮﺭﺩ ﺩﺍﺭﻧﺪ‬
‫ﺍﮔﺮ ﻭ ﺗﻨﻬﺎ ﺍﮔﺮ‪:‬‬
‫‪ .1‬ﺍﻳﻦ ﺩﻭ ﻋﻤﻠﮕﺮ ﻣﺮﺑﻮﻁ ﺑﻪ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﻣﺘﻤﺎﻳﺰ ﺑﺎﺷﻨﺪ ) ‪ ti‬ﻧﺎﻣﺴﺎﻭﻱ ‪( tj‬‬
‫‪ .2‬ﻫﺮ ﺩﻭ ﻋﻤﻠﮕﺮ ﺑﻪ ﻳﻚ ﺩﺍﺩﻩ ﺩﺳﺘﺮﺳﻲ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻨﺪ‪.‬‬
‫‪ .3‬ﺣﺪﺍﻗﻞ ﻳﻜﻲ ﺍﺯ ﺍﻳﻦ ﺩﻭ ﻋﻤﻠﮕﺮ‪ ،‬ﻋﻤﻠﮕﺮ ﻧﻮﺷﺘﻦ ))(‪(write‬ﺑﺎﺷﺪ‪.‬‬

‫ﺯﻣﺎﻧﺒﻨﺪ ﻣﻌﺎﺩﻝ ﺩﺭ ﺑﺮﺧﻮﺭﺩ )‪(conflict equivalent‬‬

‫‪29‬‬
‫ﺯﻣﺎﻧﺒﻨﺪﻫﺎﻱ '‪ s‬ﻭ‪ s‬ﻣﻌﺎﺩﻝ ﺩﺭ ﺑﺮﺧﻮﺭﺩ ﻫﺴﺘﻨﺪ‪ ،‬ﺍﮔﺮ ﻫﺮ ﺩﻭ ﺭﻭﻱ ﻳﻚ ﻣﺠﻤﻮﻋﻪ ﺍﺯ ﺩﺳﺘﻮﺭﺍﺕ ﻭ ﺗﺮﺍﻛﻨﺶ‬
‫ﻫﺎ ﻛﺎﺭ ﻛﻨﻨﺪ ﻭ ﺑﺎ ﺟﺎﺑﺠﺎ ﻛﺮﺩﻥ ﺩﺳﺘﻮﺭﺍﺕ ﺑﺪﻭﻥ ﺑﺮﺧﻮﺭﺩ ﺩﺭ ﺯﻣﺎﻧﺒﻨﺪ ‪ ،s‬ﺑﺘﻮﺍﻧﻴﻢ ﺯﻣﺎﻧﺒﻨﺪ'‪ s‬ﺭﺍ ﺗﻮﻟﻴﺪ ﻛﻨﻴﻢ‬
‫‪.‬‬
‫ﻧﻜﺘﻪ ﻣﻬﻢ‪ :‬ﺑﻪ ﻫﻴﭻ ﻋﻨﻮﺍﻥ ﻧﻤﻲ ﺗﻮﺍﻧﻴﻢ ﺗﺮﺗﻴﺐ ﺩﺳﺘﻮﺭﺍﺕ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺭﺍ ﻋﻮﺽ ﻛﻨﻴﻢ ﭼﻮﻥ ﻣﻨﻄﻖ ﺁﻥ‬
‫ﺗﺮﺍﻛﻨﺶ ﻋﻮﺽ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﺗﻌﺮﻳﻒ‪ :‬ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮﻱ ﺩﺭ ﺑﺮﺧﻮﺭﺩ‬

‫ﺍﮔﺮ ﺩﺳﺘﻮﺭﺍﺕ ﺧﻮﺍﻧﺪﻥ ﻭ ﻧﻮﺷﺘﻦ ﻣﺮﺑﻮﻁ ﺑﻪ ﺗﺮﺍﻛﻨﺶ ﻩ ﺍﻱ ﻣﺨﺘﻠﻒ ﺑﻪ ﺻﻮﺭﺕ ﻫﻤﺮﻭﻧﺪ ﺍﺟﺮﺍ ﺷﻮﻧﺪ‪ ،‬ﺑﻌﻀﺎً‬
‫ﺑﺎ ﻫﻢ ﺑﺮﺧﻮﺭﺩ ﺩﺍﺭﻧﺪ ﻭ ﺑﻌﻀﺎً ﻧﺪﺍﺭﻧﺪ‪ .‬ﺑﺮﺧﻮﺭﺩ ﺩﺍﺷﺘﻦ ﻳﻌﻨﻲ ﺗﺎﺛﻴﺮ ﻣﺨﺮﺏ ﺭﻭﻱ ﻫﻢ ﮔﺬﺍﺷﺘﻦ ‪ .‬ﻣﺎ‬
‫ﺩﺳﺘﻮﺭﺍﺗﻲ ﺭﺍ ﻛﻪ ﺑﺎ ﻫﻢ ﺑﺮﺧﻮﺭﺩ ﻧﺪﺍﺭﻧﺪ ﻣﻲ ﺗﻮﺍﻧﻴﻢ ﻫﻤﺮﻭﻧﺪ ﺍﺟﺮﺍ ﻛﻨﻴﻢ ﻭ ﺁﻧﻬﺎﻳﻲ ﺭﺍ ﺑﺮﺧﻮﺭﺩ ﺩﺍﺭﻧﺪ‪ ،‬ﺑﺎﻳﺪ‬
‫ﺑﻪ ﺻﻮﺭﺕ ﺳﺮﻳﺎﻟﻲ ﺍﺟﺮﺍ ﻛﻨﻴﻢ‪.‬‬

‫ﺗﺸﺨﻴﺺ ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮﻱ ﺩﺭ ﺑﺮﺧﻮﺭﺩ )‪:(Conflict Serializability‬‬

‫ﺩﺭ ﻣﻔﻬﻮﻡ ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮﻱ ﺩﺭ ﺑﺮﺧﻮﺭﺩ ﺍﮔﺮ ﺯﻣﺎﻧﺒﻨﺪﻱ ﺑﺎ ‪ n‬ﺗﺮﺍﻛﻨﺶ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻴﻢ‪ ،‬ﺩﺳﺘﻮﺭﻫﺎﻱ ﺑﺪﻭﻥ‬
‫ﺑﺮﺧﻮﺭﺩ ﺭﺍ ﺁﻥ ﻗﺪﺭ ﺟﺎﺑﺠﺎ ﻣﻴﻜﻨﻴﻢ ﺗﺎ ﺑﺒﻴﻨﻴﻢ ﺑﻪ ﻳﻚ ﺗﺮﻛﻴﺐ ﺳﺮﻳﺎﻟﻲ ﻣﻲ ﺭﺳﻴﻢ ﻳﺎ ﻧﻪ ‪ .‬ﺍﻳﻦ ﻛﺎﺭ ﺑﺴﻴﺎﺭ‬
‫ﺯﻣﺎﻧﺒﺮ ﺍﺳﺖ ﺯﻳﺮﺍ ﺯﻣﺎﻧﺒﻨﺪﻫﺎ ﻧﻮﻋﺎ ﺍﺯ ﺗﻌﺪﺍﺩ ﺯﻳﺎﺩﻱ ﺗﺮﺍﻛﻨﺶ ﺗﺸﻜﻴﻞ ﺷﺪﻩ ﺍﻧﺪ ﻛﻪ ﺩﺳﺘﻮﺭﺍﺕ ﮔﻮﻧﺎﮔﻮﻥ‬
‫ﺧﻮﺩ ﺭﺍ ﻓﺮﺍﺧﻮﺍﻧﻲ ﻣﻲ ﻛﻨﻨﺪ ‪ .‬ﺧﻮﺷﺒﺨﺘﺎﻧﻪ ﺭﺍﻩ ﺣﻞ ﺑﺴﻴﺎﺭ ﺳﺎﺩﻩ ﺍﻱ ﺑﺮﺍﻱ ﺍﻳﻦ ﻣﺴﺌﻠﻪ ﭘﻴﺪﺍ ﺷﺪﻩ ﻛﻪ ﺁﻥ ﺭﺍ‬
‫ﺑﺮﺭﺳﻲ ﻣﻲ ﻛﻨﻴﻢ‪.‬‬

‫ﮔﺮﺍﻑ ﺍﻭﻟﻮﻳﺖ ﺩﺍﺭ )‪:(Serializability Graph‬‬

‫ﺭﺍﺱ ﻫﺎﻱ ﺍﻳﻦ ﮔﺮﺍﻑ ﻫﻤﺎﻥ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﻓﻌﺎﻝ ﺩﺭ ﺳﻴﺴﺘﻢ ﻫﺴﺘﻨﺪ ﻭ ﻳﺎﻟﻬﺎﻱ ﺁﻥ ﺟﻬﺖ ﺩﺍﺭ ﻣﻲ ﺑﺎﺷﻨﺪ ‪.‬‬
‫ﭘﺲ ﺍﻳﻦ ﮔﺮﺍﻑ‪ ،‬ﻳﻚ ﮔﺮﺍﻑ ﺟﻬﺖ ﺩﺍﺭ ﺍﺳﺖ ﻛﻪ ﺭﺍﺱ ﻫﺎﻱ ﺁﻥ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺗﺮﺍﻛﻨﺶ ﻫﺎﺳﺖ ﻭ ﻛﻤﺎﻥ‬
‫ﻫﺎﻱ ﺁﻥ ﺑﺮﺧﻮﺭﺩﻫﺎﻱ ﺑﻴﻦ ﺗﺮﺍﻛﻨﺸﻬﺎﺳﺖ ﻭ ﻟﻴﺒﻞ )‪ (Label‬ﻛﻤﺎﻥ ﻫﺎﻱ ﺟﻬﺖ ﺩﺍﺭ‪ ،‬ﺩﻳﺘﺎﺁﻳﺘﻢ ﺍﻱ ﺍﺳﺖ‬
‫ﻛﻪ ﺭﻭﻱ ﺁﻥ ﺑﺮﺧﻮﺭﺩ ﺍﻳﺠﺎﺩ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬

‫‪30‬‬
‫ﮔﺮﺍﻑ ﻫﺎ ﺑﻪ ﻃﻮﺭ ﻣﺮﺗﺐ ﺗﻐﻴﻴﺮ ﻣﻲ ﻛﻨﻨﺪ؛ ﻳﻌﻨﻲ ﺑﻪ ﻣﺤﺾ ﺍﻳﻨﻜﻪ ﻳﻚ ﺩﺳﺘﻮﺭ ﺧﻮﺍﻧﺪﻥ ﻭ ﻳﺎ ﻧﻮﺷﺘﻦ‬
‫ﻭﺍﺭﺩﺳﻴﺴﺘﻢ ﻣﻲ ﺷﻮﺩ‪ ،‬ﻣﻤﻜﻦ ﺍﺳﺖ ﻳﻚ ﻳﺎﻝ ﺭﺍ ﺑﻪ ﮔﺮﺍﻑ ﺍﺿﺎﻓﻪ ﻛﻨﺪ ﻭ ﺯﻣﺎﻧﻲ ﻛﻪ ﺗﺮﺍﻛﻨﺸﻲ ﺳﺎﻗﻂ ﻭ ﻳﺎ‬
‫ﺗﺜﺒﻴﺖ ﺷﺪﻩ ﻭ ﺍﺯ ﺳﻴﺴﺘﻢ ﺧﺎﺭﺝ ﻣﻲ ﺷﻮﺩ‪ ،‬ﻳﺎﻝ ﻫﺎﻳﻲ ﻭ ﻧﻴﺰ ﮔﺮﻩ ﺍﻱ ﺍﺯ ﮔﺮﺍﻑ ﺣﺬﻑ ﺧﻮﺍﻫﺪ ﺷﺪ‪.‬‬

‫‪x‬‬

‫‪y‬‬
‫ﻫﺮﮔﺎﻩ ﺩﺭ ﮔﺮﺍﻑ ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮ‪ ،‬ﺣﻠﻘﻪ )‪ (Cycle‬ﻭﺟﻮﺩ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ‪ ،‬ﺁﻧﮕﺎﻩ ﺯﻣﺎﻧﺒﻨﺪﻱ ﺁﻥ ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮ‬
‫ﻧﻴﺴﺖ ﻭ ﺑﺎﻟﻌﻜﺲ‪.‬‬

‫ﻣﺜﺎﻝ‪ :‬ﻛﻤﺎﻧﻲ ﺍﺯ ‪ T1‬ﺑﻪ ‪ T2‬ﺭﺳﻢ ﺷﺪﻩ ﺍﺳﺖ ﻳﻌﻨﻲ ﺍﻳﻦ ﺩﻭ ﺗﺮﺍﻛﻨﺶ ﺑﺎ ﻫﻢ ﺑﺮﺧﻮﺭﺩ ﺩﺍﺭﻧﺪ ﻣﺜﻼً ‪ T1‬ﻃﺒﻖ‬
‫ﺍﻳﻦ ﺷﻜﻞ ﺭﻭﻱ ﺩﻳﺘﺎﺁﻳﺘﻢ ‪ Y‬ﺑﺎ ‪ ، T2‬ﺑﺮﺧﻮﺭﺩ ﺩﺍﺭﻧﺪ‪.‬‬

‫‪31‬‬
‫ﺑﺮﺍﺳﺎﺱ ﮔﺮﺍﻑ ﻓﻮﻕ‪ ،‬ﺑﻪ ﺩﻭ ﺷﻜﻞ ﻣﻲ ﺗﻮﺍﻥ ﺗﺮﺗﻴﺐ ﺍﺟﺮﺍﻱ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺭﺍ ﺑﻪ ﺻﻮﺭﺕ ﺳﺮﻳﺎﻝ )ﻣﺘﻮﺍﻟﻲ(‬
‫ﻧﻮﺷﺖ‪:‬‬
‫)‪1‬‬ ‫‪T1 , T 2 , T 3 , T 4‬‬
‫)‪2‬‬ ‫‪T1 , T3 , T2 , T4‬‬

‫ﺯﻣﺎﻧﺒﻨﺪ ﻣﻌﺎﺩﻝ ﺩﺭ ﺩﻱﺩ )‪(view equivalent‬‬

‫ﺯﻣﺎﻧﺒﻨﺪﻫﺎﻱ ‪ s‬ﻭ`‪ s‬ﻣﻌﺎﺩﻝ ﺩﺭ ﺩﻳﺪ ﻫﺴﺘﻨﺪ ﺍﮔﺮ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﻭ ﻣﺠﻤﻮﻋﻪ ﻋﻤﻠﮕﺮﻫﺎﻱ ‪ s‬ﻭ `‪ s‬ﻳﻜﺴﺎﻥ ﺑﺎﺷﺪ‬
‫ﻭ ﺳﻪ ﺷﺮﻁ ﺯﻳﺮ ﺑﺮﻗﺮﺍﺭ ﺑﺎﺷﺪ‪:‬‬

‫‪ (1‬ﺑﺮﺍﻱ ﻫﺮ ﺩﺍﺩﻩ ‪ ،Q‬ﺍﮔﺮ ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﻣﻘﺪﺍﺭ ﺍﻭﻟﻴﻪ ‪ Q‬ﺭﺍ ﺩﺭ ‪ s‬ﻣﻲ ﺧﻮﺍﻧﺪ‪ ،‬ﺁﻧﮕﺎﻩ ‪ Tj‬ﻣﻘﺪﺍﺭ ﺍﻭﻟﻴﻪ ‪Q‬‬
‫ﺭﺍ ﺩﺭ `‪ s‬ﻧﻴﺰ ﺑﺨﻮﺍﻧﺪ‪.‬‬
‫‪ (2‬ﺑﺮﺍﻱ ﻫﺮ ﺩﺍﺩﻩ ‪ Q‬ﺍﮔﺮ ‪ Ti‬ﺩﺭ ‪ s‬ﺩﺍﺩﻩ ‪ Q‬ﺭﺍ ﺍﺯ ‪ Tj‬ﻣﻲ ﺧﻮﺍﻧﺪ‪ ،‬ﺩﺭ`‪ s‬ﻧﻴﺰ ﺩﺍﺩﻩ ‪ Q‬ﺭﺍ ﺍﺯ ‪ Tj‬ﺑﺨﻮﺍﻧﺪ‪.‬‬
‫‪ (3‬ﺑﺮﺍﻱ ﻫﺮ ﺩﺍﺩﻩ ‪ ،Q‬ﺁﺧﺮﻳﻦ ﺗﺮ ﺍﻛﻨﺸﻲ ﺍﺯ ﺯﻣﺎﻧﺒﻨﺪ ‪ s‬ﻛﻪ ﺭﻭﻱ ‪ Q‬ﻣﻲ ﻧﻮﻳﺴﺪ‪ ،‬ﻫﻤﺎﻥ ﺗﺮﺍﻛﻨﺸﻲ‬
‫ﺑﺎﺷﺪ ﻛﻪ ﺩﺭ ﺯﻣﺎﻧﺒﻨﺪ `‪ s‬ﺁﺧﺮﻳﻦ ﺑﺎﺭ ﺭﻭﻱ ‪ Q‬ﻣﻲ ﻧﻮﻳﺴﺪ‪.‬‬

‫ﻭﻗﺘﻲ ﺍﺯ ﺩﻳﺪﮔﺎﻩ ﻳﻜﺴﺎﻧﻲ ﺑﻪ ﺍﻳﻦ ﺩﻭ ﺯﻣﺎﻧﺒﻨﺪ ﻧﮕﺎﻩ ﻛﻨﻴﻢ ﻭ ﺍﮔﺮ ﻳﻜﺴﺎﻥ ﺑﻮﺩﻧﺪ ﻣﻲ ﮔﻮﻳﻴﻢ ﻣﻌﺎﺩﻝ ﺩﺭ ﺩﻳﺪ‬
‫ﻫﺴﺘﻨﺪ‪.‬‬

‫ﺗﻌﺮﻳﻒ‪ :‬ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮﻱ ﺩﺭ ﺩﻳﺪ‬

‫ﺯﻣﺎﻧﺒﻨﺪ ‪ S‬ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮ ﺩﺭ ﺩﻳﺪ ﺍﺳﺖ ﺍﮔﺮ ﻣﻌﺎﺩﻝ ﺩﺭ ﺩﻳﺪ ﺑﺎ ﻳﻚ ﺯﻣﺎﻧﺒﻨﺪ ﺳﺮﻳﺎﻟﻲ ﺑﺎﺷﺪ ‪.‬‬
‫ﻣﺜﺎﻝ‪ :‬ﺁﻳﺎ ﺯﻣﺎﻧﺒﻨﺪ ﺯﻳﺮ ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮ ﺩﺭ ﺑﺮﺧﻮﺭﺩ ﻣﻲ ﺑﺎﺷﺪ؟ ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮ ﺩﺭ ﺩﻳﺪ ﭼﻄﻮﺭ؟‬

‫ﺣﻞ‪ :‬ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮ ﺩﺭ ﺑﺮﺧﻮﺭﺩ ﻧﻴﺴﺖ ﻭﻟﻲ ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮ ﺩﺭ ﺩﻳﺪ ﻣﻲ ﺑﺎﺷﺪ ‪ .‬ﺑﺮﺍﻱ ﺗﺸﺨﻴﺺ ﺳﺮﻳﺎﻟﻲ‬
‫ﭘﺬﻳﺮ ﺑﻮﺩﻥ ﺩﺭ ﺩﻳﺪ‪ ،‬ﻓﻘﻂ ﻳﻚ ﺩﺍﺩﻩ ﺑﻪ ﻧﺎﻡ ‪ Q‬ﺩﺍﺭﻳﻢ ﻛﻪ‪:‬‬

‫‪32‬‬
 ‫‪ .1‬ﺩﺭ ﻫﺮ ﺩﻭ ﺯﻣﺎﻧﺒﻨﺪ‪ T3 ،‬ﻣﻘﺪﺍﺭ ﺍﻭﻟﻴﻪ ‪ Q‬ﺭﺍ ﻣﻲ ﺧﻮﺍﻧﺪ‪.‬‬
‫‪ .2‬ﻫﻴﭻ ﺗﺮﺍﻛﻨﺸﻲ ﻣﻘﺪﺍﺭﻱ ﺑﺮﺍﻱ ‪ Q‬ﺍﺯ ﺗﺮﺍﻛﻨﺸﻲ ﺩﻳﮕﺮ ﻧﻤﻲ ﺧﻮﺍﻧﺪ‪.‬‬
‫‪ .3‬ﺩﺭ ﻫﺮ ﺩﻭ ﺯﻣﺎﻧﺒﻨﺪ‪ ،‬ﺗﺮﺍﻛﻨﺶ ‪ T5‬ﺁﺧﺮﻳﻦ ﻋﻤﻞ ﻧﻮﺷﺘﻦ ﺭﻭﻱ ‪ Q‬ﺭﺍ ﺍﻧﺠﺎﻡ ﻣﻲ ﺩﻫﺪ‪.‬‬
‫** ﭘﺲ ﺍﻳﻦ ﺯﻣﺎﻧﺒﻨﺪ ‪ CSR‬ﻧﻴﺴﺖ ﺍﻣﺎ ‪ VSR‬ﻫﺴﺖ **‬

‫ﻣﺤﺪﻭﺩﻩ ﻋﻤﻠﻜﺮﺩ ﺩﻭ ﺭﻭﺵ ﺍﺻﻠﻲ ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮﻱ ‪:‬‬

‫ﻣﻔﻬﻮﻡ ﺗﺮﺍﻛﻨﺶ ﺩﺭ ‪: SQL‬‬

‫ﺩﺭ ﺍﻳﻨﺠﺎ ﺗﺮﺍﻛﻨﺶ ﺑﺎ "‪ "begin‬ﺷﺮﻭﻉ ﻭ ﺑﺎ "‪ "end‬ﺷﺮﻭﻉ ﻣﻴﺸﻮﺩ‪ ،‬ﻋﻤﻞ "‪ "end‬ﻣﻴﺘﻮﺍﻧﺪ ﺑﻪ‬ ‫‪‬‬
‫ﺩﻭ ﺷﻜﻞ ﺯﻳﺮ ﺑﺎﺷﺪ‪:‬‬
‫‪Commit Work‬‬ ‫•‬
‫‪Rollback Work‬‬ ‫•‬
‫‪Undo .1‬‬
‫‪Redo .2‬‬
‫ﻣﺮﺍﺣﻞ ﺳﺎﺯﮔﺎﺭﻱ ﺩﺭ ‪:SQL‬‬ ‫‪‬‬
‫‪Serializable‬‬ ‫•‬
‫‪Repeatable read‬‬ ‫•‬
‫‪Read committed‬‬ ‫•‬
‫‪Read uncommitted‬‬ ‫•‬

‫ﺳﻄﻮﺡ ﺳﺎﺯﮔﺎﺭﻱ ﺩﺭ ‪: SQL-22‬‬

‫ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ‪ SQL‬ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﺩﻳﺪ ﺳﺎﺯﮔﺎﺭ ﺑﻮﺩﻥ ﭼﻬﺎﺭ ﺣﺎﻟﺖ ﺑﺮﺍﻱ ﻫﺮ ﺗﺮﺍﻛﻨﺶ ﺩﺭ ﻧﻈﺮ ﻣﻲ ‌ﮔﻴﺮﺩ‪:‬‬
‫‪33‬‬
 ‫‪ .1‬ﺳﺮﻳﺎﻟﻲ ﺑﻮﺩﻥ)‪ :(Serializabale‬ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺑﻪ ﺻﻮﺭﺕ ﺳﺮﻳﺎﻟﻲ ﺗﻌﺮﻳﻒ ﻣﻲ ﺷﻮﺩ‪.‬‬
‫‪ .2‬ﺧﻮﺍﻧﺪﻥ ﻫﺎﻱ ﻣﻜﺮّﺭ )‪ :(Repeatable Read‬ﻓﻘﻂ ﺭﻛﻮﺭﺩﻫﺎ ﻭ ﺍﻃﻼﻋﺎﺕﻱ ﻛﻪ ﺑﺎ ﻣﻮﻓﻘﻴﺖ ﺑﻪ‬
‫ﺍﺗﻤﺎﻡ ﺭﺳﻴﺪﻩ ﺍﺳﺖ ﺧﻮﺍﻧﺪﻩ ﻣﻲ ﺷﻮﺩ ﻭ ﺧﻮﺍﻧﺪﻥ ﻫﺎﻱ ﻣﻜﺮﺭ ﺑﻌﺪ ﺍﺯ ﺁﻥ ﻫﻤﺎﻥ ﻣﻘﺪﺍﺭﻱ ﻛﻪ ﺍﻭﻝ‬
‫ﺧﻮﺍﻧﺪﻩ ﺷﺪﻩ ﺍﺳﺖ ﺭﺍ ﺑﺮ ﻣﻲ ﮔﺮﺩﺍﻧﺪ‪.‬‬
‫‪ .3‬ﺧﻮﺍﻧﺪﻥ ﺭﻛﻮﺭﺩ ﭘﺬﻳﺮﻓﺘﻪ ﺷﺪﻩ )‪ :(Read Committed‬ﻳﻚ ﺭﻛﻮﺭﺩ ﻱ ﻛﻪ ﺑﺎ ﻣﻮﻓﻘﻴﺖ ﺗﻤﺎﻡ‬
‫ﺷﺪﻩ ﺧﻮﺍﻧﺪﻩ ﻣﻲ ﺷﻮﺩ ﻭﻟﻲ ﻣﻘﺪﺍﺭ ﺑﻌﺪﻱ ﺧﻮﺍﻧﺪﻩ ﺷﺪﻩ ﺑﺮ ﺭﻭﻱ ﻫﻤﺎﻥ ﺭﻛﻮﺭﺩ ﻣﻤﻜﻦ ﺍﺳﺖ‬
‫ﻣﻘﺎﺩﻳﺮ ﻣﺨﺘﻠﻔﻲ ﺭﺍ ﺑﺮﮔﺮﺩﺍﻧﺪ‪.‬‬
‫‪ .4‬ﺧﻮﺍﻧﺪﻥ ﺭﻛﻮﺭﺩﻫﺎﻱ ﭘﺬﻳﺮﻓﺘﻪ ﻧﺸﺪﻩ )‪ :(Read Uncommitted‬ﺍﻳﻨﺠﺎ ﻣﻤﻜﻦ ﺍﺳﺖ ﺧﺘﻲ‬
‫ﺭﻛﻮﺭﺩﻫﺎﻳﻲ ﻛﻪ ﺑﺎ ﻣﻮﻓﻘﻴﺖ ﺗﻤﺎﻡ ﻧﺸﺪﻩ ﻧﻴﺰ ﺧﻮﺍﻧﺪﻩ ﺷﻮﺩ ‪.‬‬

‫ﺳﻄﻮﺡ ﭘﺎﻳﻴﻦ ﺳﺎﺯﮔﺎﺭﻱ‪:‬‬

‫ﺩﺭ ﺣﺎﻻﺕ ﻣﺨﺘﻠﻒ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻛﻪ ﺳﺎﺯﮔﺎﺭﻱ ﺩﺭ ﺳﻄﺢ ﭘﺎﻳﻴﻦ ﺩﺍﺭﻧﺪ ﺍﻣﻜﺎﻥ ﺩﺍﺭﺩ ﻧﺎﺳﺎﺯﮔﺎﺭﻱ ﺑﻮﺟﻮﺩ ﺁﻳﺪ ‪.‬‬

‫‪ .‬ﺳﻄﻮﺡ ﺿﻌﻴﻒ‬ ‫ﺑﻌﻀﻲ ﻛﺎﺭﺑﺮﺩﻫﺎ ﺗﻤﺎﻳﻞ ﺑﺮ ﺍﻳﻦ ﺩﺍﺭﻧﺪ ﻛﻪ ﺩﺭ ﺳﻄﺢ ﺳﺎﺯﮔﺎﺭﻱ ﺿﻌﻴﻔﻲ ﻗﺮﺍﺭ ﺑﮕﻴﺮﻧﺪ‬
‫ﻧﺎﺳﺎﺯﮔﺎﺭﻱ ﺍﺯ ﺯﻣﺎﻥ ﺑﻨﺪ ﻫﺎﻳﻲ ﻛﻪ ﺳﺮﻳﺎﻝ ﻧﻴﺴﺘﻨﺪ ﺑﻪ ﻭﺟﻮﺩ ﻣﻲ ﺁﻳﺪ‪.‬‬

‫ﻣﺎﻧﻨﺪ ﺗﺮﺍﻛﻨﺸﻲ ﻛﻪ ﻓﻘﻂ ﺩﺳﺘﻮﺭﺍﺕ ﺧﻮﺍﻧﺪﻥ ﺩﺍﺭﺩ‪ ،‬ﻣﺎﻧﻨﺪ ﺗﺮﺍﻛﻨﺸﻲ ﻛﻪ ﺩﺭ ﺁﻥ ﻣﻲ ﺧﻮﺍﻩ ﻳﻢ ﺑﺎﻻﻧﺲ ﻛﻠﻲ‬
‫ﻱ‪.‬‬
‫ﺍﺯ ﻫﻤﻪ ﻱ ﺣﺴﺎﺏ ﻫﺎﻳﻲ ﻛﻪ ﺩﺭ ﺑﺎﻧﻚ ﻭﺟﻮﺩ ﺩﺍﺭﺩ ﺭﺍ ﺑﺪﺳﺖ ﺁ ﺩ‬

‫" ﺍﻳﻦ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺍﺣﺘﻴﺎﺝ ﻧﺪﺍﺭﻧﺪ ﻛﻪ ﺩﺭ ﻣﻘﺎﺑﻞ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﺩﻳﮕﺮ ﺳﺮﻳﺎﻝ ﺑﺎﺷﻨﺪ ‪ .‬ﺍﮔﺮ ﺳﺮﻳﺎﻝﻱ ﻧﺒﺎﺷﺪ‬
‫ﺳﺎﺯﮔﺎﺭﻱ ﺿﻌﻴﻒ ﻣﻲ‌ﺷﻮﺩ‪" .‬‬

‫ﺟﺎﻣﻌﻴﺖ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺑﺎﻳﺪ ﺩﻭ ﺷﺮﻁ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ‪:‬‬

‫‪ .1‬ﺍﺯ ﻧﻈﺮ ﻛﻨﺘﺮﻝ ﻫﻤﺮﻭﻧﺪﻱ ﻣﺸﻜﻠﻲ ﻭﺟﻮﺩ ﻧﺪﺍﺷﺘﻪ ﺑﺎﺷﺪ ‪ .‬ﺍﺯ ﻃﺮﻓﻲ ﻧﻴﺰ ﻣﻲ ﺩﺍﻧﻴﻢ ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮﻱ‬
‫ﻳﻜﻲ ﺍﺯ ﺭﻭﺵ ﻫﺎﻱ ﻛﻨﺘﺮﻝ ﻫﻤﺮﻭﻧﺪﻱ ﺍﺳﺖ ‪.‬‬
‫‪ .2‬ﺗﺮﻣﻴﻢ ﭘﺬﻳﺮ ﺑﺎﺷﺪ‪.‬‬

‫‪34‬‬
 ‫ﻧﻜﺘﻪ‪ :‬ﭘﻴﺎﺩﻩ ﺳﺎﺯﻱ ﺧﺎﺻﻴ‪‬ﺖ ﻫﺎﻱ ﺍﺗﻮﻣﻴﻚ ﺑﻮﺩﻥ ﻭ ﭘﺎﻳﺪﺍﺭﻱ ﺗﻮﺳﻂ ﻣﻮﻟﻔﻪ ﺍﻱ ﺩﺭ ﻳﻚ ﺳﻴﺴﺘﻢ ﺑﺎﻧﻚ‬
‫ﺍﻃﻼﻋﺎﺗﻲ ﺻﻮﺭﺕ ﻣﻲ ﮔﻴﺮﺩ ﻛﻪ ﻣﺪﻳﺮﻳﺖ ﺗﺮﻣﻴﻢ )‪ (Recovery management‬ﻧﺎﻡ ﺩﺍﺭﺩ‪.‬‬

‫ﺗﺮﻣﻴﻢ ﭘﺬﻳﺮﻱ‪:‬‬
‫ﺯﻣﺎﻧﺒﻨﺪﻱ ﺭﺍ ﺗﺮﻣﻴﻢ ﭘﺬﻳﺮ ﮔﻮﻳﻨﺪ ﻛﻪ ﺑﺮﺍﻱ ﺗﻤﺎﻡ ‪ Tj‬ﻫﺎ ﻛﻪ ﺍﺯ ‪ Ti‬ﻣﻲ ﺧﻮﺍﻧﻨﺪ‪ ،‬ﺗﺜﺒﻴﺖ ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﻗﺒﻞ ﺍﺯ‬
‫ﺗﺜﺒﻴﺖ ﺗﺮﺍﻛﻨﺶ ‪ Tj‬ﺻﻮﺭﺕ ﮔﻴﺮﺩ‪.‬‬

‫‪Ti‬‬ ‫‪Tj‬‬
‫)‪Read (Q‬‬

‫)‪Write (Q‬‬
‫‪commit‬‬
‫‪Commit‬‬

‫ﺍﻳﻦ ﺗﺮﺍﻛﻨﺶ ﻗﺎﺑﻞ ﺗﺮﻣﻴﻢ ﺍﺳﺖ ﭼﻮﻥ ﻋﻤﻞ ﺧﻮﺍﻧﺪﻥ ﺩﻳﺘﺎﺁﻳﺘﻢ ‪ Q‬ﺩﺭ ﺗﺮﺍﻛﻨﺶ ‪ Tj‬ﻗﺒﻞ ﺍﺯ ﻧﻮﺷﺘﻪ ﺷﺪﻥ‬
‫‪ Q‬ﺗﻮﺳﻂ ‪ Tj‬ﺍﻧﺠﺎﻡ ﻣﻲ ‌ﺷﻮﺩ ﻭ ﺍﺗﻤﺎﻡ ﻣﻮﻓﻘﻴﺖ ﺁﻣﻴﺰ ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﻗﺒﻞ ﺍﺯ ﺍﺗﻤﺎﻡ ﻣﻮﻓﻘﻴﺖ ﺁﻣﻴﺰ ﺗﺮﺍﻛﻨﺶ‬
‫‪ Tj‬ﺍﻧﺠﺎﻡ ﻣﻲ‌ﺷﻮﺩ‪.‬‬

‫ﺑﺮﮔﺸﺖ ﻫﺎﻱ ﺁﺑﺸﺎﺭﻱ )ﭘﺸﺖ ﺳﺮ ﻫﻢ (ﻳﺎ ‪: Cascading Rollbacks‬‬

‫ﺧﻄﺎﻱ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﻣﻨﻔﺮﺩ ﺑﺎﻋﺚ ﺑﺮﮔﺸﺖ)‪ (Rollback‬ﻳﻚ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﻣﻲ ﺷﻮﺩ ‪.‬‬
‫ﺯﻣﺎﻧﺒﻨﺪ ﺯﻳﺮ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﻛﻪ ﻫﻨﻮﺯ ﺑﺎ ﻣﻮﻓﻘﻴﺖ ﺑﻪ ﺍﺗﻤﺎﻡ ﻧﺮﺳﻴﺪﻩ ﺍﺳﺖ ﺭﺍ ﺑﺮﺭﺳﻲ ﻣﻲ ﻛﻨﺪ ﺑﻨﺎﺑﺮﺍﻳﻦ‬
‫ﺯﻣﺎﻧﺒﻨﺪ ﻗﺎﺑﻞ ﺗﺮﻣﻴﻢ ﺍﺳﺖ‪.‬‬

‫ﺍﮔﺮ ﺩﺭ ‪ ،T10‬ﺧﻄﺎ ﺭﺥ ﺩﻫﺪ ‪ T11‬ﻭ ‪ T12‬ﺑﺎﻳﺪ ﺑﺮﮔﺸﺖ ﺩﺍﺩﻩ ﺷﻮﻧﺪ‪.‬‬

‫‪35‬‬
‫ﺍﻳﻦ ﺯﻣﺎﻧﺒﻨﺪ ﻛﺎﻣﻼً ﺳﺮﻳﺎﻝ ﺍﻧﺠﺎﻡ ﻣﻲ ‌ﺷﻮﺩ‪ .‬ﭼﻮﻥ ‪ T12, T11‬ﺩﻳﺘﺎﺁﻳﺘﻢ ‪ A‬ﺭﺍ ﺍﺯ ﺭﻭﻱ ﺩﻳﺴﻚ ﻣﻲ‌ﺧﻮﺍﻧﺪ ﻭ‬
‫ﭼﻮﻥ ‪ A‬ﺭﺍ ﺗﻐﻴﻴﺮ ﺩﺍﺩﻩ ﻭ ﻫﻨﻮﺯ ﺭﻭﻱ ﺩﻳﺴﻚ ﻧﻮﺷﺘﻪ ﻧﺸﺪﻩ ﺑﺎﻳﺪ ﺑﺮﮔﺸﺖ ﻛﻨﺪ ﻭ ﻛﻪ ﺑﺎﻋﺚ ﺍﻳﺠﺎﺩ ﺑﺮﺧﻮﺭﺩ‬
‫ﺧﻮﺍﻫﺪ ﺷﺪ‪.‬‬

‫ﻣﻲﮔﺬﺍﺭﺩ ﻭ ﺑﻌﺪ ﺩﺭ ‪T10‬‬

‫‪ T10‬ﺩﻳﺘﺎﺁﻳﺘﻢ ﻫﺎﻱ ‪A‬ﻭ‪ B‬ﺭﺍ ﭘﺲ ﺍﺯ ﺧﻮﺍﻧﺪﻥ ﺍﺯ ﺭﻭﻱ ﺩﻳﺴﻚ ﺭﻭﻱ ‪‌ shadow‬‬
‫ﺭﻭﻱ ﻫﻤﺎﻥ ‪ shadow‬ﻋﻤﻞ ﻧﻮﺷﺘﻦ ﺻﻮﺭﺕ ﻣﻲ ﮔﻴﺮﺩ ﻭ ﭘﺲ ﺍﺯ ﺍﺗﻤﺎﻡ ﻣﻮﻓﻘﻴﺖ ﺁﻣﻴﺰ ﺭﻭﻱ ﺩﻳﺴﻚ‬
‫ﻣﻲﻧﻮﻳﺴﺪ‪ .‬ﻋﻤﻞ ﺧﻮﺍﻧﺪﻥ ‪ A‬ﺩﺭ ‪ T11‬ﻭ ‪ T12‬ﺍﺯ ﺭﻭﻱ ﺩﻳﺴﻚ ﺍﻧﺠﺎﻡ ﻣﻲ ﺷﻮﺩ ﻧﻪ ﺍﺯ ﺭﻭﻱ ‪.shadow‬‬ ‫‌‬
‫ﭼﻮﻥ ﺭﺍﻳﺖ ‪ A‬ﺩﺭ ﺗﺮﺍﻛﻨﺶ ﻗﺒﻞ ﻫﻨﻮﺯ ﺭﻭﻱ ﺩﻳﺴﻚ ﻧﻮﺷﺘﻪ ﻧﺸﺪﻩ ﻣﺸﻜﻞ ﭘﻴﺶ ﻣﻲ ‌ﺁﻳﺪ‪ .‬ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ‬
‫‪ T11‬ﻭ ‪ T12‬ﺑﺎﻳﺪ ﺑﻪ ﺷﻜﻞ ﺁﺑﺸﺎﺭﻱ ﺑﺮﮔﺸﺖ ﻛﻨﺪ‪.‬‬

‫ﺯﻣﺎﻧﺒﻨﺪﻱ ﻏﻴﺮ ﺁﺑﺸﺎﺭﻱ ﻳﺎ ‪: Cascade less Schedules‬‬

‫ﺯﻣﺎﻥﺑﻨﺪﻱ ﻛﻪ ﻏﻴﺮﺁﺑﺸﺎﺭﻱ ﺍﺳﺖ ﻗﺎﺑﻞ ﺗﺮﻣﻴﻢ ﻣﻲ ﺑﺎﺷﺪ ‪ .‬ﻫﺪﻑ ﻧﻴﺰ ﻫﻤﻴﻦ ﺍﺳﺖ ﻛﻪ ﺗﺮﺍﻛﻨﺶ ﺭﺍ ﺩﻭﺑﺎﺭﻩ‬
‫‪ redo‬ﻧﻜﻨﻴﻢ ﻭ ‪ undo‬ﻭ ‪ redo‬ﻫﺎ ﺭﺍ ﻛﻤﺘﺮ ﻛﻨﻴﻢ ‪.‬ﺍﺯ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﺁﺑﺸﺎﺭﻱ ﺟﻠﻮﮔﻴﺮﻱ ﻛﻨﻴﻢ‬
‫)ﻭﺍﺑﺴﺘﮕﻲ ﺑﻴﻦ ﺗﺮﺍﻛﻨﺶ ‌ﻫﺎ ﺭﺍ ﻛﻢ ﻛﻨﻴﻢ (‪ .‬ﺗﻘﺮﻳﺒﺎً ﻏﻴﺮﺁﺑﺸﺎﺭﻱ ﺑﻮﺩﻥ ﻭ ﻗﺎﺑﻞ ﺗﺮﻣﻴﻢ ﺑﻮﺩﻥ ﺑﻪ ﻳﻚ ﻣﻌﻨﻲ‬
‫ﻣﻲﺑﺎﺵﻥﺩ ‪.‬‬
‫‌‬

‫ﻛﻨﺘﺮﻝ ﻫﻤﺮﻭﻧﺪﻱ‪:‬‬

‫ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺑﺎﻳﺪ ﻣﻜﺎﻧﻴﺰﻣﻲ ﺍﻳﺠﺎﺩ ﻛﻨﺪ ﻛﻪ ﻫﻤﻪ ﻱ ﺯﻣﺎﻧﺒﻨﺪﻫﺎ ‪:‬‬

‫ﺍﻟﻒ ‪ -‬ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮ ﺩﺭ ﺑﺮﺧﻮﺭﺩ ﻭ ﻳﺎ ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮ ﺩﺭ ﺩﻳﺪ ﺑﺎﺷﺪ‪.‬‬

‫ﺏ – ﻗﺎﺑﻞ ﺗﺮﻣﻴﻢ ﻭ ﻏﻴﺮﺁﺑﺸﺎﺭﻱ ﺑﺎﺷﺪ‪ .‬ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮ ﺩﺭ ﺑﺮﺧﻮﺭﺩ ﻳﻌﻨﻲ ﺯﻣﺎﻧﺒﻨﺪﻱ ‌ﻫﺎ ﺑﺎ ﻭﺟﻮﺩ ﺍﻳﻨﻜﻪ‬
‫ﻫﻤﺮﻭﻧﺪ ﺍﻧﺠﺎﻡ ﻣﻲ ‌ﺷﻮﻧﺪ ﺳﺮﻳﺎﻟﻲ ﭘﺬﻳﺮ ﺩﺭ ﺩﻳﺪ ﺑﺎﺷﻨﺪ‪ .‬ﺍﮔﺮ ﻣﻜﺎﻧﻴﺰﻡ ﻫﺎﻳﻲ ﺑﺎ ﺯﻣﺎﻧﺒﻨﺪﻱ ﺳﻴﺎﻝ ﻭﻗﺎﺑﻞ‬

‫‪36‬‬
 ‫ﺗﺮﻣﻴﻢ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻴﻢ ﺩﺭ ﻭﺍﻗﻊ ﻣﻲ ‌ﺗﻮﺍﻥ ﺑﺎ ﺍﻳﻦ ﻣﻜﺎﻧﻴﺰﻡ ﻫﺎ ﻛﻨﺘﺮﻝ ﻫﻤﺮﻭﻧﺪﻱ ﺍﻧﺠﺎﻡ ﺩﻫﻴﻢ ﻭ ﻫﻴﭻ‬
‫ﻧﺎﺳﺎﺯﮔﺎﺭﻱ ﺩﺭ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺍﻳﺠﺎﺩ ﻧﺸﻮﺩ‪ ،‬ﺍﻳﻦ ﺭﻭﺵ ﻳﻜﻲ ﺍﺯ ﺳﺎﺩﻩ ‌ﺗﺮﻳﻦ ﺭﻭﺵﻫﺎﻱ ﺳﻴﺎﻟﻲ ﺍﺳﺖ ﻛﻪ‬
‫ﻣﻲﺗﻮﺍﻧﻴﻢ ﺍﻧﺠﺎﻡ ﺩﻫﻴﻢ ﻛﻪ ﻫﻢ ﺳﺮﻳﺎﻝ ﻭ ﻫﻢ ﻗﺎﺑﻞ ﺗﺮﻣﻴﻢ ﺑﺎﺷﺪ ‪.‬‬
‫‌‬

‫ﻳﻚ ﺳﻴﺎﺳﺖ ﺍﻳﻦ ﺍﺳﺖ ﻛﻪ ﻓﻘﻂ ﺩﺭ ﻫﺮ ﻟﺤﻈﻪ ﺑﺘﻮﺍﻥ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺑﻪ ﺷﻜﻞ ﺱﺭﻳﺎﻝ ﺍﺟﺮﺍ ﺷﻮﺩ ﻛﻪ ﺍﻟﺒﺘﻪ‬
‫ﺩﺭ ﻭﺍﻗﻊ ﺩﺭ ﺍﻳﻦﺻﻮﺭﺕ ﻫﻤﺮﻭﻧﺪﻱ ﻧﺪﺍﺭﻳﻢ ‪.‬‬

‫ﻣﺼﺎﻟﺤﻪ ﺑﻴﻦ ﻫﻤﺮﻭﻧﺪﻱ ﺑﻴﺸﺘﺮ ﻭ ﺗﺮﻣﻴﻢ ﻛﻤﺘﺮ‬

‫ﺩﺭ ﺯﻣﺎﻧﺒﻨﺪﻱ ﻫﻤﺮﻭﻧﺪ ﺗﺮﺍﻛﻨﺶﻫﺎ ﺩﻭ ﻫﺪﻑ ﺑﺎﻳﺪ ﻫﻤﺰﻣﺎﻥ ﻣﻮﺭﺩ ﺗﻮﺟﻪ ﻗﺮﺍﺭ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ ‪:‬‬

‫)ﻣﺎﻛﺰﻳﻤﻢ ﻛﺮﺩﻥ‬ ‫‪- 1‬ﺣﺪﺍﻛﺜﺮ ﻛﺮﺩﻥ ﺗﻌﺪﺍﺩ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻳﻲ ﻛﻪ ﺑﻪ ﺻﻮﺭﺕ ﻫﻤﺰﻣﺎﻥ ﻓﻌﺎﻝ ﻫﺴﺘﻨﺪ‬
‫ﻫﻤﺮﻭﻧﺪ(‬
‫‪- 2‬ﺣﺪﺍﻗﻞ ﻛﺮﺩﻥ ﻧﻴﺎﺯ ﺑﻪ ﺗﺮﻣﻴﻢ‬

‫ﺍﻳﻦ ﺩﻭ ﻫﺪﻑ ﺑﺎ ﻫﻢ ﺩﺭ ﺗﻨﺎﻗﺾ ﻫﺴﺘﻨﺪ‪ ،‬ﺑﻪ ﻃﻮﺭﻱ ﻛﻪ ﻫﺮ ﭼﻪ ﺗﻌﺪﺍﺩ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﻏﻴﺮﻓﻌﺎﻝ ﺑﻴﺸﺘﺮ‬
‫ﺷﻮﺩ‪ ،‬ﻣﻴﺰﺍﻥ ﻧﻴﺎﺯ ﺑﻪ ﺗﺮﻣﻴﻢ ﻧﻴﺰ ﺍﻓﺰﺍﻳﺶ ﻣﻲ ﻳﺎﺑﺪ ﻭ ﺑﺮﻋﻜﺲ ‪ .‬ﺑﻨﺎﺑﺮﺍﻳﻦ ﺑﺮﺍﻱ ﺑﺮﻃﺮﻑ ﻛﺮﺩﻥ ﺍﻳﻦ ﺗﻨﺎﻗﻀﺎﺕ‬
‫ﻧﻴﺎﺯ ﺑﻪ ﻧﻮﻋﻲ ﻣﺼﺎﻟﺤﻪ ‪ 1‬ﺑﻴﻦ ﺍﻳﻦ ﺩﻭ ﻫﺪﻑ ﻭﺟﻮﺩ ﺩﺍﺭﺩ ﻛﻪ ﺍﻳﻦ ﻛﺎﺭ ﻣﺴﺘﻠﺰﻡ ﻫﻮﺷﻤﻨﺪﻱ ﺍﺳﺖ ‪.‬‬
‫‪F6‬‬

‫‪1‬‬
‫‪Tradeoff‬‬

‫‪37‬‬
 ‫ﻓﺼﻞ ‪:3‬‬
‫ﻣﺪﻳﺮﻳﺖ ﺗﺮﻣﻴﻢ‬

‫‪38‬‬
 ‫ﻣﻘﺪﻣﻪ‪:‬‬
‫ﺩﺭ ﻣﺪﻳﺮﻳﺖ ﺗﺮﺍﻛﻨﺶ ﻫﺎ‪،‬ﺍﻳﻦ ﺍﻣﻜﺎﻥ ﻭﺟﻮﺩ ﺩﺍﺭﺩ ﻛﻪ ﺩﺭ ﻫﺮ ﻟﺤﻈﻪ ﺍﺯ ﺍﺟﺮﺍﻱ ﺗﺮﺍﻛﻨﺶ‪ ،‬ﻳﻚ ﺧﺮﺍﺑﻲ ﻳﺎ‬
‫ﺑﻲﻓﺘﺪ ﻭ ﺗﺮﺍﻛﻨﺶ ﻧﺘﻮﺍﻧﺪ ﺍﺩﺍﻣﻪ ﻳﺎﺑﺪ‪ .‬ﺑﻪ ﻋﺒﺎﺭﺗﻲ ﺑﺮﺧﻲ ﺍﺯ ﺩﺳﺘﻮﺭﺍﺕ ﺁﻥ ﺍﺟﺮﺍ ﺷﺪﻩ‬‫ﺧﻄﺎ)‪ (failure‬ﺍﺗﻔﺎﻕ ﺍ‬
‫ﻭ ﺑﻘﻴﻪ ﺑﻪ ﺩﻟﻴﻞ ﻭﻗﻮﻉ ﺧﺮﺍﺑﻲ ﻣﺘﻮﻗﻒ ﺷﻮﻧﺪ‪ .‬ﻫﻤﭽﻨﻴﻦ ﺯﻣﺎﻧﻲ ﻛﻪ ﺗﺮﺍﻛﻨﺸﻲ ﺍﻧﺠﺎﻡ )‪ (commit‬ﻣﻲ‬
‫ﺑﻲﻓﺘﺪ ﻛﻪ ﺗﺄﺛﻴﺮ ﺁﻥ ﺭﺍ ﺍﺯ ﺑﻴﻦ ﺑﺒﺮﺩ‪.‬‬
‫ﺷﻮﺩ‪ ،‬ﺑﻌﺪﺍً ﺍﺗﻔﺎﻗﻲ ﺍ‬

‫ﺍﻧﻮﺍﻉ ﺧﺮﺍﺑﻲ )‪:(failure‬‬

‫ﺧﺮﺍﺑﻲ ﻫﺎﻳﻲ ﻛﻪ ﺩﺭ ﻳﻚ ﺳﻴﺴﺘﻢ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﻣﻤﻜﻦ ﺍﺳﺖ ﺍﺗﻔﺎﻕ ﺑﻴﺎﻓﺘﺪ ﺭﺍ ﻣﻲ ﺗﻮﺍﻥ ﺑﻪ ﺳﻪ ﺩﺳﺘﻪ‬
‫ﺗﻘﺴﻴﻢ ﺑﻨﺪﻱ ﻧﻤﻮﺩ‪:‬‬

‫‪ - 1‬ﺧﺮﺍﺑﻲ ﺗﺮﺍﻛﻨﺶ )‪(transaction failure‬‬

‫ﻣﻤﻜﻦ ﺍﺳﺖ ﺟﺎﻳﻲ ﺩﺭ ﺍﺟﺮﺍﻱ ﺗﺮﺍﻛﻨﺶ‪ ،‬ﻣﻨﻄﻖ ﺁﻥ ﺩﭼﺎﺭ ﺧﻄﺎ ﺷﻮﺩ ‪ .‬ﻟﺬﺍ ﺧﺮﺍﺑﻲ ﺗﺮﺍﻛﻨﺶ ﺷﺎﻣﻞ ﺩﻭ‬
‫ﻣﻮﺭﺩ ﺯﻳﺮ ﺍﺳﺖ‪:‬‬

‫‪ ‬ﺧﻄﺎﻱ ﻣﻨﻄﻘﻲ‪ :‬ﺗﺮﺍﻛﻨﺶ ﺑﻪ ﺩﻟﻴﻞ ﺷﺮﺍﻳﻂ ﺩﺍﺧﻠﻲ ﺧﻮﺩ ﻧﺘﻮﺍﻧﺪ ﻛﺎﻣﻞ ﺍﺟﺮﺍ ﺷﻮﺩ‪.‬‬
‫‪ ‬ﺧﻄﺎﻱ ﺳﻴﺴﺘﻤﻲ‪ :‬ﺗﺮﺍﻛﻨﺶ ﺑﻪ ﺧﻮﺩﻱ ﺧﻮﺩ ﺩﺭﺳﺖ ﻛﺎﺭ ﻣﻲ ﻛﻨﺪ ﺍﻣﺎ ﺩﺭ ﺳﻴﺴﺘﻢ ﺷﺮﺍﻳﻄﻲ ﭘﻴﺶ‬
‫ﻣﻲ ﺁﻳﺪ ﻛﻪ ﺍﻳﻦ ﺗﺮﺍﻛﻨﺶ ﺭﺍ ﺍﺯ ﻛﺎﺭ ﻣﻲ ﺍﻥﺩﺍﺯﺩ ‪ .‬ﻣﺎﻧﻨﺪ ﺑﻦ ﺑﺴﺖ‬
‫‪- 2‬ﺧﺮﺍﺑﻲ ﺳﻴﺴﺘﻢ )‪(system crash‬‬

‫ﺧﺮﺍﺑﻲ ﻫﺎﻱ ﺳﺨﺖ ﺍﻓﺰﺍﺭﻱ ﻳﺎ ﻧﺮﻡ ﺍﻓﺰﺍﺭﻱ ﻛﻪ ﺳﺒﺐ ﺍﺯ ﻛﺎﺭ ﺍﻓﺘﺎﺩﻥ ﺳﻴﺴﺘﻢ ﻣﻲ ﺷﻮﺩ ﻣﺎﻧﻨﺪ ﻗﻄﻊ‬
‫ﺑﺮﻕ‪ ،‬ﺩﺭ ﺍﻳﻦ ﻧﻮﻉ ﺧﺮﺍﺑﻲ ﻫﺎ ﺍﻃﻼﻋﺎﺕ ﺣﺎﻓﻈﻪ ﺍﺻﻠﻲ ﺳﻴﺴﺘﻢ ﺍﺯ ﺑﻴﻦ ﻣﻲ ﺭﻭﺩ ﻭﻟﻲ ﺁﺳﻴﺒﻲ ﺑﻪ‬
‫ﺍﻃﻼﻋﺎﺕ ﺭﻭﻱ ﺩﻳﺴﻚ ﻭﺍﺭﺩ ﻧﻤﻲ ﺷﻮﺩ‪.‬‬

‫‪- 3‬ﺧﺮﺍﺑﻲ ﺭﺳﺎﻧﻪ‬

‫‪39‬‬
 ‫ﺧﺮﺍﺑﻲ ﻛﻪ ﺳﺒﺐ ﺷﻮﺩ ﺍﻃﻼﻋﺎﺕ ﺭﻭﻱ ﺭﺳﺎﻧﻪ ﺍﺯ ﺑﻴﻦ ﺑﺮﻭﺩ ﻳﺎ ﻗﺎﺑﻞ ﺑﺎﺯﻳﺎﺑﻲ ﻧﺒﺎﺷﺪ‪ ،‬ﻣﺎﻧﻨﺪ ﺧﺮﺍﺑﻲ‬
‫ﺩﻳﺴﻚ ﻳﺎ ﺧﺮﺍﺑﻲ ﻫﺪ‪.‬‬

‫ﺍﻧﻮﺍﻉ ﺭﺳﺎﻧﻪ ﻫﺎﻱ ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ‬

‫ﺍﺯ ﺟﻬﺖ ﻗﺎﺑﻠﻴﺖ ﻧﮕﻬﺪﺍﺭﻱ ﺍﻃﻼﻋﺎﺕ ﺩﺭ ﺻﻮﺭﺕ ﻭﻗﻮﻉ ﺧﺮﺍﺑﻲ‪ ،‬ﺭﺳﺎﻧﻪ ﻫﺎﻱ ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ ﺑﻪ ﺳﻪ ﺩﺳﺘﻪ‬
‫ﺗﻘﺴﻴﻢ ﺑﻨﺪﻱ ﻣﻲ ﺷﻮﺩ‪:‬‬

‫‪ ‬ﺭﺳﺎﻧﻪ ﻫﺎﻱ ﻓﺮﺍﺭ )‪(volatile storage‬‬

‫ﺭﺳﺎﻧﻪ ﺍﻱ ﻛﻪ ﺩﺭ ﺻﻮﺭﺕ ﻭﻗﻮﻉ ﺧﺮﺍﺑﻲ ﺳﻴﺴﺘﻢ‪ ،‬ﺍﻃﻼﻋﺎﺕ ﺁﻥ ﺍﺯ ﺑﻴﻦ ﻣﻲ ﺭﻭﺩ‪ ،‬ﻣﺎﻧﻨﺪ ﺛﺒﺎﺕ‪.‬‬
‫‪ ‬ﺭﺳﺎﻧﻪ ﻫﺎﻱ ﻏﻴﺮﻓﺮﺍﺭ )‪(non-volatile storage‬‬
‫ﺩﺭ ﺻﻮﺭﺕ ﺧﺮﺍﺑﻲ ﺳﻴﺴﺘﻢ‪ ،‬ﺍﻃﻼﻋﺎﺕ ﺁﻧﻬﺎ ﻗﺎﺑﻞ ﺑﺎﺯﻳﺎﺑﻲ ﺍﺳﺖ ‪ .‬ﻣﺎﻧﻨﺪ ﺩﻳﺴﻚ ﻭ ﻧﻮﺍﺭﻣﻐﻨﺎﻃﻴﺴﻲ‬
‫‪ ‬ﺭﺳﺎﻧﻪ ﭘﺎﻳﺪﺍﺭ )‪(stable storage‬‬
‫ﺍﻳﻦ ﺭﺳﺎﻧﻪ ﺩﺭ ﺑﺮﺍﺑﺮ ﺗﻤﺎﻡ ﺧﺮﺍﺑﻲ ﻫﺎ ﻣﺺﻭﻥ ﻣﻲ ﺑﺎﺷﺪ ‪ .‬ﺍﻳﻦ ﻧﻮﻉ ﺭﺳﺎﻧﻪ ﺩﺭ ﻭﺍﻗﻊ ﻳﻚ ﻣﻔﻬﻮﻡ‬
‫ﻣﻨﻄﻘﻲ ﺍﺳﺖ ﻧﻪ ﻓﻴﺰﻳﻜﻲ ﻭ ﻫﻨﻮﺯ ﭼﻨﻴﻦ ﺭﺳﺎﻧﻪ ﺍﻱ ﻭﺟﻮﺩ ﺧﺎﺭﺟﻲ ﻧﺪﺍﺭﺩ ‪ .‬ﻣﺘﺪﺍﻭﻝ ﺗﺮﻳﻦ ﺭﺍﻫﻜﺎﺭ‪‌ ،‬‬
‫ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﭼﻨﺪﻳﻦ ﻛﭙﻲ ﺍﺯ ﺩﺍﺩﻩ ﻫﺎ )‪ (backup‬ﺭﻭﻱ ﺭﺳﺎﻧﻪ ﻫﺎﻱ ﻣﺨﺘﻠﻒ ﺍﺳﺖ ‪ .‬ﺑﺎ ﺍﻳﻦ ﻛﺎﺭ‬
‫ﺍﺣﺘﻤﺎﻝ ﺍﺯ ﺩﺳﺖ ﺭﻓﺘﻦ ﺍﻃﻼﻋﺎﺕ ﺩﺭ ﺻﻮﺭﺕ ﻭﻗﻮﻉ ﺧﺮﺍﺑﻲ ﻛﻤﺘﺮ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﭘﻴﺎﺩﻩﺳﺎﺯﻱ ﺣﺎﻓﻈﻪ ﺫﺧﻴﺮﻩﺳﺎﺯﻱ ﭘﺎﻳﺪﺍﺭ‬

‫ﻧﺴﺨﻪﻫﺎﻳﻲ ﺍﺯ ﻳﻚ ﺑﻼﻙ ﻣﻤﻜﻦ ﺍﺳﺖ ﺩﺭ ﻃﻮﻝ ﻋﻤﻠﻴﺎﺕ ﺧﺮﻭﺟﻲ‪ ،‬ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﺧﻄﺎﻫﺎﻱ ﻣﻮﺟﻮﺩ ﺩﺭ ﺁﻥ‬
‫ﻣﺘﻔﺎﻭﺕ ﺑﺎﺷﻨﺪ‪ .‬ﺑﺮﺍﻱ ﺗﺮﻣﻴﻢ ﻣﺮﺍﺣﻞ ﺯﻳﺮ ﺑﺎﻳﺪ ﺍﻧﺠﺎﻡ ﺷﻮﺩ ‪:‬‬

‫‪- 1‬ﭘﻴﺪﺍ ﻧﻤﻮﺩﻥ ﺑﻼﻛﻬﺎﻱ ﻧﺎﺳﺎﺯﮔﺎﺭ‬

‫ﺳﺎﺩﻩﺗﺮﻳﻦ ﺭﺍﻩ ﺑﺮﺍﻱ ﺍﻳﻦ ﻣﻨﻈﻮﺭ ﺍﻳﻦ ﺍﺳﺖ ﻛﻪ ﺩﻭ ﺑﻼﻙ ﺑﺎ ﻫﻢ ﻣﻘﺎﻳﺴﻪ ﺷﻮﻧﺪ ﺗﺎ ﺑﺒﻴﻨﻴﻢ ﺩﻭ ﻧﺴﺨﻪ ﺑﺎ‬
‫ﻫﻢ ﻓﺮﻕ ﺩﺍﺭﻧﺪ ﻳﺎ ﻧﻪ‪ .‬ﺍﻳﻦ ﺭﻭﺵ ﺯﻣﺎﻧﺒﺮ ﺍﺳﺖ ﭼﻮﻥ ﺍﺯ ﺣﺎﻓﻈﻪ ﺩﻳﺴﻚ ﻭﺍﻛﺸﻲ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫ﺭﺍﻩ ﺣﻞ ﺑﻬﺘﺮ ﺍﻳﻦ ﺍﺳﺖ ﻛﻪ ﺯﻣﺎﻧﻲ ﻛﻪ ﺍﻧﺘﻘﺎﻝ ﺩﺍﺩﻩ ﺍﻧﺠﺎﻡ ﻣﻲ ﺷﻮﺩ‪ ،‬ﺁﻥ ﺍﻧﺘﻘﺎﻝ ﺩﺍﺩﻩ ﺭﺍ ﺭﻭﻱ ﺩﻳﺴﻚ‬
‫ﺛﺒﺖ ﻛﻨﺪ ‪ .‬ﺭﻛﻮﺭﺩﻫﺎ ﺛﺒﺖ ﻣﻲ ﺷﻮﻧﺪ ﻭ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺁﻧﻬﺎ ﻣﻘﺎﻳﺴﻪ ﻣﻲ ﮔﺮﺩﺩ‪ .‬ﺩﺭ ﺍﻳﻦ ﺭﻭﺵ ﺍﺯ ﺍﺑﺘﺪﺍ‬

‫‪40‬‬
‫ﻓﻘﻂ ﺟﺎﻫﺎﻳﻲ ﻣﻘﺎﻳﺴﻪ ﻣﻲ ﺷﻮﺩ ﻛﻪ ﺑﺎ ﻫﻢ ﺍﺧﺘﻼﻑ ﺩﺍﺭﻧﺪ ‪ acknowledge .‬ﺩﺭﻳﺎﻓﺖ ﻧﺸﺪﻩ ﺍﺳﺖ‬
‫)ﺍﻃﻼﻋﺎﺕ ﻣﺘﻨﺎﻗﺾ(‪.‬‬

‫ﭼﻨﺪ ﻋﻤﻞ ﺑﺎ ﻫﻢ ﺩﺭ ﺻﻒ ﻗﺮﺍﺭ ﻣﻲ ﮔﻴﺮﺩ ﻭ ﺑﻼﻙ ﺑﻪ ﺑﻼﻙ ﺩﺭ ﺩﻳﺴﻚ ﺛﺒﺖ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪- 2‬ﺍﮔﺮ ﻫﺮ ﺩﻭ ﻛﭙﻲ ﺍﺯ ﺑﻼﻙ‪ ،‬ﻣﺘﻨﺎﻗﺺ ﻭ ﺩﺍﺭﺍﻱ ﻳﻚ ﺧﻄﺎ )‪ (Bad Checksum‬ﺗﺸﺨﻴﺺ ﺩﺍﺩﻩ‬
‫ﺵﺩ‪ ،‬ﺁﻥ ﺭﺍ ﺭﻭﻱ ﻳﻚ ﻛﭙﻲ ﺩﻳﮕﺮ ﺑﺎﺯﻧﻮﻳﺴﻲ ﻣﻲ ﻛﻨﻴﻢ‪ .‬ﺍﮔﺮ ﻫﺮ ﺩﻭ ﺑﺪﻭﻥ ﺧﻄﺎ ﺍﻣﺎ ﻣﺘﻔﺎﻭﺕ‬
‫ﻫﺴﺘﻨﺪ‪ ،‬ﺑﺎﺯﻧﻮﻳﺴﻲ ﺑﻼﻙ ﺩﻭﻡ ﺗﻮﺳﻂ ﺑﻼﻙ ﺍﻭﻝ ﺻﻮﺭﺕ ﻣﻲ ﮔﻴﺮﺩ )ﺍﻃﻼﻋﺎﺕ ‪ overwrite‬ﻣﻲ‪-‬‬
‫ﺷﻮﺩ(‪.‬‬

‫‪) RAID‬ﻧﺴﺨﻪﻫﺎﻱ ﻣﺘﻌﺪﺩ ﺍﺯ ﻳﻚ ﺑﻼﻙ (‪ :‬ﺍﺯ ﺩﻳﺴﻚ ﻫﺎﻱ ﻣﺴﺘﻘﻞ ﺍﺯ ﻫﻢ )ﺁﺭﺍﻳﻪ( ﻛﻪ ﺍﻓﺰﻭﻧﮕﻲ ﺩﺍﺩﻩ‬
‫ﺩﺍﺭﺩ‪ ،‬ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﻛﻨﺪ‪ .‬ﺑﻪ ﻋﺒﺎﺭﺕ ﺩﻳﮕﺮ ﺗﻜﻨﻴﻚ ‪ RAID‬ﺍﺯ ﭼﻨﺪﻳﻦ ﺩﻳﺴﻚ ﺳﺨﺖ ﺩﺭ ﺟﻬﺖ‬
‫ﺑﺎﻻﺑﺮﺩﻥ ﻛﻴﻔﻴﺖ ﻭ ﺍﻃﻤﻴﻨﺎﻥ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﻛﻨﺪ ﻛﻪ ﻃﺒﻖ ﻗﺎﻋﺪﻩ ﺧﺎﺻﻲ ﺑﻪ ﻫﻢ ﻣﺮﺗﺒﻂ ﻣﻲ ﺷﻮﻧﺪ ﻭ‬
‫ﻇﺮﻓﻴﺖ ﺑﺎﻻﻳﻲ ﺭﺍ ﺍﺯ ﺣﺎﻓﻈﻪ ﺑﺎ ﻛﻴﻔﻴﺖ ﺑﺎﻻ ﻣﻲ ﺳﺎﺯﻧﺪ‪.‬‬

‫ﻣﺰﺍﻳﺎﻱ ‪:RAID‬‬

‫• ﺍﻓﺰﻭﻧﮕﻲ‪ :‬ﺩﺭ ﻭﺍﻗﻊ ﺍﻓﺰﻭﻧﮕﻲ ﻛﭙﻲﻫﺎﻳﻲ ﻛﻪ ﺩﺍﺭﻳﻢ ﻭﺟﻮﺩ ﻧﺪﺍﺭﺩ ﻭ ﻧﻪ ﺍﻓﺰ ﻭﻧﮕﻲ ﺩﺍﺩﻩﻫﺎ‪ .‬ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ‬
‫ﺍﻳﻦ ﻭﻳﮋﮔﻲ‪ ،‬ﺍﮔﺮ ﺩﺭ ﻳﻚ ﻫﺎﺭﺩ ﺧﻄﺎ ﭘﻴﺶ ﺁﻳﺪ‪ ،‬ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺩﻳﮕﺮ ﺩﻳﺴﻜﻬﺎ ﺍﻃﻼﻋﺎﺕ ﺭﺍ ﺑﺎﺯﻳﺎﺑﻲ‬
‫ﻣﻲﻛﻨﻴﻢ‪.‬‬
‫• ﺑﺎﻻﺑﺮﺩﻥ ﻛﻴﻔﻴﺖ‬
‫• ﻫﺰﻳﻦ ﭘﺎﻳﻴﻦ‬

‫ﺭﻭﺍﻝ ﺩﺳﺘﺮﺳﻲ ﺑﻪ ﺩﺍﺩﻩ ﻫﺎ ﺑﺮﺍﻱ ﺍﻧﺠﺎﻡ ﺗﺮﺍﻛﻨﺶ‬

‫ﺩﺍﺩﻩ ﻫﺎﻱ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺕ ﺭﻭﻱ ﺭﺳﺎﻧﻪ ﻫﺎ ﻗﺮﺍﺭ ﺩﺍﺭﻧﺪ ﻛﻪ ﻣﻌﻤﻮﻝ ﺗﺮﻳﻦ ﺁﻧﻬﺎ ﺩﻳﺴﻚ ﻣﻲ ﺑﺎﺷﺪ‪ .‬ﻋﻤﺪﻩ‬
‫ﺗﺮﻳﻦ ﻋﺎﻣﻞ ﺳﺮﻋﺖ ﺩﺭ ﭘﺮﺩﺍﺯﺵ ﺍﻃﻼﻋﺎﺕ‪ ،‬ﺗﻌﺪﺍﺩ ﺩﻓﻌﺎﺕ ﺩﺳﺘﻴﺎﺑﻲ ﺑﻪ ﺭﺳﺎﻧﻪ ﺍﺳﺖ‪ .‬ﺩﺭ ﺑﺎﻧﻚ ﻫﺎﻱ‬
‫ﺍﻃﻼﻋﺎﺕ ﻧﺎﻣﺘﻤﺮﻛﺰ ﻋﻼﻭﻩ ﺑﺮ ﺍﻳﻦ ﻋﺎﻣﻞ‪ ،‬ﭘﻬﻨﺎﻱ ﺑﺎﻧﺪ ﻧﻴﺰ ﺍﻫﻤﻴﺖ ﺩﺍﺭﺩ‪.‬‬
‫‪41‬‬
‫ﺑﺮﺍﻱ ﻛﺎﻫﺶ ﺗﻌﺪﺍﺩ ﺩﻓﻌﺎﺕ ﻣﺮﺍﺟﻌﻪ ﺑﻪ ﺩﻳﺴﻚ‪ ،‬ﺭﻛﻮﺭﺩﻫﺎ ﺑﺎ ﻫﻢ ﺑﻠﻮﻙ ﺑﻨﺪﻱ ﻣﻲ ﺷﻮﻧﺪ ﻭ ﺩﺭ ﻫﺮ ﺑﺎﺭ‬
‫ﻣﺮﺍﺟﻌﻪ ﺑﻪ ﺩﻳﺴﻚ ﺑﺮﺍﻱ ﺧﻮﺍﻧﺪﻥ ﻳﺎ ﻧﻮﺷﺘﻦ‪ ،‬ﺑﻪ ﺟﺎﻱ ﻳﻚ ﺭﻛﻮﺭﺩ‪ ،‬ﻳﻚ ﺑﻠﻮﻙ ﺧﻮﺍﻧﺪﻩ ﻣﻲ ﺷﻮﺩ‪ .‬ﭘﺲ ﺍﺯ‬
‫ﺧﻮﺍﻧﺪﻥ ﻳﻚ ﺑﻠﻮﻙ‪ ،‬ﺍﻃﻼﻋﺎﺕ ﺁﻥ ﺑﻠﻮﻙ ﺑﻪ ﺑﺨﺸﻲ ﺍﺯ ﺣﺎﻓﻈﻪ ﺍﺻﻠﻲ ﻣﻨﺘﻘﻞ ﻣﻲ ﺷﻮﻧﺪ ﻛﻪ ﺑﺎﻓﺮ‬
‫)‪ (buffer‬ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﺩ‪ .‬ﺳﭙﺲ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺍﺯ ﺑﺎﻓﺮ ﻛﻪ ﺑﻪ ﺻﻮﺭﺕ ‪ RAM‬ﻫﺴﺘﻨﺪ ﻭ ﺳﺮﻋﺖ ﺧﻴﻠﻲ‬
‫ﺑﺎﻻﻳﻲ ﺩﺍﺭﻧﺪ‪ ،‬ﺭﻛﻮﺭﺩﻫﺎ ﺭﺍ ﺑﺮﺩﺍﺷﺘﻪ ﻭ ﻋﻤﻠﻴﺎﺕ ﻫﺎ ﺭﺍ ﺍﻧﺠﺎﻡ ﻣﻲ ﺩﻫﻨﺪ‪ .‬ﺧﺮﻭﺟﻲ ﺑﺮﻋﻜﺲ ﺍﻳﻦ ﻓﺮﺍﻳﻨﺪ‬
‫ﺍﺳﺖ؛ ﻳﻌﻨﻲ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺍﺯ ﻧﺎﺣﻴﻪ ﻛﺎﺭﻱ )‪ (work area‬ﺧﻮﺩ‪ ،‬ﺭﻛﻮﺭﺩﻫﺎ ﺭﺍ ﺭﻭﻱ ﺑﺎﻓﺮﻫﺎ ﻧﻮﺷﺘﻪ ﻭ ﺑﺎ‬
‫ﭘﺮﺷﺪﻥ ﺑﺎﻓﺮ‪ ،‬ﺑﻪ ﻳﻜﺒﺎﺭﻩ ﺑﻪ ﺩﻳﺴﻚ ﻣﻨﺘﻘﻞ ﻣﻲ ﺷﻮﻧﺪ‪.‬‬
‫ﻣﺮﺣﻠﻪ ﺍﻭﻝ ﺍﻳﻦ ﻛﺎﺭ ﻳﻌﻨﻲ ﺍﻧﺘﻘﺎﻝ ﺍﻃﻼﻋﺎﺕ ﺍﺯ ﺩﻳﺴﻚ ﺑﻪ ﺑﺎﻓﺮ ﻭ ﺑﺎﻟﻌﻜﺲ‪ input() ،‬ﻭ )(‪ output‬ﻧﺎﻡ‬
‫ﺩﺍﺭﺩ ﻭ ﻣﺮﺣﻠﻪ ﺩﻭﻡ ﻳﻌﻨﻲ ﺍﻧﺘﻘﺎﻝ ﺍﻃﻼﻋﺎﺕ ﺍﺯ ﺑﺎﻓﺮﻫﺎ ﺑﻪ ﻧﺎﺣﻴﻪ ﻫﺎﻱ ﻛﺎﺭﻱ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﻭ ﺑﺎﻟﻌﻜﺲ ﺑﻪ ﻧﺎﻡ‬
‫)(‪ read‬ﻭ )(‪ write‬ﻣﺮﺳﻮﻡ ﺍﺳﺖ‪.‬‬
‫ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﻣﺨﺘﻠﻒ ﺗﻌﺪﺍﺩﻱ ﺑﺎﻓﺮ ﺩﺍﺭﻧﺪ ﻛﻪ ﺩﺍﺭﺍﻱ ﺳﺎﻳﺰ ﻣﺸﺨﺼﻲ ﻫﺴﺘﻨﺪ‪ .‬ﺑﻨﺎﺑﺮﺍﻳﻦ ﺳﺎﻳﺰ ﺑﻠﻮﻙ ﻫﺎﻳﻲ‬
‫ﻛﻪ ﺗﻌﻴﻴﻦ ﻣﻲ ﻛﻨﻴﻢ‪ ،‬ﻳﻌﻨﻲ ﺗﻌﺪﺍﺩ ﺭﻛﻮﺭﺩﻫﺎﻳﻲ ﻛﻪ ﺩﺭ ﺑﻠﻮﻙ ﻫﺎ ﻗﺮﺍﺭ ﻣﻲ ﺩﻫﻴﻢ ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﺳﺎﻳﺰ ﺑﺎﻓﺮ‬
‫ﻣﻲ ﺑﺎﺷﺪ‪ .‬ﺍﮔﺮ ﺳﺎﻳﺰ ﺑﻠﻮﻙ ﻫﺎ ﺍﺯ ﺳﺎﻳﺰ ﺑﺎﻓﺮﻫﺎ ﻛﻤﺘﺮ ﺑﺎﺷﺪ ﺩﺭ ﺍﻳﻦ ﺻﻮﺭﺕ ﻓﻀﺎﻱ ﺑﺎﻓﺮ ﺭﺍ ﺍﺯ ﺩﺳﺖ ﺩﺍﺩﻩ‬
‫ﺍﻳﻢ ﻭ ﺍﮔﺮ ﺳﺎﻳﺰ ﺑﻠﻮﻙ ﻫﺎ ﺍﺯ ﺳﺎﻳﺰ ﺑﺎﻓﺮﻫﺎ ﺑﻴﺸﺘﺮ ﺑﺎﺷﺪ‪ ،‬ﺩﺭ ﺍﻳﻦ ﺻﻮﺭﺕ ﺑﺎﻳﺪ ﺑﺎﻓﺮﻫﺎ ﺭﺍ ﺑﻪ ﻫﻢ ﭼﺴﺒﺎﻧﺪ ﻭ‬
‫ﺁﻧﻬﺎ ﺭﺍ ﺧﻮﺍﻧﺪ ﻛﻪ ﻛﺎﺭ ﺑﺴﻴﺎﺭ ﺳﺨﺘﻲ ﺍﺳﺖ‪ .‬ﺑﻨﺎﺑﺮﺍﻳﻦ ﺳﺎﻳﺰ ﺑﻠﻮﻙ ﻫﺎ ﺑﺎﻳﺪ ﻫﺮ ﭼﻪ ﻧﺰﺩﻳﻜﺘﺮ ﺑﻪ ﺳﺎﻳﺰ ﺑﺎﻓﺮﻫﺎ‬
‫ﺑﺎﺷﺪ ﻭ ﻧﻪ ﺑﻴﺸﺘﺮ‪.‬‬

‫‪42‬‬
 ‫‪buffer‬‬
‫‪Buffer Block A‬‬ ‫)‪input(A‬‬
‫‪X‬‬ ‫‪A‬‬
‫‪Buffer Block B‬‬ ‫‪Y‬‬ ‫‪B‬‬
‫)‪output(B‬‬
‫)‪read(X‬‬
‫)‪write(Y‬‬

‫‪x2‬‬
‫‪x1‬‬
‫‪y1‬‬

‫‪work area work area‬‬

‫‪of T1‬‬ ‫‪of T2‬‬

‫‪memory‬‬ ‫‪disk‬‬

‫ﺩﺳﺘﺮﺳﻲ ﺑﻪ ﺩﺍﺩﻩ ﻫﺎ‪:‬‬

‫‪ ‬ﺑﻼﻛﻬﺎﻱ ﻓﻴﺰﻳﻜﻲ ﻛﻪ ﺭﻭﻱ ﺩﻳﺴﻚ ﻣﻘﻴﻢ ﻫﺴﺘﻨﺪ‪.‬‬
‫‪ ‬ﺑﻼﻙ ﻫﺎﻱ ﺑﺎﻓﺮﻱ ﻛﻪ ﺩﺭ ﺣﺎﻓﻈﻪ ﺍﺻﻠﻲ ﺑﺼﻮﺭﺕ ﻣﻮﻗﺖ ﻣﻘﻴﻢ ﺷﺪﻩ ﺍﻧﺪ ‪.‬‬
‫‪ ‬ﺍﻟﺒﺘﻪ ﺍﻧﺘﻘﺎﻝ ﺩﺍﺩﻩ ﺑﻴﻦ ﺩﻳﺴﻚ ﻭ ﺣﺎﻓﻈﻪ ﺍﺻﻠﻲ ﺍﺯ ﺩﻭ ﻃﺮﻳﻖ ﺻﻮﺭﺕ ﻣﻴﮕﻴﺮﺩ ‪:‬‬
‫‪ :Input ‬ﺍﻧﺘﻘﺎﻝ ﺑﻼﻙ ﻓﻴﺰﻳﻜﻲ )ﺑﻄﻮﺭ ﻣﺜﺎﻝ ﺑﻼﻙ ‪ (B‬ﺑﻪ ﺣﺎﻓﻈﻪ ﺍﺻﻠﻲ‪.‬‬

‫‪43‬‬
‫‪ :Output ‬ﺍﻧﺘﻘﺎﻝ ﺑﻼﻙ ﻓﻴﺰﻳﻜﻲ )ﺑﻄﻮﺭ ﻣﺜﺎﻝ ﺑﻼﻙ ‪ (B‬ﺍﺯ ﺭﻭﻱ ﺣﺎﻓﻈﻪ ﺍﺻﻠﻲ‬
‫ﺑﺮﻭﻱ ﺩﻳﺴﻚ‪.‬‬
‫‪ ‬ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺩﻳﺘﺎﺁﻳﺘﻢ ﻫﺎ ﺭﺍ ﺑﻴﻦ ﺑﻼﻛﻬﺎﻱ ﺑﺎﻓﺮ ﺩﺭ ﺳﻴﺴﺘﻢ ﺑﻪ ﻛﻤﻚ ﺩﻭ ﻋﻤﻠﻜﺮﺩ ﺯﻳﺮ ﺟﺎﺑﺠﺎ‬
‫ﻣﻴﻜﻨﺪ‪:‬‬
‫‪ ‬ﺧﻮﺍﻧﺪﻥ)‪ :(Read‬ﻣﻘﺪﺍﺭ ﺩﻳﺘﺎﺁﻳﺘﻢ )ﻣﺜﻼ ﺩﻳﺘﺎﺁﻳﺘﻢ ‪ (X‬ﺭﺍ ﺑﻪ ﻣﺘﻐﻴﺮ ﻣﺤﻠﻲ )ﻣﺜﻼ‬
‫ﻣﺘﻐﻴﺮ ﻣﺤﻠﻲ ‪ (Xi‬ﺗﻌﻠﻖ ﻣﻴﺪﻫﺪ‪.‬‬
‫‪ ‬ﻧﻮﺷﺘﻦ)‪ :(Write‬ﻣﻘﺪﺍﺭ ﻣﺤﻠﻲ )ﻣﺜﻼ ﻣﺘﻐﻴﺮ ﻣﺤﻠﻲ ‪ (Xi‬ﺭﺍ ﺑﺮﻭﻱ ﺩﻳﺘﺎﺁﻳﺘﻢ‬
‫)ﻣﺜﻼ ﺩﻳﺘﺎﺁﻳﺘﻢ ‪ (X‬ﻣﻮﺟﻮﺩ ﺩﺭ ﺑﺎﻓﺮ ﻧﺴﺒﺖ ﻣﻲ ﺩﻫﺪ‪.‬‬

‫ﺍﻟﮕﻮﺭﻳﺘﻢ ﻫﺎﻱ ﺗﺮﻣﻴﻢ‬

‫ﺑﺮﺍﻱ ﺗﻀﻤﻴﻦ ﺟﺎﻣﻌﻴﺖ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺕ ﻭ ﺍﻋﻤﺎﻝ ﺧﻮﺍﺹ ﻳﻜﭙﺎﺭﭼﮕﻲ ﻭ ﭘﺎﻳﺎﻳﻲ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺩﺭ ﺻﻮﺭﺕ‬
‫ﻭﻗﻮﻉ ﺧﺮﺍﺑﻲ‪ ،‬ﺍﺯ ﺍﻟﮕﻮﺭﻳﺘﻢ ﻫﺎﻳﻲ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﻛﻨﻴﻢ ﻛﻪ ﻋﻤﻮﻣﺎً ﺷﺎﻣﻞ ﺩﻭ ﻣﺮﺣﻠﻪ ﺯﻳﺮ ﻣﻲ ﺑﺎﺷﻨﺪ‪:‬‬

‫‪ .1‬ﻣﺮﺣﻠﻪ ﺍﻭﻝ ‪ :‬ﺩﺭ ﺣﻴﻦ ﻋﻤﻠﻜﺮﺩ ﻋﺎﺩﻱ ﺳﻴﺴﺘﻢ‪ ،‬ﺍﻃﻼﻋﺎﺗﻲ ﻛﻪ ﺑﺮﺍﻱ ﺍﻧﺠﺎﻡ ﺗﺮﻣﻴﻢ ) ﭘﺲ ﺍﺯ ﻭﻗﻮﻉ‬
‫ﺧﺮﺍﺑﻲ( ﺑﻪ ﺁﻧﻬﺎ ﻧﻴﺎﺯ ﺩﺍﺭﻳﻢ ﺩﺭ ﺟﺎﻳﻲ ﺛﺒﺖ ﺷﻮﻧﺪ ‪ .‬ﺑﻪ ﺍﻳﻦ ﻣﻜﺎﻥ ﻟﻮگ )‪ (log‬ﻣﻲ ﮔﻮﻳﻴﻢ‪.‬‬
‫‪ .2‬ﻣﺮﺣﻠﻪ ﺩﻭﻡ ‪ :‬ﭘﺲ ﺍﺯ ﻭﻗﻮﻉ ﺧﺮﺍﺑﻲ ) ﻭﻗﺘﻲ ﺳﻴﺴﺘﻢ ﺩﻭﺑﺎﺭﻩ ﺑﺎﻻ ﻣﻲ ﺁﻳﺪ (‪ ،‬ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺍﻃﻼﻋﺎﺕ‬
‫ﺛﺒﺖ ﺷﺪﻩ ﺩﺭ ﻣﺮﺣﻠﻪ ﻗﺒﻞ ﻭ ﺍﻟﮕﻮﺭﻳﺘﻢ ﻫﺎﻳﻲ ﻛﻪ ﺍﺯ ﭘﻴﺶ ﺗﻬﻴﻪ ﺷﺪﻩ ﺍﻧﺪ‪ ،‬ﺳﻴﺴﺘﻢ ﺭﺍ ﺑﻪ ﺣﺎﻟﺖ‬
‫ﺍﻭﻝ ﺑﺮﻣﻲ ﮔﺮﺩﺍﻧﻴﻢ‪.‬‬
‫ﭘﺲ ﺩﺭ ﻣﺮﺣﻠﻪ ﺩﻭﻡ ﺍﺯ ﺍﻟﮕﻮﺭﻳﺘﻢ ﺗﺮﻣﻴﻢ‪ ،‬ﻫﺮ ﺗﺮﺍﻛﻨﺸﻲ ﻛﻪ ﺑﺎ ﻣﻮﻓﻘﻴﺖ ﺍﻧﺠﺎﻡ )‪ (commit‬ﺷﺪﻩ ﺍﺳﺖ‬
‫ﺭﺍ ﺗﻜﺮﺍﺭ )‪ (redo‬ﻛﺮﺩﻩ‪ ،‬ﻳﻌﻨﻲ ﺩﻭﺑﺎﺭﻩ ﺍﺟﺮﺍ ﻣﻲ ﻛﻨﻴﻢ ﺗﺎ ﺍﺛﺮﺍﺗﺶ ﺭﻭﻱ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺕ ﺩﺍﺋﻤﻲ ﺷﻮﺩ ﻭ‬
‫ﻫﺮ ﺗﺮﺍﻛﻦﺷﻲ ﻛﻪ ﺳﺎﻗﻂ ﺷﺪﻩ ﺭﺍ ﺧﻨﺜﻲ )‪ (undo‬ﻣﻲ ﻛﻨﻴﻢ‪ .‬ﻓﺮﺽ ﻣﻲ ﻛﻨﻴﻢ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺳﺮﻳﺎﻟﻲ‬
‫ﻫﺴﺘﻨﺪ‪.‬‬
‫ﺷﻮﻥ ‪ :‬ﺭﻭﻳﻜﺮﺩ ﻛﺎﺭﻧﺎﻣﻪ )‪ (log-based‬ﻭ‬
‫ﺍﻟﮕﻮﺭﻳﺘﻢ ﻫﺎﻱ ﺗﺮﻣﻴﻢ ﺑﻪ ﺩﻭ ﺭﻭﻳﻜﺮﺩ ﻛﻠﻲ ﺗﻘﺴﻴﻢ ﻣﻲ ﺩ‬
‫ﺭﻭﻳﻜﺮﺩ ﺭﻭﻧﻮﺷﺖ )‪.(shadow paging‬‬

‫ﺭﻭﻳﻜﺮﺩ ﻛﺎﺭﻧﺎﻣﻪ )‪(log-based‬‬

‫‪44‬‬
‫ﺩﺭ ﺍﻳﻦ ﺭﻭﻳﻜﺮﺩ ﺍﻃﻼﻋﺎﺕ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﺑﺮﺍﻱ ﺍﻧﺠﺎﻡ ﺗﺮﻣﻴﻢ ﺭﺍ ﺩﺭﺣﺎﻓﻈﻪ ﭘﺎﻳﺪﺍﺭ ﺛﺒﺖ ﻣﻲ ﻛﻨﻴﻢ‪ .‬ﺑﺮﺍﻱ ﻫﺮ ﻳﻚ‬
‫ﻧﺎﻡ ﺭﻛﻮﺭﺩ ﻛﺎﺭﻧﺎﻣﻪ )‪ (log‬ﺑﺎ ﺳﺎﺧﺘﺎﺭ ﺯﻳﺮﻧﻮﺷﺘﻪ ﻣﻲ ﺷﻮﺩ‪:‬‬
‫ﺍﺯ ﺩﺳﺘﻮﺭﺍﺕ ﻳﻚ ﺗﺮﺍﻛﻨﺶ‪ ،‬ﺭﻛﻮﺭﺩﻱ ﺏ‬
‫‪ -‬ﺑﺮﺍﻱ ﺷﺮﻭﻉ ﺗﺮﺍﻛﻨﺶ ‪ ، Ti‬ﺭﻛﻮﺭﺩ > ‪< Ti , start‬‬
‫‪ -‬ﺑﺮﺍﻱ ﺩﺳﺘﻮﺭ )‪ write(x‬ﺍﺯ ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﺩﺭ ﺣﺎﻟﺖ ﻛﻠﻲ ﺭﻛﻮﺭﺩ > ‪ <Ti ,x,V1,V2‬ﻛﻪ‬
‫‪ V2 , V1‬ﺑﻪ ﺗﺮﺗﻴﺐ ﻣﻘﺎﺩﻳﺮ ﻗﺒﻞ ﻭ ﺑﻌﺪ ﺍﺯ ﺍﻧﺠﺎﻡ ﻧﻮﺷﺘﻦ ‪ x‬ﻣﻲ ﺑﺎﺷﻨﺪ‪.‬‬
‫‪ -‬ﺑﺎ ﺍﺟﺮﺍﻱ ﺁﺧﺮﻳﻦ ﺩﺳﺘﻮﺭ ﺗﺮﺍﻛﻨﺶ ‪ ،Ti‬ﺭﻛﻮﺭﺩ> ‪ <Ti, Commit‬ﺩﺭ ﻛﺎﺭﻧﺎﻣﻪ ﺛﺒﺖ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﺭﻭﻳﻜﺮﺩ ﺭﻭﻧﻮﺷﺖ )‪(shadow paging‬‬

‫ﺑﺮﺍﻱ ﭘﻴﺎﺩﻩ ﺳﺎﺯﻱ ﺍﻳﻦ ﺭﻭﻳﻜﺮﺩ ﺭﻭﻧﻮﺷﺖ ﻧﻴﺎﺯ ﺑﻪ ﻛﺎﺭﻧﺎﻣﻪ)‪ (log‬ﻣﻲ ﺑﺎﺷﺪ ‪ .‬ﻗﺒﻞ ﺍﺯ ﺍﻧﺠﺎﻡ ﺗﻐﻴﻴﺮﺍﺕ )ﻗﺒﻞ‬
‫ﺍﺯ ﺷﺮﻭﻉ ﺑﻪ ﺍﺟﺮﺍﻱ ﺗﺮﺍﻛﻦﺵ ( ﻳﻚ ﻛﭙﻲ ﺍﺯ ﺻﻔﺤﺎﺕ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﺁﻥ ﺗﺮﺍﻛﻨﺶ ﺍﺯ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺕ ﺗﻬﻴﻪ ﻣﻲ‬
‫ﻛﻨﻴﻢ‪ .‬ﺑﻪ ﺍﻳﻦ ﻧﺴﺨﻪ‪ ،‬ﻧﺴﺨﻪ ﺟﺎﺭﻱ )‪ (current‬ﻣﻲ ﮔﻮﻳﻴﻢ‪ .‬ﺗﻐﻴﻴﺮﺍﺕ ﻣﻮﺭﺩﻧﻈﺮ ﺗﺮﺍﻛﻨﺶ ﺭﻭﻱ ﻧﺴﺨﻪ‬
‫ﺟﺎﺭﻱ ﺻﻮﺭﺕ ﻣﻲ ﮔﻴﺮﺩ ﺑﻨﺎﺑﺮﺍﻳﻦ ﻧﺴﺨﻪ ﺍﺻﻠﻲ ﺩﺳﺖ ﻧﺨﻮﺭﺩﻩ ﺑﺎﻗﻲ ﻣﻲ ﻣﺎﻧﺪ‪.‬‬
‫ﭘﺲ ﻣﺎ ﺩﻭ ﻧﺴﺨﻪ ﺩﺍﺭﻳﻢ‪ :‬ﻧﺴﺨﻪ ﺍﺻﻠﻲ ﻛﻪ ﻧﺴﺨﻪ ﺳﺎﻳﻪ )‪ (shadow‬ﻧﻴﺰ ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﺩ ﺑﻪ ﺩﻟﻴﻞ‬
‫ﺍﻳﻨﻜﻪ ﻭﻗﺘﻲ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﻣﻮﻓﻖ ﻣﻲ ﺷﻮﻧﺪ‪ ،‬ﺁﻥ ﻧﺴﺨﻪ ﺩﻭﺭ ﺭﻳﺨﺘﻪ ﻣﻲ ﺷﻮﺩ ﻭ ﻧﺴﺨﻪ ﺩﻳﮕﺮ‪ ،‬ﻧﺴﺨﻪ ﺟﺎﺭﻱ‬
‫ﻛﻪ ﺗﻐﻴﻴﺮﺍﺕ ﺭﺍ ﺭﻭﻱ ﺁﻥ ﺍﻧﺠﺎﻡ ﻣﻲ ﺩﻫﻴﻢ‪ .‬ﭼﻨﺎﻧﭽﻪ ﺗﺮﺍﻛﻨﺶ ﺑﻪ ﺍﻧﺠﺎﻡ ﺭﺳﻴﺪ‪ ،‬ﺍﻳﻦ ﻧﺴﺨﻪ ﺟﺎﺭﻱ ﺭﺍ ﺑﻪ‬
‫ﻋﻨﻮﺍﻥ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺕ ﺟﺪﻳﺪ ﺗﻠﻘﻲ ﻣﻲ ﻛﻨﻴﻢ ﻭ ﻧﺴﺨﻪ ﺳﺎﻳﻪ ﺭﺍ ﺍﺯ ﺑﻴﻦ ﻣﻲ ﺑﺮﻳﻢ‪ .‬ﺍﻣﺎ ﺍﮔﺮ ﺗﺮﺍﻛﻨﺶ‬
‫ﻧﺘﻮﺍﻧﺴﺖ ﺍﻧﺠﺎﻡ ﺷﻮﺩ‪ ،‬ﻧﺴﺨﻪ ﺟﺎﺭﻱ ﺭﺍ ﺍﺯ ﺑﻴﻦ ﺑﺮﺩﻩ ﻭ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺕ ﻣﻌﺎﺩﻝ ﻫﻤﺎﻥ ﻧﺴﺨﻪ ﺳﺎﻳﻪ ﺧﻮﺍﻫﺪ‬
‫ﺑﻮﺩ‪.‬‬

‫ﻣﺰﺍﻳﺎﻱ ﺭﻭﻳﻜﺮﺩ ﺭﻭﻧﻮﺷﺖ‪:‬‬

‫‪ (1‬ﺳﺮﺑﺎﺭ ﻣﺮﺑﻮﻁ ﺑﻪ ﻧﻮﺷﺘﻦ ﺭﻛﻮﺭﺩﻫﺎﻱ ﻛﺎﺭﻧﺎﻣﻪ ﺭﺍ ﻧﺪﺍﺭﺩ‪.‬‬
‫‪ (2‬ﺍﻧﺠﺎﻡ ﻋﻤﻠﻴﺎﺕ ﺗﺮﻣﻴﻢ ﺑﺴﻴﺎﺭ ﺳﺎﺩﻩ ﻭ ﻧﺎﭼﻴﺰ ﺍﺳﺖ‪.‬‬
‫ﻣﻌﺎﻳﺐ ﺭﻭﻳﻜﺮﺩ ﺭﻭﻧﻮﺷﺖ ‪:‬‬
‫‪ . 1‬ﺑﺮﺍﻱ ﭘﻴﺎﺩﻩ ﺳﺎﺯﻱ ﺁﻥ ﻧﻴﺎﺯ ﺑﻪ ﻛﺎﺭﻧﺎﻣﻪ ﺩﺍﺭﻳﻢ‪.‬‬

‫‪45‬‬
‫‪ . 2‬ﻛﭗﻱ ﻛﺮﺩﻥ ﺍﻃﻼﻋﺎﺕ ﻣﺮﺑﻮﻃﻪ ﺍﺯ ﺑﺎﻧﻚ ﺑﺴﻴﺎﺭ ﭘﺮﻫﺰﻳﻨﻪ ﺍﺳﺖ )ﺣﺘﻲ ﺑﺎ ﻭﺟﻮﺩ ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﻫﺎﻱ‬
‫ﺍﻧﺠﺎﻡ ﺷﺪﻩ ﺩﺭ ﺍﻳﻦ ﺯﻣﻴﻨﻪ(‬
‫‪ . 3‬ﺗﻮﺳﻌﻪ ﺍﻳﻦ ﺭﻭﺵ ﺑﺮﺍﻱ ﺣﺎﻟﺘﻲ ﻛﻪ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺑﺘﻮﺍﻧﻨﺪ ﻫﻤﺮﻭﻧﺪ ﺍﺟﺮﺍ ﺷﻮﻧﺪ ﺑﺴﻴﺎﺭ ﻣﺸﻜﻞ ﺍﺳﺖ )ﺑﺮ‬
‫ﺧﻼﻑ ﺭﻭﺵ ﻛﺎﺭﻧﺎﻣﻪ(‪.‬‬
‫‪ . 4‬ﺳﺮﺑﺎﺭ ﺍﻧﺠﺎﻡ ﻋﻤﻠﻴﺎﺕ ﺍﻧﺠﺎﻡ ﺯﻳﺎﺩ ﺍﺳﺖ‪.‬‬
‫ﻧﺘﻴﺠﻪ‪ :‬ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﻣﻘﺎﻳﺴﻪ ﺍﺟﻤﺎﻟﻲ ﺍﻳﻦ ﺩﻭ ﺭﻭﻳﻜﺮﺩ‪ ،‬ﺩﺭ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺕ ﺭﻭﻳﻜﺮﺩ ﻛﺎﺭﻧﺎﻣﻪ‬
‫ﺍﺯ ﺭﻭﻳﻜﺮﺩ ﺭﻭﻧﻮﺷﺖ ﺑﻬﺘﺮ ﺍﺳﺖ ﻭ ﻋﻤﻮﻣﺎً ﺍﺯ ﺭﻭﺵ ﻛﺎﺭﻧﺎﻣﻪ ﺑﺮﺍﻱ ﺍﻧﺠﺎﻡ ﺗﺮﻣﻴﻢ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﺭﻭﻳﻜﺮﺩ ﻛﺎﺭﻧﺎﻣﻪ ﻣﻌﺎﻳﺒﻲ ﺩﺍﺭﺩ ﻛﻪ ﻣﻬﻤﺘﺮﻳﻦ ﺁﻧﻬﺎ ﻋﺒﺎﺭﺗﻨﺪ ﺍﺯ‪:‬‬

‫‪ .1‬ﺑﺰﺭگ ﺷﺪﻥ ﺍﻧﺪﺍﺯﻩ ﻓﺎﻳﻞ ﻛﺎﺭﻧﺎﻣﻪ‪.‬‬
‫‪ .2‬ﺯﻣﺎﻥ گﻳﺮ ﺑﻮﺩﻥ ﺟﺴﺘﺠﻮ ﺩﺭ ﻛﺎﺭﻧﺎﻣﻪ‪.‬‬
‫‪ .3‬ﺍﺣﺘﻤﺎﻝ ﺗﻜﺮﺍﺭ ﻣﺠﺪﺩ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻳﻲ ﻛﻪ ﻗﺒﻼً ﺁﻧﻬﺎ ﺭﺍ ﺗﻜﺮﺍﺭ ﻛﺮﺩﻩ ﺍﻳﻢ‪.‬‬
‫ﺭﻭﺯﺭﺳﺎﻧﻲ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺕ ﺑﻪ ﺩﻟﻴﻞ ﻛﺎﺭ ﺑﺎ ﺭﺳﺎﻧﻪ ﻫﺎ‪.‬‬ ‫‪ .4‬ﺍﻓﺰﺍﻳﺶ ﻫﺰﻳﻨﻪ ﺏ‬
‫ﺑﺮﺍﻱ ﺑﻬﺒﻮﺩ ﺍﻳﻦ ﻣﻌﺎﻳﺐ ﺭﻭﺷﻬﺎﻳﻲ ﺍﺭﺍﺋﻪ ﺷﺪﻩ ﻛﻪ ﺍﺯ ﺑﻬﺘﺮﻳﻦ ﺁﻥﻫﺎ‪ ،‬ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻧﻘﻄﻪ ﺑﺎﺯﺭﺳﻲ‬
‫)‪ (checkpoint‬ﻣﻲ ﺑﺎﺷﺪ‪.‬‬

‫ﻧﻘﻄﻪ ﺑﺎﺯﺭﺳﻲ )‪(checkpoint‬‬

‫ﻃﻮﺭ ﻣﺘﻨﺎﻭﺏ )‪ (periodic‬ﺩﺭ ﺑﺮﻫﻪ ﻫﺎﻱ ﺯﻣﺎﻧﻲ ﻣﻌﻴﻨﻲ ﺍﺯ ﻋﻤﻠﻜﺮﺩ ﻋﺎﺩﻱ ﺳﻴﺴﺘﻢ‪،‬‬ ‫ﺩﺭ ﺍﻳﻦ ﺭﻭﺵ‪ ،‬ﺏ‬
‫ﻛﺎﺭﻫﺎﻱ ﺍﻧﺠﺎﻡ ﺷﺪﻩ ﺭﻭﻱ ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺕ )ﺗﺎ ﺁﻥ ﺯﻣﺎﻥ( ﺭﺍ ﻗﻄﻌﻲ ﻭ ﻧﻬﺎﻳﻲ ﻣﻲ ﻛﻨﻴﻢ ﻛﻪ ﺍﻳﻦ ﻟﺤﻈﺎﺕ‬
‫ﻧﻘﺎﻁ ﺑﺎﺯﺭﺳﻲ ﺳﻴﺴﺘﻢ ﻧﺎﻡ ﺩﺍﺭﻧﺪ‪ .‬ﺑﻨﺎﺑﺮﺍﻳﻦ ﺍﮔﺮ ﺍﺗﻔﺎﻗﻲ ﺩﺭ ﺳﻴﺴﺘﻢ ﺭﺥ ﺑﺪﻫﺪ‪ ،‬ﻧﺒﺎﻳﺪ ﺗﺎ ﺍﺑﺘﺪﺍﻱ ‪ log‬ﺭﺍ‬
‫ﺑﺮﺭﺳﻲ ﻛﻨﻴﻢ‪ ،‬ﺑﻠﻜﻪ ﻛﺎﻓﻴﺴﺖ ﺗﺎ ﺍﻭﻟﻴﻦ ﻧﻘﻄﻪ ﺑﺎﺯﺭﺳﻲ )‪ (checkpoint‬ﺍﻳﻦ ﻛﺎﺭ ﺭﺍ ﺍﻧﺠﺎﻡ ﺩﻫﻴﻢ‪.‬‬
‫ﺑﻌﺒﺎﺭﺕ ﺩﻳﮕﺮ ﻛﺎﻓﻴﺴﺖ ‪ log‬ﺭﺍ ﺍﺯ ﺁﺧﺮ ﺷﺮﻭﻉ ﻛﻨﻴﻢ‪ ،‬ﺍﻃﻼﻋﺎﺕ ﺁﻧﺮﺍ ﻣﻮﺭﺩ ﺍﺳﺘﻔﺎﺩﻩ ﻗﺮﺍﺭ ﺩﻫﻴﻢ‪ ،‬ﺳﻴﺴﺘﻢ‬
‫ﺭﺍ ﺗﺮﻣﻴﻢ ﻛﻨﻴﻢ ﻭ ﺯﻣﺎﻧﻲ ﻛﻪ ﺑﻪ ﺍﻭﻟﻴﻦ ﻧﻘﻄﻪ ﺑﺎﺯﺭﺳﻲ ﺭﺳﻴﺪﻳﻢ ﻣﺘﻮﻗﻒ ﺷﻮﻳﻢ‪.‬‬

‫ﺩﺭ ﻫﺮ ﻧﻘﻄﻪ ﺑﺎﺯﺭﺳﻲ ﻣﺠﻤﻮﻋﻪ ﻋﻤﻠﻴﺎﺕ ﺯﻳﺮ ﺑﻪ ﺍﻧﺠﺎﻡ ﻣﻲ ﺭﺳﻨﺪ‪:‬‬

‫‪46‬‬
‫‪ .1‬ﺭﻛﻮﺭﺩﻫﺎﻱ ﻛﺎﺭﻧﺎﻣﻪ )ﻛﻪ ﺩﺭ ﺣﻴﻦ ﻋﻤﻠﻜﺮﺩ ﻋﺎﺩﻱ ﺳﻴﺴﺘﻢ ﺩﺭ ﺑﺎﻓﺮ ﻧﻮﺷﺘﻪ ﻣﻲ ﺷﺪﻧﺪ ( ﺑﻪ ﺣﺎﻓﻈﻪ‬
‫ﭘﺎﻳﺪﺍﺭ ﻣﻨﺘﻘﻞ ﻣﻲ ﺷﻮﻧﺪ‪.‬‬
‫‪ .2‬ﺩﺍﺩﻩ ﻫﺎﻱ ﺗﻐﻴﻴﺮ ﻳﺎﻓﺘﻪ ﺩﺭ ﺑﺎﻓﺮ ﺑﻪ ﺩﻳﺴﻚ ﻣﻨﺘﻘﻞ ﻣﻲ ﮔﺮﺩﻧﺪ‪.‬‬
‫ﻧﺎﻡ ﺭﻛﻮﺭﺩ ﺑﺎﺯﺭﺳﻲ ﺑﺎ ﺳﺎﺧﺘﺎﺭ >‪ <checkpoint‬ﺛﺒﺖ ﻣﻲ ﺷﻮﺩ‪.‬‬ ‫‪ .3‬ﺭﻛﻮﺭﺩﻱ ﺏ‬

‫ﻣﺜﺎﻝ‪ :‬ﺩﺭ ﺍﻳﻦ ﺷﻜﻞ ﭘﺲ ﺍﺯ ﻭﻗﻮﻉ ﺧﺮﺍﺑﻲ ﺑﻪ ﺻﻮﺭﺕ ﺯﻳﺮ ﻋﻤﻞ ﻣﻲ ﺷﻮﺩ‪:‬‬

‫‪ -‬ﺗﺮﺍﻛﻨﺶ ‪ T1‬ﺩﺭ ﺯﻣﺎﻥ ‪ T2‬ﻧﻬﺎﻳﻲ ﺷﺪﻩ ﺍﺳﺖ ﻭ ﻧﻴﺎﺯ ﺑﻪ ﺗﺮﻣﻴﻢ ﻧﺪﺍﺭﺩ‪.‬‬

‫‪ -‬ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ‪ T2‬ﻭ ‪ T3‬ﺭﺍ ﺑﻨﺎ ﺑﻪ ﺧﺎﺻﻴﺖ ﭘﺎﻳﺎﻳﻲ ﺑﺎﻳﺪ ﺗﻜﺮﺍﺭ ﻧﻤﻮﺩ ﺗﺎ ﺍﺛﺮ ﺁﻧﻬﺎ ﺩﺭ ﺳﻴﺴﺘﻢ ﻧﻬﺎﻳﻲ‬
‫ﻭ ﺩﺍﺋﻤﻲ ﺷﻮﺩ ) ﺑﺮﺍﻱ ‪ T2‬ﻓﻘﻂ ﺑﺨﺸﻲ ﺍﺯ ﺩﺳﺘﻮﺭﺍﺕ ﻛﻪ ﭘﺲ ﺍﺯ ‪ Tc‬ﺍﺟﺮﺍ ﺷﺪﻩ ﺍﻧﺪ ﺭﺍ ﺗﻜﺮﺍﺭ ﻣﻲ‬
‫ﻛﻨﻴﻢ(‪.‬‬
‫‪ -‬ﺗﺮﺍﻛﻨﺶ ‪ T4‬ﻛﻪ ﻧﻴﻤﻪ ﻛﺎﺭﻩ ﻣﺎﻧﺪﻩ ﺍﺳﺖ ﺭﺍ ﺑﻨﺎ ﺑﻪ ﺧﺎﺻﻴﺖ ﻳﻜﭙﺎﺭﭼﮕﻲ‪ ،‬ﺧﻨﺜﻲ ﻣﻲ ﻛﻨﻴﻢ ‪.‬‬

‫‪Fuzzy Checkpoint‬‬

‫ﺗﺎ ﻛﻨﻮﻥ ﻓﺮﺽ ﺑﺮ ﺍﻳﻦ ﺑﻮﺩ ﻛﻪ ﺗﺮﺍﻛﻨﺶﻫﺎ ﺩﺭ ﺣﻴﻦ ﺍﻳﺠﺎﺩ ﻧﻘﻄﻪ ﺑﺮﺭﺳﻲ )ﭼﻚﭘﻮﻳﻨﺖ( ﻓﻌﺎﻝ ﻧﻴﺴﺘﻨﺪ‪ .‬ﺍﻣﺎ‬
‫ﺩﺭ ﺑﺮﺧﻲ ﺍﺯ ﺳﻴﺴﺘﻤﻬﺎ ﻳﻚ ﺳﺮﻱ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺍﺟﺎﺯﻩ ﺩﺍﺭﻧﺪ ﺩﺭ ﺣﻴﻦ ﭼﻚ ﭘﻮﻳﻨﺖ ﻓﻌﺎﻝ ﺑﺎﺷﻨﺪ ‪ .‬ﻫﺪﻑ‬
‫ﭼﻚﭘﻮﻳﻨﺖ ﻓﺎﺯﻱ ﺍﻳﻦ ﺍﺳﺖ ﻛﻪ ﺍﻳﻦ ﺗﺮﺍﻛﻨﺸﻬﺎ ﺭﺍ ﺗﺸﺨﻴﺺ ﺩﻫﺪ ﻳﻌﻨﻲ ﺧﺮﻭﺟﻲ ﺗﺮﺍﻛﻨﺸﻬﺎ ﻣﺠﺎﺯ ﺍﺳﺖ ‪.‬‬
‫ﻗﺒﻞ ﺍﺯ ﺁﺧﺮﻳﻦ ﭼﻚﭘﻮﻳﻨﺖ ﻫﻨﻮﺯ ﺑﻌﻀﻲﻫﺎ ﺍﺟﺮﺍ ﻧﺸﺪﻩﺍﻧﺪ‪.‬‬
‫‪47‬‬
‫ﺑﺮﺍﺳﺎﺱ ﻓﺎﻛﺘﻮﺭﻫﺎﻱ ﺗﺮﺍﻛﻨﺶ‪ ،‬ﺁﻧﻬﺎﻳﻲ ﻛﻪ ﺩﺭﮔﻴﺮ ﻫﺴﺘﻨﺪ‪ ،‬ﺑﺎﻳﺪ ﻣﺘﻮﻗﻒ ﺷﻮﻧﺪ ‪ .‬ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﮔﺮﺍﻓﻬﺎﻱ‬
‫ﻧﺎﻫﻤﺒﻨﺪ ﻣﻲﺗﻮﺍﻥ ﺗﺮﺍﻛﻨﺸﻬﺎ ﺭﺍ ﺑﻪ ﺩﻭ ﺩﺳﺘﻪ ﺗﻘﺴﻴﻢ ﻛﺮﺩ‪.‬‬

‫ﺧﻄﺎ ﺑﺎ ﺍﺯ ﺩﺳﺖ ﺩﺍﺩﻥ ﺣﺎﻓﻈﻪ ﺫﺧﻴﺮﻩﺳﺎﺯﻱ ﻏﻴﺮﻓﺮﺍﺭ‬

‫ﺩﺭﺻﻮﺭﺗﻲ ﻛﻪ ﺑﺨﻮﺍﻫﻴﻢ ﺑﺎ ﻣﺸﻜﻞ ﺍﺯ ﺩﺳﺖ ﺩﺍﺩﻥ ﺩﺍﺩﻩ ﻫﺎﻱ ﺣﺎﻓﻈﻪ ﻏﻴﺮﻓﺮﺍﺭ ﻣﻘﺎﺑﻠﻪ ﻛﻨﻴﻢ‪ ،‬ﺗﻜﻨﻴﻜﻲ‬
‫)‪ (Dumping‬ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﺩ ‪.‬‬ ‫ﻣﺸﺎﺑﻪ ﻧﻘﺎﻁ ﻭﺍﺭﺳﻲ ﻗﺎﺑﻞ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺳﺖ ﻛﻪ ﺍﺻﻄﻼﺣﺎً ﺭﻭﺑﺮﺩﺍﺭﻱ‬
‫ﺏ ﺻﻮﺭﺕ ﺯﻳﺮ ﺍﺳﺖ ‪:‬‬ ‫ﻣﺮﺍﺣﻞ ﺍﻳﻦ ﺗﻜﻨﻴﻚ ﺑﻪ ﻃﻮﺭ ﺧﻼﺻﻪ ﻩ‬
‫‪- 1‬ﻫﺮ ﭼﻪ ﺭﻛﻮﺭﺩ ﺩﺭ ‪ log‬ﻭﺟﻮﺩ ﺩﺍﺭﺩ‪ ،‬ﺩﺭ ﺣﺎﻓﻈﻪ ﭘﺎﻳﺪﺍﺭ ﺛﺒﺖ ﺷﻮﺩ ‪.‬‬
‫‪- 2‬ﺩﺍﺩﻩﻫﺎ ﺛﺒﺖ ﺷﻮﺩ )ﻛﻠﻴﻪ ﺑﻼﻛﻬﺎﻱ ﺑﺎﻓﺮ ﺑﺮﺭﻭﻱ ﺩﻳﺴﻚ ﺛﺒﺖ ﺷﻮﺩ (‪.‬‬
‫‪- 3‬ﺍﻧﺠﺎﻡ ﻋﻤﻠﻴﺎﺕ ‪) Dumping‬ﺭﻭﺑﺮﺩﺍﺭﻱ( ﺭﻭﻱ ﺩﻳﺴﻚ‬
‫‪- 4‬ﺛﺒﺖ ﻳﻚ ﻛﻮﺭﺩ ﺑﻪ ﺷﻜﻞ >‪ <Dump‬ﺑﺮﺭﻭﻱ ﺣﺎﻓﻈﻪ ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ ﭘﺎﻳﺪﺍﺭ‬
‫ﻧﻜﺘﻪ‪ :‬ﺗﺮﺗﻴﺐ ﺍﺟﺮﺍﻱ ‪ Checkpoint‬ﻭ ‪ Dumping‬ﻫﻴﭻ ﺍﺭﺗﺒﺎﻃﻲ ﺑﺎ ﻫﻢ ﻧﺪﺍﺭﻧﺪ ‪ .‬ﺩﺭ ‪Checkpoint‬‬
‫ﺍﻧﺘﻘﺎﻝ ﺍﺯ ﺑﺎﻓﺮ ﺑﻪ ﺩﻳﺴﻚ ﺻﻮﺭﺕ ﻣﻲ ﮔﻴﺮﺩ ﺍﻣﺎ ﺩﺭ ‪ Dumping‬ﺍﻧﺘﻘﺎﻝ ﺍﺯ ﺩﻳﺴﻚ ﺑﻪ ﺩﻳﺴﻚ ﺻﻮﺭﺕ ﻣﻲ ‪-‬‬
‫ﮔﻴﺮﺩ‪.‬‬

‫‪(Algorithms for Recovery and Isolation Exploiting Semantics) ARIES‬‬

‫ﺑﺪﻳﻬﻲ ﺍﺳﺖ ﻛﻪ ﺁﻧﭽﻪ ﺩﺭ ﻣﻮﺭﺩ ﺭﻭﻳﻪ ﺗﺮﻣﻴﻢ ﺳﻴﺴﺘﻢ ﮔﻔﺘﻪ ﻣﻲ ﺷﻮﺩ‪ ،‬ﺑﺴﻴﺎﺭ ﺳﺎﺩﻩ ﺍﺳﺖ ‪ .‬ﺩﺭ ﺁﻥ ﺭﻭﻳﻪ‪،‬‬
‫ﻋﻤﻞ "ﻟﻐﻮ ﻛﺮﺩﻥ " ﻗﺒﻞ ﺍﺯ ﻋﻤﻞ "ﻛﺎﺭ ﻣﺠﺪﺩ " ﺻﻮﺭﺕ ﻣﻲ ﮔﻴﺮﺩ ‪ .‬ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﺍﻭﻟﻴﻪ ﺑﻪ ﺍﻳﻦ ﺭﻭﺵ ﻛﺎﺭ‬
‫ﻣﻲ ﻛﺮﺩﻧﺪ‪ ،‬ﺍﻣﺎ ﺑﺪﻟﻴﻞ ﻛﺎﺭﺍﻳﻲ‪ ،‬ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﺟﺪﻳﺪ ﺍﺯ ﺭﻭﺵ ﺩﻳﮕﺮﻱ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﻛﻨﻨﺪ ‪ .‬ﺩﺭ ﻭﺍﻗﻊ‪ ،‬ﺍﻏﻠﺐ‬
‫ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﺍﻣﺮﻭﺯﻱ‪ ،‬ﺍﺯ ﺭﻭﺵ ‪ ARIES‬ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﻛﻨﺪ‪ ARIES .‬ﺩﺭ ﺳﻪ ﻣﺮﺣﻠﻪ ﺍﻧﺠﺎﻡ ﻣﻲ ﺷﻮﺩ ‪:‬‬
‫‪ - 1‬ﺗﺤﻠﻴﻞ‪ ،‬ﻟﻴﺴﺖ ﻫﺎﻱ ‪ UNDO‬ﻭ ‪ REDO‬ﺭﺍ ﺗﺸﻜﻴﻞ ﻣﻲ ﺩﻫﺪ‪.‬‬
‫‪ - 2‬ﻛﺎﺭ ﻣﺠﺪﺩ ‪ ،‬ﺍﺯ ﻣﻜﺎﻧﻲ ﺷﺮﻭﻉ ﻣﻲ ﺷﻮﺩ ﻛﻪ ﺩﺭ ﻣﺮﺣﻠﻪ ﺗﺤﻠﻴﻞ ﻣﺸﺨﺺ ﺷﺪﻩ ﺍﺳﺖ ﻭ‬
‫ﺑﺎﻧﻚ ﺍﻃﻼﻋﺎﺗﻲ ﺭﺍ ﺩﺭ ﺣﺎﻟﺘﻲ ﻗﺮﺍﺭ ﻣﻲ ﺩﻫﺪ ﻛﻪ ﻗﺒﻞ ﺍﺯ ﺧﺮﺍﺑﻲ ﺩﺭ ﺁﻥ ﺑﻮﺩﻩ ﺍﺳﺖ ‪.‬‬
‫‪ - 3‬ﻟﻐﻮ‪ ،‬ﺍﺛﺮﺍﺕ ﺗﺮﺍﻛﻨﺶ ﻧﺎﻣﻮﻓﻖ ﺭﺍ ﻟﻐﻮ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫‪48‬‬
 ‫ﻓﺼﻞ ‪:4‬‬
‫ﺍﻣﻨﻴﺖ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ‬

‫‪49‬‬
 ‫ﺗﻌﺮﻳﻒ ﺍﻳﻤﻨﻲ‪:‬‬
‫ﺍﻳﻤﻨﻲ ﻋﺒﺎﺭﺗﺴﺖ ﺍﺯ ﺣﻔﺎﻇﺖ ﺩﺍﺩﻩ ﻫﺎ ﺩﺭ ﻗﺒﺎﻝ ﺩﺳﺘﻴﺎﺑﻲ ﻏﻴﺮﻣﺠﺎﺯ‪ ،‬ﺗﻐﻴﻴﺮ ﻏﻴﺮ ﻣﺠﺎﺯ ﻳﺎ ﺗﺨﺮﻳﺐ ﺁﻧﻬﺎ ﻭ‬
‫ﻧﻴﺰ ﺩﺭ ﻗﺒﺎﻝ ﺩﺳﺘﻴﺎﺑﻲ ﺑﻪ ﺩﺍﺩﻩ ﻫﺎ ﺑﺎ ﺳﻮء ﻧﻴﺖ‪.‬‬

‫ﺗﻔﺎﻭﺕ ﺍﻳﻤﻨﻲ ﺑﺎ ﺟﺎﻣﻌﻴﺖ‪:‬‬

‫ﺩﺭ ﺍﻳﻤﻨﻲ ‪ :‬ﻣﺴﺌﻠﻪ ﺣﻔﺎﻇﺖ ﺩﺍﺩﻩ ﻫﺎ ﺩﺭ ﻣﻘﺎﺑﻞ ﻛﺎﺭﺑﺮ ﻏﻴﺮ ﻣﺠﺎﺯ ﻣﻄﺮﺡ ﺍﺳﺖ ﻭ ﺣﺼﻮﻝ ﺍﻃﻤﻴﻨﺎﻥ ﺍﺯ‬
‫ﺍﻳﻨﻜﻪ ﻛﺎﺭﺑﺮﺍﻥ ﺑﻪ ﺍﻧﭽﻪ ﻛﻪ ﺑﺪﺍﻥ ﺍﻗﺪﺍﻡ ﻣﻲ ﻭﺭﺯﻧﺪ ﻣﺠﺎﺯﻧﺪ‪.‬‬

‫ﺩﺭ ﺟﺎﻣﻌﻴﺖ ‪ :‬ﻧﻮﻋﻲ ﺣﻔﺎﻇﺖ ﺩﺍﺩﻩ ﻫﺎ ﺩﺭ ﻣﻘﺎﺑﻞ ﻋﻤﻠﻴﺎﺕ ﻛﺎﺭﺑﺮ ﻣﺠﺎﺯ ﺍﻋﻤﺎﻝ ﻣﻲ ﺷﻮﺩ ﻭ ﺣﺼﻮﻝ‬
‫ﺍﻃﻤﻴﻨﺎﻥ ﺍﺯ ﺍﻳﻨﻜﻪ ﺍﻗﺪﺍﻡ ﻛﺎﺭﺑﺮ ﻣﺠﺎﺯ‪ ،‬ﺻﺤﻴﺢ ﺍﺳﺖ ﻭ ﺻﺤﺖ ﻭ ﺩﻗﺖ ﺩﺍﺩﻩ ﻫﺎ ﺭﺍ ﺧﺪﺷﻪ ﺩﺍﺭ ﻧﻤﻲ ﻛﻨﺪ ‪.‬‬

‫ﻧﻤﻮﻧﻪﺍﻱ ﺍﺯ ﻗﻮﺍﻧﻴﻦ ﺳﺎﺯﮔﺎﺭﻱ ﻳﺎ ﺟﺎﻣﻌﻴﺖ‪:‬‬

‫‪Create table EV‬‬

‫‪S# NN‬‬
‫‪C# NN‬‬
‫‪P# NN‬‬
‫)‪Score decimal (2,2‬‬
‫‪.‬‬
‫‪.‬‬
‫‪.‬‬
‫)‪Check (score>=00.00 and score<=20.00‬‬

‫ﻗﻮﺍﻧﻴﻦ ﺟﺎﻣﻌﻴﺖ ﺩﺭ ﺍﺣﻜﺎﻡ ‪ DDL‬ﺑﻴﺎﻥ ﻣﻲﺷﻮﻧﺪ‪.‬‬

‫‪50‬‬
 ‫ﺷﺒﺎﻫﺖ ﺍﻳﻤﻨﻲ ﺑﺎ ﺟﺎﻣﻌﻴﺖ‪:‬‬
‫‪ ‬ﺳﻴﺴﺘﻢ ﻣﺪﻳﺮﻳﺖ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻫﺎ ﺍﺯ ﻭﺟﻮﺩ ﭘﺎﺭﻩ ﺍﻱ ﻗﻮﺍﻋﺪ ﻣﺤﺪﻭﺩﻳﺘﻬﺎ ﻛﻪ ﻛﺎﺭﺑﺮﺍﻥ ﻧﺒﺎﻳﺪ ﻧﻘﺾ‬
‫ﻛﻨﻨﺪ‪ ،‬ﺑﺎﻳﺪ ﺁﮔﺎﻩ ﺑﺎﺷﺪ‪.‬‬

‫‪ ‬ﻣﺤﺪﻭﺩﻳﺘﻬﺎ ﻭ ﻗﻮﺍﻋﺪ ﺑﺎﻳﺪ ﺑﻪ ﻧﺤﻮﻱ ﺗﻮﺻﻴﻒ ﺷﻮﻧﺪ ﻭ ﺩﺭ ﻛﺎﺗﺎﻟﻮگ ﻧﮕﻬﺪﺍﺭﻱ ﮔﺮﺩﻧﺪ ‪.‬‬

‫‪ ‬ﺑﺎﻳﺪ ﺭﻭﻱ ﻋﻤﻠﻴﺎﺕ ﻛﺎﺭﺑﺮ ﻧﻈﺎﺭﺕ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ ‪.‬‬

‫ﺧﻄﺮﺍﺕ‪ :Threats‬ﺧﻄﺮ ﻫﺮ ﻭﺿﻌﻴﺖ ﻭ ﺭﻭﻳﺪﺍﺩﻱ ﺍﺳﺖ ﻛﻪ ﻋﻤﺪﺍ ﻳﺎ ﺳﻬﻮﺍ ﺭﻭﻱ ﺳﻴﺴﺘﻢ ﻭ ﺩﺭ‬
‫ﻧﺘﻴﺠﻪ ﺭﻭﻱ ﺳﺎﺯﻣﺎﻥ ﺗﺎﺛﻴﺮ ﻧﺎﻣﺴﺎﻋﺪ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ‪.‬‬

‫ﺭﺩﻩ ﺑﻨﺪﻱ ﺧﻄﺮﺍﺕ‪:‬‬

‫‪ ‬ﺧﻄﺮﺍﺕ ﻓﻴﺰﻳﻜﻲ‬

‫‪ ‬ﺧﻄﺮﺍﺕ ﻣﻨﻄﻘﻲ‬

‫ﺍﻧﻮﺍﻉ ﺧﻄﺮﺍﺕ‪:‬‬

‫‪ ‬ﺧﻄﺮﺍﺕ ﻏﻴﺮﻋﻤﺪﻱ)ﺍﺗﻔﺎﻗﻲ(‬

‫‪ ‬ﺩﺭﺧﻮﺍﺳﺖ ﺳﻬﻮﻱ ﺩﺍﺩﻩ ﺍﺯ ﺳﻮﻱ ﻛﺎﺭﺑﺮ ﻏﻴﺮ ﻣﺠﺎﺯ‬

‫‪ ‬ﺑﺮﻭﺯ ﺧﺮﺍﺑﻲ ﺳﺨﺖ ﺍﻓﺰﺍﺭﻱ‬

‫‪ ‬ﻗﻄﻊ ﺑﺮﻕ ﻭ ﺍﺯ ﺑﻴﻦ ﺭﻓﺘﻦ ﺩﺍﺩﻩ ﻫﺎ ﻭ ‪....‬‬

‫‪ ‬ﺧﻄﺮﺍﺕ ﻋﻤﺪﻱ‬

‫‪ ‬ﺍﺳﺘﺮﺍﻕ ﺳﻤﻊ‬

‫‪ ‬ﺳﻮء ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﮔﺬﺭﻭﺍژﻩ ﻳﻚ ﻛﺎﺭﺑﺮ ﻣﺠﺎﺯ‬

‫‪ ‬ﺗﻬﺪﻳﺪ ﻭ ﺍﺭﻋﺎﺏ ﻛﺎﺭﺑﺮﺍﻥ ﻣﺠﺎﺯ ﻭ ‪....‬‬

‫‪51‬‬
 ‫ﺷﻲء ﺍﻳﻤﻨﻲ‪:‬‬
‫‪ ‬ﺩﺍﺩﻩ ﻫﺎ )ﺍﺯ ﺳﻄﺢ ﺻﻔﺖ ﺗﺎ ﻛﻞ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻫﺎ(‬

‫‪ ‬ﺑﺮﻧﺎﻣﻪ ﻫﺎ ﻭ ﺗﺮﺍﻛﻨﺶ ﻫﺎ‬

‫‪ ‬ﻛﺎﺗﺎﻟﻮگ ﺳﻴﺴﺘﻢ‬

‫‪ ‬ﻧﺴﺨﻪ ﻫﺎﻱ ﭘﺸﺘﻴﺒﺎﻥ‬

‫‪ ‬ﻓﺎﻳﻞ ﻫﺎﻱ ﺛﺒﺖ‬

‫‪ ‬ﺧﻮﺩ ﺳﻴﺴﺘﻢ ﻣﺪﻳﺮﻳﺖ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻫﺎ‬

‫‪ ‬ﺳﻴﺴﺘﻢ ﻋﺎﻣﻞ‬

‫‪ ‬ﺳﺨﺖ ﺍﻓﺰﺍﺭ‬

‫‪ ‬ﺷﺒﻜﻪ‬

‫‪ ‬ﻣﺤﻴﻂ ﻓﻴﺰﻳﻜﻲ‬

‫‪ ‬ﺍﻋﻀﺎ ﺗﻴﻢ ﻫﺎﻱ ﻣﺪﻳﺮﻳﺘﻲ‬

‫‪ ‬ﻛﺎﺭﺑﺮﺍﻥ ﺳﻴﺴﺘﻢ ﻫﺎ‬

‫ﺗﺪﺍﺑﻴﺮ ﺍﻳﻤﻨﻲ‪:‬‬

‫‪ ‬ﺗﺪﺍﺑﻴﺮ ﻏﻴﺮ ﻛﺎﻣﭙﻴﻮﺗﺮﻱ‬

‫‪ ‬ﻭﺿﻊ ﺳﻴﺎﺳﺖ ﻫﺎﻱ ﺍﻳﻤﻨﻲ ﻭ ﺗﻌﻴﻴﻦ ﻃﺮﺡ ﻣﻘﺎﺑﻠﻪ‬

‫‪ ‬ﻛﻨﺘﺮﻝ ﺍﻓﺮﺍﺩ‬

‫‪ ‬ﺟﺎﻳﺪﻫﻲ ﺗﺠﻬﻴﺰﺍﺕ ﺩﺭ ﻣﻜﺎﻥ ﻫﺎﻱ ﺍﻣﻦ‬

‫‪ ‬ﺗﻨﻈﻴﻢ ﺗﻮﺍﻓﻖ ﻧﺎﻣﻪ ﻫﺎ ﺍﺯ ﺳﻮﻱ ﻃﺮﻑ ﺛﺎﻟﺚ‬

‫‪52‬‬
 ‫‪ ‬ﺗﻨﻈﻴﻢ ﺗﻮﺍﻓﻖ ﻧﺎﻣﻪ ﻫﺎﻱ ﻧﮕﻬﺪﺍﺭﻱ‬

‫‪ ‬ﻛﻨﺘﺮﻝ ﺩﺳﺘﻴﺎﺑﻲ ﻓﻴﺰﻳﻜﻲ ﺍﻓﺮﺍﺩ‬

‫‪ ‬ﺗﺪﺍﺑﻴﺮ ﻛﺎﻣﭙﻴﻮﺗﺮﻱ‬

‫‪ ‬ﺷﻨﺎﺳﺎﻳﻲ ﻛﺎﺭﺑﺮ‬

‫‪ ‬ﺗﺸﺨﻴﺺ ﺍﺻﻠﻴﺖ ﻛﺎﺭﺑﺮ‬

‫‪ ‬ﻣﺠﺎﺯ ﺷﻤﺎﺭﻱ ﻭ ﻣﺪﻳﺮﻳﺖ ﺁﻥ‬

‫‪ ‬ﻛﻨﺘﺮﻝ ﺩﺳﺘﻴﺎﺑﻲ‬

‫‪ ‬ﻛﻨﺘﺮﻝ ﺩﺳﺘﻴﺎﺑﻲ ﺍﺯ ﻃﺮﻳﻖ ﻗﻒﻝ ﮔﺬﺍﺭﻱ‬

‫‪ ‬ﻧﻬﺎﻥ ﻧﮕﺎﺭﻱ‬

‫‪ ‬ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻣﻔﻬﻮﻡ ﺩﻳﺪ ﺧﺎﺭﺟﻲ‬

‫‪ ‬ﻓﺮﺍﻳﻨﺪ ﺭﺩ ﮔﻴﺮﻱ‬

‫‪ ‬ﻛﻨﺘﺮﻝ ﮔﺮﺩﺵ ﺍﻃﻼﻋﺎﺕ‬

‫‪ ‬ﻛﻨﺘﺮﻝ ﺍﺳﺘﻨﺘﺎﺝ‬

‫ﺷﻨﺎﺳﺎﻳﻲ ﻛﺎﺭﺑﺮ) ‪( User Identification‬‬

‫‪ ‬ﺑﻪ ﻫﺮ ﻛﺎﺭﺑﺮ ﻳﻚ ﺷﻨﺎﺳﻪ ﺩﺍﺩﻩ ﻣﻲ ﺷﻮﺩ ﻭ ﺍﺯ ﻃﺮﻳﻖ ﺁﻥ ﺳﻴﺴﺘﻢ‪ ،‬ﻛﺎﺭﺑﺮ ﺭﺍ ﻣﻲ ﺷﻨﺎﺳﺪ ‪.‬‬

‫‪ ‬ﻛﺎﺭﺑﺮ ﻣﻤﻜﻦ ﺍﺳﺖ‪ :‬ﻣﻨﻔﺮﺩ‪ ،‬ﮔﺮﻭﻩ ﻛﺎﺭﺑﺮﺍﻥ ﻳﺎ ﺳﻠﺴﻠﻪ ﻣﺮﺍﺗﺐ ﻛﺎﺭﺑﺮﺍﻥ ﺑﺎﺷﺪ‪.‬‬

‫ﺗﺸﺨﻴﺺ ﺍﺻﻠﻴﺖ ﻛﺎﺭﺑﺮ)‪(User Authentication‬‬

‫‪ ‬ﺁﻳﺎ ﻛﺎﺭﺑﺮ ﻫﻤﺎﻥ ﺍﺳﺖ ﻛﻪ ﺍﺩﻋﺎ ﻣﻲ ﻛﻨﺪ ﻳﺎ ﺍﺣﻴﺎﻧﺎ ﻣﺠﻌﻮﻝ ﺍﺳﺖ!!!‬

‫‪ ‬ﻣﺒﺘﻨﻲ ﺑﺮ ﻳﻚ ﻳﺎ ﭼﻨﺪ ﻣﻮﺭﺩ ﺯﻳﺮ ﺍﺳﺖ‪:‬‬

‫‪53‬‬
 ‫‪ ‬ﭼﻴﺰﻱ ﻛﻪ ﻛﺎﺭﺑﺮ ﺑﺪﺍﻧﺪ‪.‬‬

‫‪ ‬ﭼﻴﺰﻱ ﻛﻪ ﻛﺎﺭﺑﺮ ﻣﺎﻟﻚ ﺁﻥ ﺍﺳﺖ‪.‬‬

‫‪ ‬ﭼﻴﺰﻱ ﻛﻪ ﻛﺎﺭﺑﺮ ﺍﺯ ﻧﻈﺮ ﺑﻴﻮﻟﻮژﻳﻚ ﺩﺍﺭﺩ‪.‬‬

‫‪ ‬ﺍﻳﻦ ﺍﻣﻜﺎﻧﺎﺕ ﻭﺟﻮﺩ ﺩﺍﺭﻧﺪ‪:‬‬

‫‪ ‬ﮔﺬﺭ ﻭﺍژﻩ‬

‫‪ ‬ﺍﺛﺮ ﺍﻧﮕﺸﺖ‬

‫‪ ‬ﺍﻟﮕﻮﻱ ﺷﺒﻜﻪ ﭼﺸﻢ‬

‫‪ ‬ﻛﺎﺭﺕ ﺍﻟﻜﺘﺮﻭﻧﻴﻜﻲ‬

‫‪ ‬ﭘﺮﺳﺸﻨﺎﻣﻪ‬

‫‪ ‬ﻧﺸﺎﻥ‬

‫‪ ‬ﺑﻌﻀﻲ ﻭﻳﮋﮔﻲ ﻫﺎﻱ ﻓﻴﺰﻳﻮﻟﻮژﻳﻚ‬

‫ﻣﺠﺎﺯ ﺷﻤﺎﺭﻱ)‪(Authorization‬‬
‫ﻋﺒﺎﺭﺗﺴﺖ ﺍﺯ ﺍﻋﻄﺎﻱ ﻳﻚ ﺣﻖ ﻳﺎ ﺍﻣﺘﻴﺎﺯ ﻳﺎ ﻣﺠﻮﺯ ﺑﻪ ﻛﺎﺭﺑﺮ ﺑﻪ ﻧﺤﻮﻱ ﻛﻪ ﻛﺎﺭﺑﺮ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺁﻥ ﺩﺭ‬
‫ﺩﺳﺘﻴﺎﺑﻲ ﺑﻪ ﺩﺍﺩﻩ ﻫﺎﻱ ﻣﻮﺭﺩ ﻧﻈﺮﺵ ﻣﺠﺎﺯ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫ﺑﺮﺍﻱ ﺍﻗﺪﺍﻣﺎﺕ ﺯﻳﺮ ﻣﻲ ﺗﻮﺍﻥ ﺑﻪ ﻛﺎﺭﺑﺮ ﺍﻣﺘﻴﺎﺯ ﺩﺍﺩ‪:‬‬

‫‪ ‬ﺍﻳﺠﺎﺩ ﻭ ﺣﺬﻑ ﻣﻴﺪﺍﻥ‬

‫‪ ‬ﺍﻳﺠﺎﺩ ﻭ ﺣﺬﻑ ﺭﺍﺑﻄﻪ )ﺟﺪﻭﻝ( ﻣﺒﻨﺎ‬

‫‪ ‬ﺍﺿﺎﻓﻪ ﻛﺮﺩﻥ ﻳﺎ ﺣﺬﻑ ﻛﺮﺩﻥ ﺳﺘﻮﻧﻲ ﺍﺯ ﺟﺪﻭﻝ ﻣﺒﻨﺎ‬

‫‪ ‬ﺍﻳﺠﺎﺩ ﻭ ﺣﺬﻑ ﺩﻳﺪ‬

‫‪54‬‬
 ‫‪ ‬ﺍﻳﺠﺎﺩ ﻭ ﺣﺬﻑ ﻳﻚ ﻣﺤﺪﻭﺩﻳﺖ ﺟﺎﻣﻌﻴﺘﻲ‬

‫‪ ‬ﺍﻳﺠﺎﺩ ﻭ ﺣﺬﻑ ﻳﻚ ﺍﺳﺘﺮﺍﺗﮋﻱ ﺩﺳﺘﻴﺎﺑﻲ‬

‫‪ ‬ﺍﻋﻄﺎ ﻳﺎ ﺳﻠﺐ ﻳﻚ ﺣﻖ ﺩﺳﺘﻴﺎﺑﻲ ﺑﺎ ﻣﻨﻈﻮﺭ ﻣﺸﺨﺺ‬

‫‪ ‬ﺩﺭﺧﻮﺍﺱﺕ ﻳﻚ ﺟﺪﻭﻝ ﻟﺤﻈﻪ ﺍﻱ‬

‫‪ ‬ﺑﺎﺯﻳﺎﺑﻲ ﺍﺯ ﺟﺪﻭﻝ ﻣﺒﻨﺎ ﻳﺎ ﺟﺪﻭﻝ ﻣﺸﺘﻖ‬

‫‪ ‬ﺩﺭﺝ ﺩﺭ ﺟﺪﻭﻝ‬

‫‪ ‬ﺣﺬﻑ ﺍﺯ ﺟﺪﻭﻝ‬

‫‪ ‬ﺑﻬﻨﮕﺎﻡ ﺳﺎﺯﻱ ﺟﺪﻭﻝ‬

‫‪ ‬ﺗﻮﻟﻴﺪ ﺁﺭﺷﻴﻮ ﺍﺯ ﺭﻭﻱ ﺟﺪﻭﻝ‬

‫ﺩﺳﺘﻮﺭﻫﺎﻱ ‪ GRANT‬ﻭ ‪REVOKE‬‬

‫‪ Grant‬ﻭ ‪ Revoke‬ﭘﻴﺎﺩﻩﺳﺎﺯﻱ ﻣﻲ ﺷﻮﺩ‪ .‬ﺑﺮﺍﻱ ﺍﻋﻄﺎﻱ‬ ‫ﻣﺠﺎﺯﺷﻤﺎﺭﻱ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺩﺳﺘﻮﺭﺍﺕ‬
‫ﻣﺠﻮﺯﻫﺎﻱ ﻣﻮﺭﺩ ﻧﻈﺮ ﺑﻪ ﻛﺎﺭﺑﺮﺍﻥ ﺍﺯ ﺩﺳﺘﻮﺭ ‪ Grant‬ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲﺷﻮﺩ‪ .‬ﻧﻤﻮﻧﻪﺍﻱ ﺍﺯ ﻛﺎﺭﺑﺮﺩ ﺍﻳﻦ ﺩﺳﺘﻮﺭ ﺑﻪ‬
‫ﺻﻮﺭﺕ ﺯﻳﺮ ﺍﺳﺖ‪:‬‬

‫;‪Grant select (Name, Avg) , Delete on (student) to Ali‬‬

‫ﺩﺳﺘﻮﺭ ﻓﻮﻕ ﺑﻪ ﺍﻳﻦ ﻣﻌﻨﻲ ﺍﺳﺖ ﻛﻪ ﻣﺠﻮﺯ ﺑﺎﺯﻳﺎﺑﻲ ﻓﻴﻠﺪﻫﺎﻱ ﻧﺎﻡ ﻭ ﻣﻌﺪﻝ ﻭ ﻧﻴﺰ ﻣﺠﻮﺯ ﺣﺬﻑ ﺭﻛﻮﺭ ﺩ ﺍﺯ‬
‫ﺟﺪﻭﻝ ‪ student‬ﺑﻪ ﻛﺎﺭﺑﺮ ‪ Ali‬ﺍﻋﻄﺎ ﺷﻮﺩ‪.‬‬
‫ﺑﺮﺍﻱ ﺳﻠﺐ )ﻟﻐﻮ ﻳﺎ ﺑﺎﺯﭘﺲ ﮔﻴﺮﻱ( ﻣﺠﻮﺯﻫﺎﻱ ﺍﻋﻄﺎﺷﺪﻩ ﺍﺯ ﻛﺎﺭﺑﺮﺍﻥ ﺍﺯ ﺩﺳﺘﻮﺭ ‪ Revoke‬ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ‪-‬‬
‫ﺷﻮﺩ‪.‬‬

‫‪55‬‬
 :‫ﻧﻤﻮﻧﻪ ﺍﻱ ﺍﺯ ﻗﻮﺍﻋﺪ ﻳﺎ ﻣﺤﺪﻭﺩﻳﺖ ﻫﺎﻱ ﺍﻳﻤﻨﻲ‬

:‫ﻛﻨﺘﺮﻝ ﺩﺳﺘﻴﺎﺑﻲ‬
DAC (Discretionary Access Control) ‫ ﺭﻭﺵ‬

MAC (Mandatory Access Control) ‫ ﺭﻭﺵ‬

RBAC (Role Based Access Control) ‫ ﺭﻭﺵ ﻣﺒﺘﻨﻲ ﺑﺮ ﻧﻘﺶ‬

56
 ‫ﺭﻭﺵ ﺍﺧﺘﻴﺎﺭﻱ )‪(DAC‬‬

‫‪ ‬ﺍﻳﻦ ﺭﻭﺵ ﺑﺮﺍﻱ ﻛﻨﺘﺮﻝ ﺩﺳﺘﻴﺎﺑﻲ ﺍﻣﺘﻴﺎﺯﺍﺗﻲ ﺑﻪ ﻛﺎﺭﺑﺮﺍﻥ ﺍﻋﻄﺎ ﻣﻲ ﻛﻨﺪ ﻭ ﻫﺮﮔﺎﻩ ﻻﺯﻡ‬
‫ﺑﺎﺷﺪ ﺍﺯ ﺁﻧﻬﺎ ﺳﻠﺐ ﻣﻲ ﻧﻤﺎﻳﺪ‪.‬‬

‫‪ ‬ﻛﺎﺭﺑﺮﺍﻥ ﻣﺨﺘﻠﻒ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﺣﻘﻮﻕ ﻣﺘﻔﺎﻭﺕ ﺭﻭﻱ ﺷﻴﺊ ﻭﺍﺣﺪ ﺩﺍﺩﻩ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻨﺪ‪.‬‬

‫‪ ‬ﻣﺪﻳﺮﻳﺖ ﻣﺠﺎﺯ ﺷﻤﺎﺭﻱ ﻟﺰﻭﻣﺎ ﺣﺎﻟﺖ ﻣﺘﻤﺮﻛﺰ ﻧﺪﺍﺭﺩ ﻭ ﺣﺎﻻﺕ ﺯﻳﺮ ﻣﻲ ﺗﻮﺍﻧﺪ ﻭﺟﻮﺩ‬
‫ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ‪:‬‬

‫‪ .1‬ﻣﺪﻳﺮﻳﺖ ﻣﺘﻤﺮﻛﺰ‪ :‬ﻳﻚ ﻧﻔﺮ ﺍﺟﺎﺯﻩ ﺩﻫﻨﺪﻩ ﻣﺠﻮﺯﻫﺎ ﺭﺍ ﻣﻲ ﺩﻫﺪ ﻭ ﺳﻠﺐ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫‪ .2‬ﻣﺪﻳﺮﻳﺖ ﺳﻠﺴﻠﻪ ﻣﺮﺍﺗﺒﻲ ‪ :‬ﻳﻚ ﻧﻔﺮ ﻣﺴﺌﻮﻝ ﻭﺍﮔﺬﺍﺭﻱ ﻣﺴﺌﻮﻟﻴﺘﻬﺎﻱ ﻣﺪﻳﺮﻳﺘﻲ ﺑﻪ ﺳﺎﻳﺮ‬
‫ﻣﺪﻳﺮﺍﻥ ﺍﺳﺖ ‪ .‬ﺍﻳﻦ ﻣﺪﻳﺮﺍﻥ ﺧﻮﺩ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﻣﺠﻮﺯ ﻫﺎ ﺭﺍ ﺑﻪ ﻛﺎﺭﺑﺮﺍﻥ ﺑﺪﻫﻨﺪ ﻳﺎ ﺳﻠﺐ‬
‫ﻛﻨﻨﺪ‪.‬‬

‫‪ .3‬ﻣﺪﻳﺮﻳﺖ ﺟﻤﻌﻲ )ﺗﻌﺎﻭﻧﻲ(‪ :‬ﺑﺎ ﻫﻤﻴﺎﺭﻱ ﻭ ﻫﻤﻜﺎﺭﻱ ﭼﻨﺪ ﻧﻔﺮ ﺍﻋﻄﺎ ﻣﺠﻮﺯ ﺍﻧﺠﺎﻡ ﻣﻲ‬
‫ﺷﻮﺩ‪.‬‬

‫‪ .4‬ﻣﺪﻳﺮﻳﺖ ﺗﻤﻠﻜﻲ ‪ :‬ﻛﺎﺭﺑﺮ ﺍﻳﺠﺎﺩ ﻛﻨﻨﺪﻩ ﻫﺮ ﺷﻴﺊ ﻣﺎﻟﻚ ﺁﻥ ﺍﺳﺖ ﻭ ﻣﺠﻮﺯﻫﺎ ﺭﺍ ﺍﻭ ﻣﻲ‬
‫ﺩﻫﺪ‪.‬‬

‫‪ .5‬ﻣﺪﻳﺮﻳﺖ ﻧﺎﻣﺘﻤﺮﻛﺰ ‪ :‬ﻣﺎﻟﻚ ﺷﻴﺊ ﻣﻲ ﺗﻮﺍﻧﺪ ﺍﺧﺘﻴﺎﺭ ﻣﺪﻳﺮﻳﺖ ﻣﺠﺎﺯ ﺷﻤﺎﺭﻱ ﺁﻥ ﺭﺍ ﺑﻪ‬
‫ﺩﻳﮕﺮﺍﻥ ﺗﻔﻮﻳﺾ ﻛﻨﺪ‪.‬‬

‫‪ .6‬ﺗﺮﻛﻴﺐ ﺣﺎﻻﺕ ﻗﺒﻠﻲ‪.‬‬

‫ﻗﻮﺍﻋﺪ ﻛﻨﺘﺮﻝ ﺍﻳﻤﻨﻲ‪:‬‬

‫‪ ‬ﻫﺮ ﻗﺎﻋﺪﻩ ﺩﺍﺭﺍﻱ ﺍﺟﺰﺍﻱ ﺯﻳﺮ ﺍﺳﺖ‪:‬‬

‫‪ ‬ﻧﺎﻡ ﻗﺎﻋﺪﻩ‬

‫‪57‬‬
 ‫ ﻳﻚ ﻳﺎ ﺑﻴﺶ ﺍﺯ ﻳﻚ ﺍﻣﺘﻴﺎﺯ‬

‫ ﺣﻴﻄﻪ ﺍﻋﻤﺎﻝ ﻗﺎﻋﺪﻩ‬

‫ ﻳﻚ ﻳﺎ ﺑﻴﺶ ﺍﺯ ﻳﻚ ﻛﺎﺭﺑﺮ‬

‫ ﺍﻗﺪﺍﻡ ﺩﺭ ﺻﻮﺭﺕ ﻋﺪﻡ ﺭﻋﺎﻳﺖ ﻗﺎﻋﺪﻩ‬

CREATE SECURITY RULE <rule name>

GRANT <privileges list>
ON <expression>
TO <users>
ON VIOLATION <action>

:‫ﺑﺮﺍﻱ ﺣﺬﻑ ﻗﺎﻋﺪﻩ ﺍﺯ ﺩﺳﺘﻮﺭ ﺯﻳﺮ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺷﻮﺩ‬

DESTROY SECURITY RULE <rule name>

:‫ﻣﺜﺎﻝ‬
CREATE SECURITY RULE SR1
CREDIT) COTITLE, RETRIEVE (COID, GRANT
COT.CEID=D202 COT WHERE COT.DEID=D101 OR ON
U2 U1, TO
ON VIOLATION REJECT

:‫ﻣﺰﺍﻳﺎ ﻭ ﻣﻌﺎﻳﺐ‬

58
 ‫‪ ‬ﻣﺰﻳﺖ ﺭﻭﺵ ﺍﺧﺘﻴﺎﺭﻱ ﺍﻳﻦ ﺍﺳﺖ ﻛﻪ ﺑﺴﻴﺎﺭ ﺍﻧﻌﻄﺎﻑ ﭘﺬﻳﺮ ﺍﺳﺖ ‪.‬‬

‫‪ ‬ﻋﻴﺐ ﺁﻥ ﺍﻳﻦ ﺍﺳﺖ ﻛﻪ ﻛﺎﺭﺑﺮ ﻏﻴﺮﻣﺠﺎﺯ ﺍﺯﻃﺮﻳﻖ ﻛﺎﺭﺑﺮ ﻣﺠﺎﺯ ﻣﻲ ﺗﻮﺍﻧﺪ ﺍﻣﺘﻴﺎﺯﺍﺗﻲ ﺭﺍ ﻛﺴﺐ ﻛﻨﺪ ‪.‬‬
‫ﺍﺯ ﺍﻳﻦ ﻧﻈﺮ ﺭﻭﺵ ﻗﺎﺑﻞ ﺍﻃﻤﻴﻨﺎﻧﻲ ﻧﻴﺴﺖ ‪ .‬ﺑﻌﻼﻭﻩ ﻛﺎﺭﺑﺮ ﻣﻲ ﺗﻮﺍﻧﺪ ﺍﻣﺘﻴﺎﺯ ﺧﻮﺩ ﺭﺍ ﺑﻪ ﻛﺎﺭﺑﺮ ﺩﻳﮕﺮ‬
‫ﺍﻋﻄﺎ ﻛﻨﺪ‪) ،‬ﺍﻧﺘﺸﺎﺭ ﺍﻣﺘﻴﺎﺯ( ﺩﺭ ﺍﻳﻦ ﺻﻮﺭﺕ ﻋﻤﻠﻴﺎﺕ ﺳﻠﺐ ﺍﻣﺘﻴﺎﺯ ﺑﺎ ﻣﺸﻜﻼﺗﻲ ﻣﻮﺍﺟﻪ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫ﺍﻋﻄﺎ ﻭ ﺳﻠﺐ ﺍﻣﺘﻴﺎﺯ‪:‬‬

‫‪ U1‬ﺍﻣﺘﻴﺎﺯ ‪ p1‬ﺭﺍ ﺑﻪ ‪ U2‬ﻭ‪ U3‬ﻣﻲ ﺩﻫﺪ‪ U2.‬ﻭ ‪ U3‬ﻣﺘﻘﺎﺑﻼ ﺍﻳﻦ ﺍﻣﺘﻴﺎﺯ ﺭﺍ ﺑﻪ ﻫﻢ ﻣﻲ ﺩﻫﻨﺪ ‪.‬‬

‫ﺣﺎﻝ ‪ U1‬ﺍﻣﺘﻴﺎﺯ ‪ p1‬ﺭﺍ ﺍﺯ ‪ U2‬ﺳﻠﺐ ﻣﻲ ﻛﻨﺪ‪ U2 .‬ﻫﻨﻮﺯ ﺍﺯ ﻃﺮﻳﻖ ‪ U3‬ﺍﻳﻦ ﺍﻣﺘﻴﺎﺯ ﺭﺍ ﺩﺭ ﺍﺧﺘﻴﺎﺭ ﺩﺍﺭﺩ‪.‬‬

‫ﺣﺘﻲ ﺍﮔﺮ ‪ U1‬ﺍﺯ ‪ U3‬ﻧﻴﺰ ﺳﻠﺐ ﺍﻣﺘﻴﺎﺯ ﻧﻤﺎﻳﺪ‪ ،‬ﺑﺎﺯ ﻫﻢ ﺍﻣﺘﻴﺎﺯ ﺍﺯ ‪ U2‬ﻭ ‪ U3‬ﺳﻠﺐ ﻧﻤﻲ ﺷﻮﺩ‪.‬‬

‫ﺗﻜﻨﻴﻚ ﻫﺎﻱ ﻣﺤﺪﻭﺩ ﻛﺮﺩﻥ ﺍﻧﺘﺸﺎﺭ ﺍﻣﺘﻴﺎﺯ‪:‬‬

‫‪59‬‬
 ‫‪ ‬ﺍﻧﺘﺸﺎﺭ ﺍﻓﻘﻲ‬

‫‪ ‬ﺑﺎ ﻣﺸﺨﺺ ﻛﺮﺩﻥ ﻳﻚ ﻋﺪﺩ ﺻﺤﻴﺢ ‪ i‬ﺑﺮﺍﻱ ﻛﺎﺭﺑﺮ ﺩﺍﺭﺍﻱ ﻳﻚ ﺍﻣﺘﻴﺎﺯ‪ ،‬ﺑﻪ ﺍﻳﻦ ﻛﺎﺭﺑﺮ ﺍﺟﺎﺯﻩ‬
‫ﺍﻋﻄﺎﻱ ﺍﻣﺘﻴﺎﺯ ﺑﻪ ﺣﺪﺍﻛﺜﺮ ‪ i‬ﻛﺎﺭﺑﺮ ﺩﻳﮕﺮ ﺩﺍﺩﻩ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪ ‬ﺍﻧﺘﺸﺎﺭ ﻋﻤﻮﺩﻱ‬

‫‪ ‬ﺩﺭ ﻭﺍﻗﻊ ﻋﻤﻖ ﻣﺴﻴﺮ ﮔﺮﺍﻑ ﻣﺠﺎﺯ ﺷﻤﺎﺭﻱ ﻛﻨﺘﺮﻝ ﻣﻲ ﺷﻮﺩ ‪ .‬ﺑﻪ ﺍﻳﻦ ﺗﺮﺗﻴﺐ ﻛﻪ ﺍﮔﺮ ﻛﺎﺭﺑﺮ‬
‫‪ U1‬ﺩﺍﺭﺍﻱ ﻳﻚ ﺍﻣﺘﻴﺎﺯ ﺑﺎﺷﺪ‪ ،‬ﺑﺎ ﺗﻌﻴﻴﻦ ﻣﻘﺪﺍﺭ ‪ j>0‬ﺑﺮﺍﻱ ﺍﻳﻦ ﻛﺎﺭﺑﺮ ﺑﻪ ‪ U1‬ﺍﺟﺎﺯﻩ‬
‫ﺍﻋﻄﺎﻱ ﺍﻣﺘﻴﺎﺯ ﺑﻪ ‪ U2‬ﺩﺍﺩﻩ ﻣﻲ ﺷﻮﺩ‪ .‬ﺩﺭ ﺍﻳﻦ ﺻﻮﺭﺕ ‪ U1‬ﻳﻚ ﻭﺍﺣﺪ ﺍﺯ ‪ j‬ﻛﻢ ﻣﻲ ﻛﻨﺪ ﻭ‬
‫ﺩﺭ ﻧﺘﻴﺠﻪ ‪ U2‬ﻓﻘﻂ ﺑﻪ ‪ j-1‬ﻛﺎﺭﺑﺮ ﺩﻳﮕﺮ ﻣﻲ ﺗﻮﺍﻧﺪ ﺁﻥ ﺍﻣﺘﻴﺎﺯ ﺭﺍ ﺑﻄﻮﺭ ﻋﻤﻮﺩﻱ ﺍﻋﻄﺎ‬
‫ﻛﻨﺪ‪ .‬ﻫﻤﻴﻨﻄﻮﺭ ﺍﻳﻦ ﺭﻭﺍﻝ ﺗﻜﺮﺍﺭ ﻣﻲ ﺷﻮﺩ ﺗﺎ ﺯﻣﺎﻧﻴﻜﻪ ﻣﻘﺪﺍﺭ ‪ j‬ﺻﻔﺮ ﺷﻮﺩ‪.‬‬

‫ﺭﻭﺵ ﺍﺟﺒﺎﺭﻱ)‪( MAC‬‬

‫‪ ‬ﺍﻳﻦ ﺭﻭﺵ “ﻳﺎ ﻫﻤﻪ ﻳﺎ ﻫﻴﭻ ” ﺍﺳﺖ‪.‬‬

‫‪ ‬ﻳﻚ ﻛﺎﺭﺑﺮ ﻳﺎ ﺍﻣﺘﻴﺎﺯ ﺍﻧﺠﺎﻡ ﻳﻚ ﻋﻤﻞ ﺭﺍ ﺩﺍﺭﺩ ﻳﺎ ﻧﺪﺍﺭﺩ‪.‬‬

‫‪ ‬ﺍﻳﻦ ﺭﻭﺵ ﺩﺭ ﻣﺤﻴﻂ ﻫﺎﻱ ﺑﺴﻴﺎﺭ ﻧﻈﻢ ﻣﻨﺪ ﻣﺜﻞ ﻣﺤﻴﻂ ﻧﻈﺎﻣﻲ ﺍﻳﺠﺎﺩ ﺷﺪﻩ ﺍﺳﺖ ‪.‬‬

‫ﺳﻄﺢ ﻣﺤﺮﻣﺎﻧﮕﻲ ﺩﺍﺩﻩ ﻫﺎ ﻭ ﻛﺎﺭﺑﺮﺍﻥ‪:‬‬

‫‪ ‬ﺑﻪ ﻫﺮ ﺷﻲء ﺩﺍﺩﻩ ﺍﻱ ﻳﻚ ﻋﺪﺩ ﻛﻪ ﻧﺸﺎﻥ ﺩﻫﻨﺪﻩ ﺳﻄﺢ ﻣﺤﺮﻣﺎﻧﮕﻲ ﺁﻥ ﺍﺳﺖ ﻣﻨﺘﺴﺐ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫‪ ‬ﺑﻌﻼﻭﻩ ﻫﺮ ﻛﺎﺭﺑﺮ ﻧﻴﺰ ﻳﻚ ﻣﺠﻮﺯ ﺑﺮﺍﻱ ﺩﺳﺘﻴﺎﺑﻲ ﺑﻪ ﺳﻄﺢ ﻣﺸﺨﺼﻲ ﺍﺯ ﺷﻴﺊ ﺩﺍﺩﻩ ﺍﻱ ﺩﺍﺭﺩ ‪.‬‬
‫ﺑﻨﺎﺑﺮﺍﻳﻦ ﺗﻨﻬﺎ ﻛﺎﺭﺑﺮ ﺩﺍﺭﺍﻱ ﻣﺠﻮﺯ ﺩﺳﺘﻴﺎﺑﻲ ﺑﻪ ﻳﻚ ﺷﻴﺊ ﺩﺍﺩﻩ ﺍﻱ ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﻪ ﺁﻥ ﺩﺳﺘﻴﺎﺑﻲ‬
‫ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ‪.‬‬

‫‪ ‬ﺩﺍﺩﻩ ﻫﺎ ﺑﻪ ﭼﻬﺎﺭ ﺭﺩﻩ ﺯﻳﺮ ﺗﻘﺴﻴﻢ ﻣﻲ ﺷﻮﻧﺪ‪:‬‬

‫‪60‬‬
 ‫‪ ‬ﺧﻴﻠﻲ ﺳﺮﻱ )‪TS (Top Secret‬‬

‫‪ ‬ﺳﺮﻱ )‪S (Secret‬‬

‫‪ ‬ﻣﺤﺮﻣﺎﻧﻪ )‪C (Confidential‬‬

‫‪ ‬ﺭﺩﻩ ﺑﻨﺪﻱ ﻧﺸﺪﻩ )‪U (Unclassified‬‬

‫ﺭﺍﺑﻄﻪ ﺑﻴﻦ ﺁﻧﻬﺎ ﺑﻪ ﺍﻳﻦ ﺷﻜﻞ ﺍﺳﺖ‪:‬‬

‫‪TS>S>C>U‬‬

‫ﻛﺎﺭﺑﺮﺍﻥ ﻧﻴﺰ ﺑﻪ ﺭﺩﻩ ﻫﺎﻱ ﺯﻳﺮ ﺗﻘﺴﻴﻢ ﻣﻲ ﺷﻮﻧﺪ‪:‬‬

‫‪ ‬ﺭﺩﻩ ﻧﺸﺎﻥ ﺩﻫﻨﺪﻩ ﺳﻄﺢ ﻣﺤﺮﻣﺎﻧﮕﻲ ﺩﺍﺩﻩ ﺭﺍ ﺑﺎ )‪ cl(D‬ﻭ ﺭﺩﻩ ﻣﺠﻮﺯ ﻛﺎﺭﺑﺮ ﺭﺍ ﺑﺎ )‪cl(U‬‬
‫ﻧﻤﺎﻳﺶ ﻣﻲ ﺩﻫﻴﻢ‪.‬‬

‫‪ ‬ﺧﺼﻮﺻﻴﺖ ﺍﻳﻤﻨﻲ ﺳﺎﺩﻩ‪ :‬ﻛﺎﺭﺑﺮ ‪ U‬ﻧﻤﻲ ﺗﻮﺍﻧﺪ ﺩﺍﺩﻩ ‪ D‬ﺭﺍ ﺑﺨﻮﺍﻧﺪ ﻣﮕﺮ‬
‫ﺍﻳﻨﻜﻪ‪cl(U)>=cl(D) :‬‬

‫‪ ‬ﺧﺼﻮﺻﻴﺖ ﺳﺘﺎﺭﻩ ﺩﺍﺭ‪ :‬ﻛﺎﺭﺑﺮ ‪ U‬ﻧﻤﻲ ﺗﻮﺍﻧﺪ ﺩﺍﺩﻩ ‪ D‬ﺭﺍ ﺑﻨﻮﻳﺴﺪ ﻣﮕﺮ‬
‫ﺍﻳﻨﻜﻪ‪cl(U)<=cl(D) :‬‬

‫ﻣﺜﺎﻝ‪:‬‬
‫‪ ‬ﺭﺍﺑﻄﻪ ﻫﺎﻱ ‪ STT‬ﻭ ‪ STCOT‬ﺭﺍ ﺩﺭ ﻧﻈﺮ ﻣﻲ ﮔﻴﺮﻳﻢ‪.‬‬

‫‪cl(STT)=C & cl(STCOT)=TS ‬‬

‫ﺍﺳﺘﺎﺩ‪TS:‬‬ ‫ﻛﺎﺭﻣﻨﺪ ﺁﻣﻮﺯﺵ‪S:‬‬ ‫‪ ‬ﺩﺍﻧﺸﺠﻮ‪C:‬‬

‫ﺑﻪ ﺍﻳﻦ ﺗﺮﺗﻴﺐ ﺩﺳﺘﻴﺎﺑﻲ ﻛﺎﺭﺑﺮﺍﻥ ﻛﻨﺘﺮﻝ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﺭﻭﺵ ﻣﺒﺘﻨﻲ ﺑﺮ ﻧﻘﺶ)‪(RBAC‬‬

‫‪61‬‬
 ‫‪ ‬ﺩﺭ ﺍﻳﻦ ﺭﻭﺵ ﺍﺟﺎﺯﻩ ﺩﺳﺘﻴﺎﺑﻲ ﻛﺎﺭﺑﺮ ﺑﻪ ﺩﺍﺩﻩ ﻫﺎ‪ ،‬ﻭﺍﺑﺴﺘﻪ ﺑﻪ ﻧﻘﺶ ﻛﺎﺭﺑﺮ ﺩﺭ ﺳﺎﺯﻣﺎﻥ ﺍﺳﺖ ‪.‬‬

‫‪ ‬ﻛﺎﺭﺑﺮﺍﻥ ﺑﻪ ﻋﻨﻮﺍﻥ ﺍﻋﻀﺎ ﻧﻘﺸﻬﺎ ﺗﻌﺮﻳﻒ ﻣﻲ ﺷﻮﻧﺪ ﻭ ﻫﺮ ﻛﺎﺭﺑﺮ ﻣﻲ ﺗﻮﺍﻧﺪ ﻣﺠﻮﺯ ﻣﺮﺑﻮﻁ ﺑﻪ ﻧﻘﺶ‬
‫ﺧﻮﺩ ﺭﺍ ﺑﺮﺍﻱ ﺩﺳﺘﻴﺎﺑﻲ ﺑﻪ ﺩﺍﺩﻩ ﻫﺎ ﺩﺭﻳﺎﻓﺖ ﻛﻨﺪ ‪.‬‬

‫‪ ‬ﺑﺮﺍﻱ ﻭﻇﺎﻳﻒ ﻛﺎﺭﻱ ﻫﺮ ﻓﺮﺩ ﻳﺎ ﮔﺮﻭﻩ ﺩﺍﺭﺍﻱ ﻭﻇﺎﻳﻒ ﻣﺸﺎﺑﻪ ﺩﺭ ﺳﺎﺯﻣﺎﻥ‪ ،‬ﻧﻘﺶ ﺗﻌﺮﻳﻒ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫ﻛﻨﺘﺮﻝ ﺍﻳﻤﻨﻲ ﭼﻨﺪ ﺳﻄﺤﻲ ﺩﺭ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺭﺍﺑﻄﻪ ﺍﻱ‪:‬‬

‫ﻣﻲ ﺗﻮﺍﻥ ﻋﻼﻭﻩ ﺑﺮ ﺍﻋﻤﺎﻝ ﻣﺤﺪﻭﺩﻳﺖ ﺍﻳﻤﻨﻲ ﺭﻭﻱ ﺭﺍﺑﻄﻪ‪ ،‬ﻣﺤﺪﻭﺩﻳﺖ ﺭﺍ ﺩﺭ ﺳﻄﺢ ﺗﺎﭘﻞ ﻭ ﺣﺘﻲ ﺻﻔﺖ‬
‫ﻫﻢ ﺍﻋﻤﺎﻝ ﻛﺮﺩ‪.‬‬

‫ﺑﺮﺍﻱ ﺍﻋﻤﺎﻝ ﺿﺎﺑﻄﻪ ﺍﻳﻤﻨﻲ ﺩﺭ ﺳﻄﺢ ﺻﻔﺖ ﺑﻪ ﻫﺮ ﺻﻔﺖ ‪ Ai‬ﻳﻚ ﺻﻔﺖ ﺭﺩﻩ ﺑﻨﺪﻱ ‪ C‬ﻣﻦﻅّﻡ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫‪ TC‬ﺑﻪ ﺻﻔﺎﺕ ﺭﺍﺑﻄﻪ ﺍﺿﺎﻓﻪ ﻣﻲ‬ ‫ﺑﺮﺍﻱ ﺍﻋﻤﺎﻝ ﺿﺎﺑﻄﻪ ﺍﻳﻤﻨﻲ ﺩﺭ ﺳﻄﺢ ﺗﺎﭘﻞ‪ ،‬ﺻﻔﺖ ﺭﺩﻩ ﺑﻨﺪﻱ ﺗﺎﭘﻞ‬
‫ﺷﻮﺩ‪.‬‬

‫)‪R (A1,C1, A2, C2, … ,An, Cn, TC‬‬

‫ﻣﺜﺎﻝ‪:‬‬

‫ﺍﮔﺮ ﺍﻳﻦ ﺭﺍﺑﻄﻪ ﺭﺍ ﺑﺮﺍﻱ ﻛﺎﺭﺑﺮﺍﻥ ﺑﺎ ﻣﺠﻮﺯ ﺭﺩﻩ ‪ C‬ﻓﻴﻠﺘﺮ ﻛﻨﻴﻢ‪:‬‬

‫‪62‬‬
 ‫ﻭ ﺍﮔﺮ ﺑﺮﺍﻱ ﻛﺎﺭﺑﺮ ﺑﺎ ﻣﺠﻮﺯ ‪ U‬ﻓﻴﻠﺘﺮ ﻛﻨﻴﻢ‪:‬‬

‫ﻛﻨﺘﺮﻝ ﺩﺳﺘﻴﺎﺑﻲ ﺍﺯ ﻃﺮﻳﻖ ﻗﻔﻞ ﮔﺬﺍﺭﻱ‪:‬‬

‫ﻣﻲ ﺗﻮﺍﻥ ﺑﺎ ﻗﻔﻞ ﮔﺬﺍﺭﻱ ﺭﻭﻱ ﻭﺍﺣﺪﻫﺎﻱ ﺩﺍﺩﻩ ﺍﻱ ﺩﺭ ﭼﻨﺪﻳﻦ ﺳﻄﺢ‪ ،‬ﺩﺳﺘﻴﺎﺑﻲ ﺑﻪ ﺁﻧﻬﺎ ﺗﻮﺳﻂ ﻛﺎﺭﺑﺮﺍﻥ ﺭﺍ‬
‫ﻛﻨﺘﺮﻝ ﻛﺮﺩ‪.‬‬

‫ﺭﻣﺰ ﻧﮕﺎﺭﻱ)‪(Encryption‬‬
‫ﻋﺒﺎﺭﺗﺴﺖ ﺍﺯ ﻛﺪﮔﺬﺍﺭﻱ ﺩﺍﺩﻩ ﻫﺎ ﺑﻮﺳﻴﻠﻪ ﻳﻚ ﺍﻟﮕﻮﺭﻳﺘﻢ ﺧﺎﺹ ﺑﻪ ﻧﺤﻮﻱ ﻛﻪ ﻛﺎﺭﺑﺮ ﻧﺘﻮﺍﻧﺪ ﺑﺪﻭﻥ‬
‫ﺩﺍﺷﺘﻦ ﻛﻠﻴﺪ ﺭﻣﺰ ﮔﺸﺎﻳﻲ ﺩﺍﺩﻩ ﻫﺎ ﺭﺍ ﺑﺨﻮﺍﻧﺪ‪.‬‬

‫ﺍﺟﺰﺍﻱ ﺳﻴﺴﺘﻢ ﺭﻣﺰ ﻧﮕﺎﺭﻱ‪:‬‬

‫‪ ‬ﻛﻠﻴﺪ ﺭﻣﺰ ﻧﮕﺎﺭﻱ‬

‫‪ ‬ﺍﻟﮕﻮﺭﻳﺘﻢ ﺭﻣﺰ ﻧﮕﺎﺭﻱ‬

‫‪ ‬ﻛﻠﻴﺪ ﺭﻣﺰ ﮔﺸﺎﻳﻲ‬

‫‪63‬‬
 ‫‪ ‬ﺍﻟﮕﻮﺭﻳﺘﻢ ﺭﻣﺰ ﮔﺸﺎﻳﻲ‬

‫ﺍﮔﺮ ﺍﺯ ﻳﻚ ﻛﻠﻴﺪ ﺑﺮﺍﻱ ﺭﻣﺰﻧﮕﺎﺭﻱ ﻭ ﺭﻣﺰ ﮔﺸﺎﻳﻲ ﺍﺳﺘﻔﺎﺩﻩ ﺷﻮﺩ‪ ،‬ﺳﻴﺴﺘﻢ ﺭﻣﺰ ﻧﮕﺎﺭﻱ ﻣﺘﻘﺎﺭﻥ ﺍﺳﺖ ﻭ ﺩﺭ‬
‫ﻏﻴﺮ ﺍﻳﻨﺼﻮﺭﺕ ﻧﺎﻣﺘﻘﺎﺭﻥ ﺧﻮﺍﻫﺪ ﺑﻮﺩ‪.‬‬

‫*** ﺍﻟﮕﻮﺭﻳﺘﻢ ‪ RSA‬ﻧﻤﻮﻧﻪ ﺭﻣﺰﻧﮕﺎﺭﻱ ﻧﺎﻣﺘﻘﺎﺭﻥ ﺍﺳﺖ‪.‬‬

‫ﻓﺮﺁﻳﻨﺪ ﺭﺩﮔﻴﺮﻱ)‪(Audit Process‬‬

‫‪ ‬ﻋﺒﺎﺭﺗﺴﺖ ﺍﺯ ﺟﻤﻊ ﺁﻭﺭﻱ ﺩﺍﺩﻩ ﻫﺎﻳﻲ ﺩﺭ ﻣﻮﺭﺩ ﻓﻌﺎﻟﻴﺘﻬﺎ ﺩﺭ ﺳﻴﺴﺘﻢ ﻭ ﺗﺤﻠﻴﻞ ﺁﻧﻬﺎ ﺑﻪ ﻣﻨﻈﻮﺭ‬
‫ﻳﺎﻓﺘﻦ ﻣﻮﺍﺭﺩ ﻋﺪﻡ ﺭﻋﺎﻳﺖ ﺿﻮﺍﺑﻂ ﺍﻳﻤﻨﻲ ﻭ ﻳﺎ ﺗﺸﺨﻴﺺ ﻋﻠﺖ ﺑﺮﻭﺯ ﺍﻳﻦ ﻣﻮﺍﺭﺩ ‪.‬‬

‫‪ ‬ﺍﻳﻦ ﺩﺍﺩﻩ ﻫﺎ ﺩﺭ ﻓﺎﻳﻠﻲ ﺑﻪ ﻧﺎﻡ ﻓﺎﻳﻞ ﺭﺩﮔﻴﺮﻱ ﺫﺧﻴﺮﻩ ﻣﻲ ﺷﻮﻧﺪ‪.‬‬

‫‪ ‬ﺗﺤﻠﻴﻞ ﺩﺍﺩﻩ ﻫﺎ ﻣﻤﻜﻦ ﺍﺳﺖ ﺑﺼﻮﺭﺕ ﺑﺮﻭﻥ ﺧﻂ ﻭ ﺑﺎ ﺗﺎﺧﻴﺮ ﻳﺎ ﺑﺮﺧﻂ ﻭ ﺑﻴﺪﺭﻧﮓ ﺑﺎﺷﺪ ‪.‬‬

‫‪ ‬ﺩﺭ ﺣﺎﻟﺖ ﺑﻴﺪﺭﻧﮓ ﻓﺮﺍﻳﻨﺪ ﺭﺩﮔﻴﺮﻱ ﺭﺍ ﻛﺸﻒ ﻣﺰﺍﺣﻤﺖ )‪ (Intrusion detection‬ﻣﻲﮔﻮﻳﻴﻢ ‪.‬‬

‫ﻛﻨﺘﺮﻝ ﮔﺮﺩﺵ ﺍﻃﻼﻋﺎﺕ‬

‫ﺑﺎ ﻛﻨﺘﺮﻝ ﮔﺮﺩﺵ ﺍﻃﻼﻋﺎﺕ ﺍﺯ ﮔﺮﺩﺵ ﻏﻴﺮ ﻣﺠﺎﺯ ﺍﻃﻼﻋﺎﺕ ﺑﻄﻮﺭ ﺻﺮﻳﺢ )ﻣﺜﻞ ﻧﺴﺨﻪ ﺍﻱ ﺍﺯ ﺍﻃﻼﻋﺎﺕ ( ﻭ‬
‫ﺍﻱﺿﻤﻨﻲ ﺑﻴﻦ ﺳﻄﻮﺡ ﻣﺨﺘﻠﻒ ﻫﺮﻡ ﻣﺪﻳﺮﻳﺘﻲ‪ -‬ﻋﻤﻠﻴﺎﺗﻲ ﺳﺎﺯﻣﺎﻥ ﺟﻠﻮﮔﻴﺮﻱ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫ﻛﻨﺘﺮﻝ ﺍﺳﺘﻨﺘﺎﺝ)‪(Inference Control‬‬

‫‪ ‬ﺩﺭ ﺍﻳﻦ ﻛﻨﺘﺮﻝ ﺍﺯ ﻛﺴﺐ ﺍﻃﻼﻋﺎﺕ ﺍﺯ ﻃﺮﻳﻖ ﺍﺳﺘﻨﺘﺎﺝ ﺟﻠﻮﮔﻴﺮﻱ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫‪ ‬ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺑﺮﺧﻲ ﭘﺮﺳﺸﻬﺎﻱ ﺧﺎﺹ ﻭ ﺟﻬﺖ ﻳﺎﻓﺘﻪ ﻣﻲ ﺗﻮﺍﻥ ﺍﻃﻼﻋﺎﺕ ﻣﻮﺭﺩ ﻧﻈﺮ ﺭﺍ ﺍﺳﺘﻨﺘﺎﺝ‬
‫ﻛﺮﺩ‪ .‬ﺩﺭ ﺑﻌﻀﻲ ﺍﺯ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻫﺎ ﻛﺎﺭﺑﺮﺍﻥ ﻓﻘﻂ ﺍﺟﺎﺯﻩ ﺑﺎﺯﻳﺎﺑﻲ ﺩﺍﺩﻩ ﻫﺎﻱ ﺁﻣﺎﺭﻱ )ﻣﺜﻞ ﻣﻴﺎﻧﮕﻴﻦ‪،‬‬
‫ﺣﺎﺻﻞ ﺟﻤﻊ‪ ،‬ﺗﻌﺪﺍﺩ‪ ،‬ﻣﺎﻛﺰﻳﻤﻢ ﻭ ‪ (...‬ﺭﺍ ﺩﺍﺭﻧﺪ ﻭ ﺣﻖ ﺩﺳﺘﺮﺳﻲ ﻣﺴﺘﻘﻴﻢ ﺑﻪ ﺩﺍﺩﻩ ﻫﺎﻱ ﻣﻨﻔﺮﺩ‬
‫ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻫﺎ ﺭﺍ ﻧﺪﺍﺭﻧﺪ‪ .‬ﺩﺭ ﭼﻨﻴﻦ ﭘﺎﻳﮕﺎﻫﻲ ﻣﻤﻜﻦ ﺍﺳﺖ ﻣﺸﻜﻼﺗﻲ ﺑﻮﺟﻮﺩ ﺑﻴﺎﻳﺪ ‪.‬‬

‫ﻣﺜﺎﻝ‪:‬‬

‫‪64‬‬
 ‫‪ ‬ﺭﺍﺑﻄﻪ ‪ PERSON‬ﺭﺍ ﺩﺭ ﻧﻈﺮ ﺑﮕﻴﺮﻳﺪ‪:‬‬
‫)‪PERSON (NAME, SSN, INCOME, ADDRESS, CITY, STATE, ZIP, SEX, LAST_DEGRE‬‬

‫ﭘﺮﺳﺶ ﻫﺎﻱ ﺯﻳﺮ ﺭﺍ ﻧﻴﺰ ﺩﺭ ﻧﻈﺮ ﻣﻲ ﮔﻴﺮﻳﻢ‪:‬‬

‫;‪Q1: SELECT COUNT (*) FROM PERSON WHERE condition‬‬
‫;‪Q2: SELECT AVG (INCOME) FROM PERSON WHERE condition‬‬

‫ﺣﺎﻝ ﺍﮔﺮ ﺑﺨﻮﺍﻫﻴﻢ ﺩﺳﺘﻤﺰﺩ ﻛﺎﺭﻣﻨﺪ ‪ e‬ﺭﺍ ﭘﻴﺪﺍ ﻛﻨﻴﻢ ﻭ ﺑﺪﺍﻧﻴﻢ ﺍﻭ ﺯﻧﻲ ﺍ ﺳﺖ ﺑﺎ ﻣﺪﺭﻙ ﺩﻛﺘﺮﺍ ﻭ ﺩﺭ ﺷﻬﺮ ‪a‬‬
‫ﺩﺭ ﺍﻳﺎﻟﺖ ‪ s1‬ﺯﻧﺪﮔﻲ ﻣﻲ ﻛﻨﺪ‪ ،‬ﺩﺭ ‪ Q1‬ﺑﺠﺎﻱ ‪ condition‬ﺷﺮﺍﻳﻂ ﺯﻳﺮ ﺭﺍ ﻗﺮﺍﺭ ﻣﻲ ﺩﻫﻴﻢ‪:‬‬
‫‪LAST_DEGREE=phd AND SEX=f AND CITY=a AND STATE= s1‬‬

‫ﺍﮔﺮ ﻧﺘﻴﺠﻪ ﺍﻳﻦ ﭘﺮﺱ ﻭ ﺟﻮ ‪ 1‬ﺑﺎﺷﺪ ﻭ ﻫﻤﻴﻦ ﺷﺮﺍﻳﻂ ﺭﺍ ﺩﺭ ‪ Q2‬ﺍﺳﺘﻔﺎﺩﻩ ﻛﻨﻴﻢ‪ ،‬ﺩﺳﺘﻤﺰﺩ ﺍﻳﻦ ﻛﺎﺭﻣﻨﺪ‬
‫ﺑﺪﺳﺖ ﻣﻲ ﺁﻳﺪ‪.‬‬

‫ﺣﺘﻲ ﺍﮔﺮ ﻧﺘﻴﺠﻪ ﭘﺮﺱ ﻭ ﺟﻮ ‪ 1‬ﻧﺒﺎﺷﺪ ﻭ ﻋﺪﺩ ﻛﻮﭼﻜﻲ ﻣﺜﻞ ‪ 2‬ﻳﺎ ‪ 3‬ﺑﺎﺷﺪ‪ ،‬ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ‪ min‬ﻳﺎ‬
‫‪ max‬ﻣﻲ ﺗﻮﺍﻥ ﻣﺤﺪﻭﺩﻩ ﺩﺳﺘﻤﺰﺩ ﻭﻱ ﺭﺍ ﺑﺪﺳﺖ ﺁﻭﺭﺩ‪.‬‬

‫ﺭﺍﻩ ﺣﻞ‪:‬‬
‫‪ ‬ﻳﻚ ﺭﺍﻩ ﺑﺮﺍﻱ ﺭﻓﻊ ﺍﻳﻦ ﻣﺸﻜﻞ ﺁﻥ ﺍﺳﺖ ﻛﻪ ﻳﻚ ﺣﺪ ﺁﺳﺘﺎﻧﻪ ﺑﺮﺍﻱ ﺗﺎﭘﻞ ﻫﺎ ﺩﺭ ﻧﻈﺮ ﺑﮕﻴﺮﻳﻢ ﻭ ﺍﮔﺮ‬
‫ﻧﺘﻴﺠﻪ ﺍﺯ ﺍﻳﻦ ﺣﺪ ﻛﻤﺘﺮ ﺍﺳﺖ ﺟﻠﻮﻱ ﺍﻧﺠﺎﻡ ﭘﺮﺳﺶ ﺭﺍ ﺑﮕﻴﺮﻳﻢ ‪.‬‬

‫‪ ‬ﺭﺍﻩ ﺩﻳﮕﺮ ﻣﻤﺎﻧﻌﺖ ﺍﺯ ﺍﺟﺮﺍﻱ ﺩﻧﺒﺎﻟﻪ ﺍﻱ ﺍﺯ ﭘﺮﺳﺶ ﻫﺎﺳﺖ ﻛﻪ ﻣﻜﺮﺭﺍ ﺑﻪ ﺟﻤﻌﻴﺖ ﺛﺎﺑﺘﻲ ﺍﺯ ﻳﻚ‬
‫ﺭﺍﺑﻄﻪ ﺩﺳﺘﻴﺎﺑﻲ ﭘﻴﺪﺍ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫ﺑﺮﻗﺮﺍﺭﻱ ﺍﻣﻨﻴﺖ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺩﻳﺪ ﺧﺎﺭﺟﻲ‬

‫ﻣﻔﻬﻮﻡ ﺩﻳﺪ ﺧﺎﺭﺟﻲ ‪ :‬ﺩﻳﺪﮔﺎﻩ ﺩﺭ ﺣﻘﻴﻘﺖ ﻳﻚ ﻓﻴﻠﺘﺮﻳﻨﮓ ﺍﺳﺖ ﻛﻪ ﺑﻪ ﺻﻮﺭﺕ ﻳﻚ ﺟﺪﻭﻝ ﻣﺠﺎﺯﻱ ﺍﺭﺍﺋﻪ‬
‫ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﺩﻳﺪ ﺧﺎﺭﺟﻲ ﺑﻪ ﻣﻨﻈﻮﺭ ﺍﻓﺰﺍﻳﺶ ﺍﻣﻨﻴﺖ ﺩﺭ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺷﻮﺩ ﻭ ﺑﺮﺍﻱ ﻫﺮ ﻛﺎﺭﺑﺮ ﻓﻘﻂ ﻧﻴﺎﺯﻫﺎﻱ‬
‫ﺧﺎﺹ ﻭﻱ ﺩﺭ ﺩﻳﺪ ﺧﺎﺭﺟﻲ ﻧﻤﺎﻳﺶ ﺩﺍﺩﻩ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪65‬‬
 ‫ﺍﻧﻮﺍﻉ ﺩﻳﺪﮔﺎﻩ‬

‫ﺩﻳﺪﮔﺎﻩ ﺳﻄﺮﻱ ‪ :‬ﻓﻘﻂ ﺳﻄﺮﻫﺎﻱ ﺧﺎﺻﻲ ﺍﺯ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺭﺍ ﻧﻤﺎﻳﺶ ﻣﻲ ﺩﻫﺪ ﻣﺜﻞ ﺭﻛﻮﺭﺩﻫﺎﻱ ﻳﻚ ﭘﺮﻭﺍﺯ‬
‫ﺧﺎﺹ ﻭ ﻳﺎ ﺍﻃﻼﻋﺎﺕ ﺩﺍﻧﺸﺠﻮﻳﺎﻧﻲ ﻛﻪ ﻣﻌﺪﻝ ﺁﻧﻬﺎ ﺑﺎﻻﺗﺮ ﺍﺯ ‪ 17‬ﺍﺳﺖ‪.‬‬

‫‪Create view top-stud‬‬

‫* ‪As select‬‬
‫‪From Stud‬‬
‫;‪Where Avg >= 17‬‬
‫ﺩﻳﺪﮔﺎﻩ ﺳﺘﻮﻧﻲ ‪ :‬ﺳﺘﻮﻥﻫﺎ ﻳﺎ ﻓﻴﻠﺪﻫﺎﻱ ﺧﺎﺻﻲ ﺍﺯ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺭﺍ ﻧﺸﺎﻥ ﻣﻲ ﺩﻫﺪ ﻣﺜﻞ ﻓﻴﻠﺪﻫﺎﻱ ﺷﻤﺎﺭﻩ‬
‫ﺩﺍﻧﺸﺠﻮﻳﻲ ﻭ ﻧﺎﻡ ﻭ ﻣﻌﺪﻝ ﻭ ﺭﺷﺘﻪ‪.‬‬

‫‪Create view stud2‬‬

‫‪As select s#, sname, avg, major‬‬
‫;‪From Stud‬‬
‫)ﺳﻄﺮﻫﺎ(ﻱ ﺧﺎﺻﻲ ﺍﺯ ﭘﺎﻳﮕﺎﻩ‬ ‫ﺩﻳﺪﮔﺎﻩ ﺳﻄﺮﻱ‪-‬ﺳﺘﻮﻧﻲ‪ :‬ﺳﺘﻮﻥﻫﺎ ﻳﺎ ﻓﻴﻠﺪﻫﺎﻱ ﺧﺎﺻﻲ ﺍﺯ ﺭﻛﻮﺭﺩﻫﺎ‬
‫ﺩﺍﺩﻩﻫﺎ ﺭﺍ ﻧﺸﺎﻥ ﻣﻲﺩﻫﺪ‪.‬‬

‫‪Create view stud3‬‬

‫‪As select s#, sname, avg, major‬‬
‫;‪From Stud‬‬
‫;‪Where Avg >= 17‬‬
‫ﺩﻫﺪ ﻣﺜﻞ‬ ‫ﺩﻳﺪﮔﺎﻩ ﺁﻣﺎﺭﻱ ‪ :‬ﺩﺍﺩﻩﻫﺎﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺭﺍ ﺑﻪ ﺻﻮﺭﺕ ﺧﻼﺻﻪ ﻭ ﻣﺤﺎﺳﺒﺎﺗﻲ ﻧﻤﺎﻳﺶ ﻣﻲ‬
‫ﻣﻴﺎﻧﮕﻴﻦ ﺩﺍﻧﺸﺠﻮﻳﺎﻥ ﻫﺮ ﺭﺷﺘﻪ‪.‬‬

‫‪Create View AvgStud‬‬

‫)‪As select major, AVG(avg‬‬
‫‪66‬‬
From stud
Group by major;
.‫ ﻳﺎ ﺍﻟﺤﺎﻕ ﺣﺪﺍﻗﻞ ﺳﻪ ﺩﻱﺩﮔﺎﻩ ﺑﻪ ﺩﺳﺖ ﻣﻲ ﺁﻳﺪ ﻣﺜﻞ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﭼﻨﺪ ﺟﺪﻭﻝ‬Join ‫ ﺍﺯ‬:‫ﺩﻳﺪﮔﺎﻩ ﺍﻟﺤﺎﻕ‬

Create View Comp

As select stud.s#, stud.sname, stud.major, course.c#, course.cname
From stud, course
Where stud.major=course.major;

67
 ‫ﻓﺼﻞ ‪:5‬‬
‫ﻛﻨﺘﺮﻝ ﻫﻤﺰﻣﺎﻧﻲ‬

‫‪68‬‬
 ‫ﻛﻨﺘﺮﻝ ﻫﻤﺰﻣﺎﻧﻲ‬

‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﻣﺒﺘﻨﻲ ﺑﺮ ﻗﻔﻞ‬

‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﻣﺒﺘﻨﻲ ﺑﺮ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ‬
‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﻣﺒﺘﻨﻲ ﺑﺮ ﺍﻋﺘﺒﺎﺭﺳﻨﺠﻲ‬
‫‪ ‬ﺧﻮﺷﻪ ﺑﻨﺪﻱ ﭼﻨﺪﮔﺎﻧﻪ‬
‫‪ ‬ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﭼﻨﺪ ﻧﺴﺨﻪ ﺍﻱ‬
‫‪ ‬ﺩﺭﺝ ﻭ ﺣﺬﻑ ﻋﻤﻠﻴﺎﺕ‬
‫‪ ‬ﻫﻤﺰﻣﺎﻧﻲ ﺩﺭ ﺳﺎﺧﺘﺎﺭ ﻓﻬﺮﺳﺖ‬

‫ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﻣﺒﺘﻨﻲ ﺑﺮ ﻗﻔﻞ‬

‫‪ ‬ﻳﻚ ﻗﻔﻞ ﺳﺎﺯﻭﻛﺎﺭﻱ ﺑﺮﺍﻱ ﻛﻨﺘﺮﻝ ﺩﺳﺘﺮﺳﻲ ﻫﻤﺰﻣﺎﻥ ﺑﻪ ﻳﻚ ﺑﺨﺶ ﺩﺍﺩﻩ ﺍﺳﺖ‪.‬‬
‫‪ ‬ﺑﺨﺶ ﻫﺎﻱ ﺩﺍﺩﻩ ﺭﺍ ﻣﻲ ﺗﻮﺍﻥ ﺩﺭ ﺩﻭ ﺣﺎﻟﺖ ﻗﻔﻞ ﺷﺪﻩ ﻗﺮﺍﺭ ﺩﺍﺩ ‪:‬‬
‫‪ ‬ﺣﺎﻟﺖ ﺍﻧﺤﺼﺎﺭﻱ )‪: (X‬ﺑﺨﺶ ﺩﺍﺩﻩ ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﻪ ‪ 2‬ﺻﻮﺭﺕ ﺧﻮﺍﻧﺪﻥ ﻭ ﻧﻮﺷﺘﻦ ﻣﻮﺭﺩ ﺍﺳﺘﻔﺎﺩﻩ ﻗﺮﺍﺭ‬
‫ﮔﻴﺮﺩ‪ X-lock.‬ﺑﻪ ﻭﺳﻴﻠﻪ ﻱ ﺩﺳﺘﻮﺭﺍﻟﻌﻤﻞ ‪ lock-X‬ﺑﺮﺍﻱ ﺍﺳﺘﻔﺎﺩﻩ‪ ،‬ﺩﺭﺧﻮﺍﺳﺖ ﺩﺍﺩﻩ ﻣﻲ ﺷﻮﺩ‪.‬‬
‫‪ ‬ﺣﺎﻟﺖ ﻣﺸﺘﺮﻙ )‪ : (S‬ﺑﺨﺶ ﺩﺍﺩﻩ ﺩﺭ ﺍﻳﻦ ﺣﺎﻟﺖ ﻓﻘﻂ ﺧﻮﺍﻧﺪﻩ ﻣﻲ ﺷﻮﺩ ‪ S-lock .‬ﺑﻪ ﻭﺳﻲﻟﻪ ﻱ‬
‫ﺩﺳﺘﻮﺭﺍﻟﻌﻤﻞ ‪ lock-S‬ﺑﺮﺍﻱ ﺍﺳﺘﻔﺎﺩﻩ‪ ،‬ﺩﺭﺧﻮﺍﺳﺖ ﺩﺍﺩﻩ ﻣﻲ ﺷﻮﺩ‪.‬‬
‫‪ ‬ﺩﺭﺧﻮﺍﺳﺖ ﻫﺎﻱ ﻗﻔﻞ ﺑﺮﺍﻱ ﻣﺪﻳﺮﻳﺖ ﻛﻨﺘﺮﻝ ﻫﻤﺰﻣﺎﻧﻲ ﺳﺎﺧﺘﻪ ﺷﺪﻩ ﺍﻧﺪ ‪ .‬ﺗﺮﺍﻛﻨﺶ ﻓﻘﻂ ﭘﺲ ﺍﺯﺍﺧﺬ‬
‫ﻣﺠﻮﺯ ﻣﻲ ﺗﻮﺍﻧﺪ ﭘﺮﺩﺍﺯﺵ ﺷﻮﺩ‪.‬‬

‫‪ ‬ﻗﻔﻞ ﻣﺎﺗﺮﻳﺲ ﺳﺎﺯﮔﺎﺭﻱ‬

‫‪69‬‬
‫‪ ‬ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﻣﻲ ﺗﻮﺍﻧﺪ ﻣﺠﻮﺯ ﻗﻔﻞ ﺑﻪ ﻳﻚ ﺑﺨﺶ ﺭﺍ ﺍﻋﻄﺎ ﻛﻨﺪ ﺍﮔﺮ ﻗﻔﻞ ﺩﺭﺧﻮﺍﺳﺖ ﺩﺍﺩﻩ ﺷﺪﻩ ﺑﺎ ﻗﻔﻞ‬
‫ﻗﺒﻠﻲ ﺍﺧﺘﺼﺎﺹ ﺩﺍﺩﻩ ﺷﺪﻩ ﺗﻮﺳﻂ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﺩﻳﮕﺮ ﺳﺎﺯﮔﺎﺭ ﺑﺎﺷﺪ‪.‬‬
‫‪ ‬ﻫﺮ ﺗﻌﺪﺍﺩ ﺍﺯ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﻗﻔﻞ ﻫﺎﻱ ﺑﻪ ﺍﺷﺘﺮﺍﻙ ﮔﺬﺍﺷﺘﻪ ﺷﺪﻩ ﺭﻭﻱ ﻳﻚ ﺑﺨﺶ ﺭﺍ ﻧﮕﻪ‬
‫ﺩﺍﺭﻧﺪ‪.‬‬
‫• ﺍﻣﺎ ﺍﮔﺮ ﻫﺮ ﺗﺮﺍﻛﻨﺸﻲ ﺑﻪ ﺻﻮﺭﺕ ﺍﻧﺤﺼﺎﺭﻱ ﺩﺭ ﻳﻚ ﺑﺨﺶ ﻧﮕﻬﺪﺍﺭﻱ ﺑﺸﻮﺩ ﺳﺎﻳﺮ ﺗﺮﺍﻛﻨﺶ ﻫﺎ‬
‫ﻧﻤﻲ ﺗﻮﺍﻧﻨﺪ ﻫﺮ ﻗﻔﻠﻲ ﺭﺍ ﺩﺭ ﺁﻥ ﺑﺨﺶ ﻧﮕﻪ ﺩﺍﺭﻧﺪ‪.‬‬
‫‪ ‬ﺍﮔﺮ ﻳﻚ ﻗﻔﻞ ﻧﺘﻮﺍﻧﺪ ﺍﺧﺬ ﻣﺠﻮﺯ ﻛﻨﺪ‪ ،‬ﺗﺮﺍﻛﻨﺶ ﻳﻚ ﺩﺭﺧﻮﺍﺳﺖ ﺣﺎﻟﺖ ﺍﻧﺘﻈﺎﺭ ﺍﻳﺠﺎﺩ ﻣﻲ ﻛﻨﺪ ﺗﺎ ﻭﻗﺘﻲ‬
‫ﻛﻪ ﻫﻤﻪ ﻱ ﻗﻔﻞ ﻫﺎﻱ ﻧﺎﺳﺎﺯﮔﺎﺭﻧﮕﻪ ﺩﺍﺷﺘﻪ ﺷﺪﻩ ﺗﻮﺳﻂ ﺳﺎﻳﺮ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺁﺯﺍﺩ ﺷﻮﻧﺪ ﺁﻧﮕﺎﻩ ﻣﺠﻮﺯ‬
‫ﻗﻔﻞ ﺻﺎﺩﺭ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪ ‬ﻣﺜﺎﻟﻲ ﺍﺯ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺍﻧﺠﺎﻡ ﻗﻔﻞ ‪:‬‬

‫;)‪T2: lock-S(A‬‬
‫;)‪read (A‬‬
‫;)‪unlock(A‬‬
‫;)‪lock-S(B‬‬
‫;)‪read (B‬‬
‫;)‪unlock(B‬‬
‫)‪display(A+B‬‬
‫‪ ‬ﻗﻔﻞ ﮔﻒﺗﻪ ﺷﺪﻩ ﺩﺭ ﺑﺎﻻ ﺑﺮﺍﻱ ﺍﺧﺬ ﻣﺠﻮﺯ ﺳﺮﻳﺎﻟﻲ ﻛﺎﻓﻲ ﻧﻴﺴﺖ – ﺍﮔﺮ ‪ A‬ﻭ ‪ B‬ﺩﺭ ﺑﻴﻦ ﺧﻮﺍﻧﺪﻥ ﺍﺯ ‪A‬‬
‫ﺑﻪ ‪ B‬ﺑﻪ ﺭﻭﺯ ﺭﺳﺎﻧﻲ ﺷﻮﺩ ‪ ،‬ﻣﻘﺪﺍﺭ ﻧﻤﺎﻳﺶ ﺩﺍﺩﻩ ﺷﺪﻩ ﺍﺷﺘﺒﺎﻩ ﻣﻲ ﺷﻮﺩ‪.‬‬
‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﻗﻔﻞ ﻣﺠﻤﻮﻋﻪ ﻗﻮﺍﻧﻴﻨﻲ ﺍﺳﺖ ﻛﻪ ﺑﻪ ﻫﻤﻪ ﻱ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﻋﻤﻞ ﺩﺭﺧﻮﺍﺳﺖ ﻭ ﺁﺯﺍﺩ ﻛﺮﺩﻥ ﻗﻔﻞ‬
‫ﺭﺍ ﺩﺭ ﭘﻴﺮﻭﻱ ﺍﺯ ﺁﻥ ﺍﻧﺠﺎﻡ ﻣﻲ ﺩﻫﻨﺪ ﭘﺮﻭﺗﻜﻞ ﻗﻔﻞ ﻣﺠﻤﻮﻋﻪ ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﻣﻤﻜﻦ ﺭﺍ ﻣﺤﺪﻭﺩ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫ﻣﺸﻜﻼﺕ ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﻣﺒﺘﻨﻲ ﺑﺮ ﻗﻔﻞ‬

‫‪70‬‬
 ‫‪ ‬ﻗﺴﻤﺘﻲ ﺍﺯ ﺍﻳﻦ ﺑﺮﻧﺎﻣﻪ ﺭﺍ ﺩﺭ ﻧﻈﺮ ﺑﮕﻴﺮﻳﺪ‬

‫‪ ‬ﻧﻪ ‪T3‬ﻭ ﻧﻪ ‪T4‬ﻧﻤﻲ ﺗﻮﺍﻧﻨﺪ ﺍﺩﺍﻣﻪ ﭘﻴﺪﺍ ﻛﻨﻨﺪ – ﺍﺟﺮﺍﻱ )‪ lock-S(B‬ﺑﺎﻋﺚ ﻣﻲ ﺷﻮﺩ ﻛﻪ ‪ T4‬ﻣﻨﺘﻈﺮ‬
‫‪ T3‬ﺑﺮﺍﻱ ﺁﺯﺍﺩﻛﺮﺩﻥ ﻗﻔﻞ ﺁﻥ ﺭﻭﻱ ‪ B‬ﺑﺎﺷﺪ‪ ،‬ﺩﺭ ﺣﺎﻟﻲ ﻛﻪ )‪ lock-X(A‬ﺑﺎﻋﺚ ﻣﻲ ﺷﻮﺩ ﻛﻪ ‪ T3‬ﻣﻨﺘﻈﺮ‬
‫‪ T4‬ﺑﺮﺍﻱ ﺁﺯﺍﺩﻛﺮﺩﻥ ﻗﻔﻞ ﺁﻥ ﺭﻭﻱ ‪ A‬ﺑﺎﺷﺪ‪.‬‬
‫‪ ‬ﭼﻨﻴﻦ ﻭﺿﻌﻴﺘﻲ ﺑﻦ ﺑﺴﺖ ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﺩ‪.‬‬
‫• ﺑﺮﺍﻱ ﺑﺮﻃﺮﻑ ﻛﺮﺩﻥ ﺑﻦ ﺑﺴﺖ ﺍﻭﻝ‪ T3 ،‬ﻳﺎ ‪ T4‬ﺑﺎﻳﺪ ﺑﺮﮔﺮﺩﺩ ﻭ ﻗﻔﻞ ﻫﺎ ﺭﺍ ﺁﺯﺍﺩ ﻛﻨﺪ‪.‬‬
‫‪ ‬ﭘﺘﺎﻧﺴﻴﻞ ﺑﺮﺍﻱ ﺑﻦ ﺑﺴﺖ ﺩﺭ ﺍﻛﺜﺮ ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﻗﻔﻞ ﻭﺟﻮﺩ ﺩﺍﺭﺩ‪ .‬ﺑﻦ ﺑﺴﺖ ﻫﺎ ﻳﻚ ﺿﺮﻭﺭﺕ ﻣﻀﺮ‬
‫ﻫﺴﺖﻧﺪ ‪.‬‬
‫‪ ‬ﻗﺤﻄﻲ ﻧﻴﺰ ﺍﻣﻜﺎﻥ ﭘﺬﻳﺮ ﺍﺳﺖ ﺍﮔﺮ ﻣﺪﻳﺮﻳﺖ ﻛﻨﺘﺮﻝ ﻫﻤﺰﻣﺎﻧﻲ ﺑﺪ ﻃﺮﺍﺣﻲ ﺷﺪﻩ ﺑﺎﺷﺪ‪ .‬ﺑﺮﺍﻱ ﻣﺜﺎﻝ‪:‬‬
‫• ﻣﻤﻜﻦ ﺍﺳﺖ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﻣﻨﺘﻈﺮ ‪ x-lock‬ﺑﺮ ﺭﻭﻱ ﻳﻚ ﺑﺨﺶ ﺑﺎﺷﺪ‪ ،‬ﺩﺭ ﺣﺎﻟﻲ ﻛﻪ‬
‫ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﺩﻳﮕﺮ ﺩﺭﺧﻮﺍﺳﺖ ﻣﺠﻮﺯ ﺩﺭ ‪ s-lock‬ﻫﻤﺎﻥ ﺑﺨﺶ ﺭﺍ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻨﺪ‪.‬‬
‫• ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﻳﻜﺴﺎﻥ ﺑﻪ ﻃﻮﺭ ﻣﺪﺍﻭﻡ ﺑﻪ ﺩﻟﻴﻞ ﺑﻦ ﺑﺴﺖ ﻩﺏﻋﻘﺐ ﺑﺮ ﻣﻲ ﮔﺮﺩﻧﺪ ‪.‬‬
‫‪ ‬ﻣﺪﻳﺮﻳﺖ ﻛﻨﺘﺮﻝ ﻫﻤﺰﻣﺎﻧﻲ ﺑﺮﺍﻱ ﺟﻠﻮﮔﻴﺮﻱ ﺍﺯ ﻗﺤﻄﻲ ﻃﺮﺍﺣﻲ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﭘﺮﻭﺗﻜﻞ ﻗﻔﻞ ‪ 2‬ﻓﺎﺯ‬

‫‪71‬‬
 ‫‪ ‬ﺍﻳﻦ ﭘﺮﻭﺗﻜﻞ ﻣﺮﺍﻗﺐ ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﺳﺮﻳﺎﻟﻲ ﻣﺘﻀﺎﺩ ﺍﺳﺖ‪.‬‬
‫‪ ‬ﻓﺎﺯ‪ : 1‬ﻓﺎﺯ ﺩﺭ ﺣﺎﻝ ﺭﺷﺪ‬
‫• ﺗﺮﺍﻛﻨﺶ ﻣﻤﻜﻦ ﺍﺳﺖ ﻗﻔﻞ ﻫﺎ ﺭﺍ ﺑﻪ ﺩﺳﺖ ﺁﻭﺭﺩ‪.‬‬
‫• ﺗﺮﺍﻛﻨﺶ ﻣﻤﻜﻦ ﻧﻴﺴﺖ ﻗﻔﻞ ﻫﺎ ﺭﺍ ﺁﺯﺍﺩ ﻛﻨﺪ‪.‬‬
‫‪ ‬ﻓﺎﺯ ‪ : 2‬ﻛﺎﻫﺶ ﻓﺎﺯ‬
‫• ﺗﺮﺍﻛﻨﺶ ﻣﻤﻜﻦ ﺍﺳﺖ ﻗﻔﻞ ﻫﺎ ﺭﺍ ﺁﺯﺍﺩ ﻛﻨﺪ‪.‬‬
‫• ﺗﺮﺍﻛﻨﺶ ﻣﻤﻜﻦ ﻧﻴﺴﺖ ﻗﻔﻞ ﻫﺎ ﺭﺍ ﺑﻪ ﺩﺳﺖ ﺁﻭﺭﺩ‪.‬‬
‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﺍﻣﻨﻴﺖ ﺳﺮﻳﺎﻟﻲ‪،‬ﻣﻲ ﺗﻮﺍﻧﺪ ﻧﺸﺎﻥ ﺩﻫﺪ ﻛﻪ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﺑﻪ ﺻﻮﺭﺕ ﺳﺮﻳﺎﻟﻲ ﺩﺭ‬
‫ﻧﻘﺎﻁ ﻗﻔﻞ ﺧﻮﺩ ﻗﺮﺍﺭ ﮔﻴﺮﻧﺪ‪).‬ﻧﻘﻄﻪ ﺍﻱ ﻛﻪ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺩﺭ ﺁﻥ ﻗﻔﻞ ﻧﻬﺎﻳﻲ ﺭﺍ ﺑﻪ ﺩﺳﺖ ﻣﻲ ﺁﻭﺭﺩ‪(.‬‬
‫‪ ‬ﻗﻔﻞ ﺩﻭ ﻓﺎﺯ ﺭﻫﺎﻳﻲ ﺍﺯ ﺑﻦ ﺑﺴﺖ ﺭﺍ ﺗﻀﻤﻴﻦ ﻧﻤﻲ ﻛﻨﺪ‪.‬‬
‫‪ ‬ﻋﻘﺒﮕﺮﺩ ﺁﺑﺸﺎﺭﻱ ﺗﻮﺳﻂ ﻗﻔﻞ ‪ 2‬ﻓﺎﺯ ﺍﻣﻜﺎﻥ ﭘﺬﻳﺮ ﺍﺳﺖ‪ ،‬ﺑﺮﺍﻱ ﺟﻠﻮﮔﻴﺮﻱ ﺍﺯ ﺍﻳﻦ ﻋﻤﻞ‪ ،‬ﺑﻪ ﺩﻧﺒﺎﻝ ﻳﻚ‬
‫ﭘﺮﻭﺗﻜﻞ ﺍﺻﻼﺡ ﺷﺪﻩ ﺩﻗﻴﻖ ‪ 2‬ﻓﺎﺯ ﻫﺴﺘﻴﻢ‪ .‬ﺩﺭ ﺍﻳﻦ ﺣﺎﻟﺖ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺑﺎﻳﺪ ﺗﻤﺎﻡ ﻗﻔﻞ ﻫﺎﻱ‬
‫ﺍﻧﺤﺼﺎﺭﻱ ﺧﻮﺩ ﺭﺍ ﺗﺎ ﺯﻣﺎﻧﻲ ﻛﻪ ﭘﺬﻳﺮﻓﺘﻪ ﻳﺎ ﺑﻲ ﻧﺘﻴﺠﻪ ﻣﻲ ﻣﺎﻧﻨﺪ ﻧﮕﻪ ﺩﺍﺭﺩ‪.‬‬
‫‪ ‬ﻗﻔﻞ ﺩﻭ ﻓﺎﺯ ﺑﺴﻴﺎﺭ ﺩﻗﻴﻖ ﺍﺳﺖ‪ :‬ﺩﺭ ﺍﻳﻦ ﺣﺎﻟﺖ ﻫﻤﻪ ﻱ ﻗﻔﻞ ﻫﺎ ﺑﻪ ﺻﻮﺭﺕ ﺗﺄﻳﻴﺪ ﺷﺪﻩ ﻳﺎ ﺑﻲ ﻧﺘﻴﺠﻪ‬
‫ﻧﮕﻪ ﺩﺍﺷﺘﻪ ﻣﻲ ﺷﻮﻧﺪ‪ .‬ﺩﺭ ﺍﻳﻦ ﭘﺮﻭﺗﻜﻞ‪ ،‬ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺭﺍ ﻣﻲ ﺗﻮﺍﻥ ﺩﺭ ﺣﺎﻟﺘﻲ ﻛﻪ ﺍﻧﺠﺎﻡ ﻣﻲ ﺷﻮﻧﺪ ﺑﻪ‬
‫ﺻﻮﺭﺕ ﺳﺮﻳﺎﻟﻲ ﻗﺮﺍﺭ ﺩﺍﺩ‪.‬‬
‫‪ ‬ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﺳﺮﻳﺎﻟﻲ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﺩﺭ ﺗﻀﺎﺩ ﺑﺎﺷﻨﺪ ﻭ ﺍﮔﺮﺍﺯ ﻗﻔﻞ ‪ 2‬ﻓﺎﺯ ﺍﺳﺘﻔﺎﺩﻩ ﺷﻮﺩ ﻧﻤﻲ ﺗﻮﺍﻧﻨﺪ ﺑﻪ‬
‫ﺩﺳﺖ ﺑﻴﺎﻳﻨﺪ‪.‬‬
‫‪ ‬ﺑﺎ ﺍﻳﻦ ﺣﺎﻝ‪ ،‬ﺑﻪ ﺩﻟﻴﻞ ﻧﺒﻮﺩ ﺍﻃﻼﻋﺎﺕ ﺍﺿﺎﻓﻲ )ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ‪ ،‬ﺩﺳﺘﻮﺭﺩﺳﺘﺮﺳﻲ ﺑﻪ ﺩﺍﺩﻩ ﻫﺎ(‪ ،‬ﻗﻔﻞ ﺩﻭ‬
‫ﻓﺎﺯ ﺑﺮﺍﻱ ﺗﻀﺎﺩﻫﺎﻱ ﺳﺮﻳﺎﻟﻲ ﺩﺭ ﻣﻔﻬﻮﻡ ﻧﺸﺎﻥ ﺩﺍﺩﻩ ﺷﺪﻩ ﺩﺭ ﺯﻳﺮ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﺍﺳﺖ‪:‬‬
‫ﺑﺎ ﺗﻮﺟﻪ ﺑﻪ ﺍﻳﻨﻜﻪ ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﺍﺯ ﻗﻔﻞ ﺩﻭ ﻓﺎﺯ ﺍﺳﺘﻔﺎﺩﻩ ﻧﻤﻲ ﻛﻨﺪ‪ ،‬ﻣﺎ ﻣﻲ ﺗﻮﺍﻧﻴﻢ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ‪Tj‬‬
‫ﻛﻪ ﺍﺯ ﻗﻔﻞ ‪ 2‬ﻓﺎﺯ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﻛﻨﻨﺪ ﺭﺍ ﭘﻴﺪﺍ ﻛﻨﻴﻢ‪ ،‬ﻭ ﺑﺮﻧﺎﻣﻪ ﺍﻱ ﺑﺮﺍﻱ ‪ Ti‬ﻭ ‪ Tj‬ﻃﺮﺍﺣﻲ ﻛﻨﻴﻢ ﻛﻪ‬
‫ﺩﺍﺭﺍﻱ ﺗﻀﺎﺩ ﺳﺮﻳﺎﻟﻲ ﻧﺒﺎﺷﺪ‪.‬‬

‫ﺗﺒﺪﻳﻼﺕ ﻗﻔﻞ‬

‫‪ ‬ﻗﻔﻞ ﺩﻭ ﻓﺎﺯ ﺑﺎ ﺗﺒﺪﻳﻞ ﻗﻔﻞ ‪:‬‬

‫‪72‬‬
 ‫‪ -‬ﻣﺮﺣﻠﻪ ﺍﻭﻝ ‪:‬‬

‫• ﻣﻲ ﺗﻮﺍﻥ ﻳﻚ ‪ lock-s‬ﺭﺍ ﺭﻭﻱ ﺑﺨﺶ ﭘﻴﺪﺍ ﻛﺮﺩ‪.‬‬

‫• ﻣﻲ ﺗﻮﺍﻥ ﻳﻚ ‪ lock-x‬ﺭﺍ ﺭﻭﻱ ﺑﺨﺶ ﭘﻴﺪﺍ ﻛﺮﺩ‪.‬‬
‫• ﻣﻲ ﺗﻮﺍﻥ ﻳﻚ ‪ lock-s‬ﺭﺍ ﺑﻪ ﻳﻚ ‪ lock-x‬ﺗﺒﺪﻳﻞ ﻛﺮﺩ )ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ (‬

‫‪ -‬ﻣﺮﺣﻠﻪ ﺩﻭﻡ ‪:‬‬

‫• ﻣﻲ ﺗﻮﺍﻥ ﻳﻚ ‪ lock-s‬ﺭﺍ ﺁﺯﺍﺩ ﻛﺮﺩ‪.‬‬

‫• ﻣﻲ ﺗﻮﺍﻥ ﻳﻚ ‪ lock-x‬ﺭﺍ ﺁﺯﺍﺩ ﻛﺮﺩ‪.‬‬
‫• ﻣﻲ ﺗﻮﺍﻥ ﻳﻚ ‪ lock-s‬ﺭﺍ ﺑﻪ ‪ lock-x‬ﺗﺒﺪﻳﻞ ﻛﺮﺩ)ﺟﻤﻊ ﻛﺮﺩﻥ(‬
‫‪ ‬ﺍﻳﻦ ﭘﺮﻭﺗﻜﻞ ﺍﻣﻨﻴﺖ ﺳﺮﻳﺎﻟﻲ ﺍﺳﺖ‪ .‬ﺍﻣﺎ ﻫﻨﻮﺯ ﻫﻢ ﺩﺭ ﺑﺮﻧﺎﻣﻪ ﻧﻮﻳﺴﻲ ﻧﻴﺎﺯ ﺑﻪ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺩﺳﺘﻮﺭﺍﻟﻌﻤﻞ‬
‫ﻫﺎﻱ ﻗﻔﻞ ﻫﺎﻱ ﻣﺨﺘﻠﻒ ﺍﺳﺖ‪.‬‬

‫ﺩﺳﺘﻴﺎﺑﻲ ﺧﻮﺩﻛﺎﺭ ﻗﻔﻞ ﻫﺎ‬

‫‪ ‬ﺩﺭ ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﻳﻚ ﻣﻮﺿﻮﻉ‪ ،‬ﺩﺳﺘﻮﺭﺍﻟﻌﻤﻞ ﻫﺎﻱ ﺍﺳﺘﺎﻧﺪﺍﺭﺩ ﺧﻮﺍﻧﺪﻥ ‪ /‬ﻧﻮﺷﺘﻦ ﺍﺳﺖ ‪ .‬ﺑﺪﻭﻥ‬
‫ﺩﺭﺧﻮﺍﺳﺖ ﺁﺷﻜﺎﺭ ﻗﻔﻞ‬
‫‪ ‬ﻋﻤﻠﻜﺮﺩ )‪ read(D‬ﻣﺎﻧﻨﺪ ﭘﺮﺩﺍﺯﺵ ﺯﻳﺮ ﺍﺳﺖ‪:‬‬

‫‪if Ti has a lock on D‬‬

‫‪then‬‬

‫)‪read(D‬‬

‫‪else begin‬‬

‫‪if necessary wait until no other‬‬

‫‪transaction has a lock-X on D‬‬

‫;‪grant Ti a lock-S on D‬‬

‫)‪read(D‬‬

‫‪73‬‬
 ‫‪end‬‬

‫‪ Write(D) ‬ﻧﻴﺰ ﻣﺎﻧﻨﺪ ﭘﺮﺩﺍﺯﺵ ﺯﻳﺮ ﺍﺳﺖ‪:‬‬

‫‪if Ti has a lock-X on D‬‬

‫‪then‬‬

‫)‪write(D‬‬

‫‪else begin‬‬

‫‪if necessary wait until no other trans. has any lock on D,‬‬

‫‪if Ti has a lock-S on D‬‬

‫‪then‬‬

‫‪upgrade lock on D to lock-X‬‬

‫‪else‬‬

‫‪grant Ti a lock-X on D‬‬

‫)‪write(D‬‬

‫;‪end‬‬

‫‪ ‬ﺗﻤﺎﻡ ﻗﻔﻞ ﻫﺎ ﺑﻌﺪ ﺍﺯ ﭘﺬﻳﺮﻓﺘﻦ ‪ /‬ﺑﻲ ﻧﺘﻴﺠﻪ ﻣﺎﻧﺪﻥ ﺁﺯﺍﺩ ﻣﻲ ﺷﻮﻧﺪ‪.‬‬

‫ﭘﻴﺎﺩﻩ ﺳﺎﺯﻱ ﻗﻔﻞ‬

‫‪ ‬ﺍﺯ ﻣﺪﻳﺮ ﻗﻔﻞ ﻣﻲ ﺗﻮﺍﻥ ﺑﻪ ﻋﻨﻮﺍﻥ ﻳﻚ ﻓﺮﺁﻳﻨﺪ ﺟﺪﺍﮔﺎﻧﻪ ﻛﻪ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﺍﺭﺳﺎﻝ ﻗﻔﻞ ﻭ ﺑﺎﺯﻛﺮﺩﻥ‬
‫ﺩﺭﺧﻮﺍﺳﺖ ﺭﺍ ﺍﻧﺠﺎﻡ ﻣﻲ ﺩﻫﺪ ﺍﺳﺘﻔﺎﺩﻩ ﻛﺮﺩ‪.‬‬
‫‪ ‬ﻣﺪﻳﺮ ﻗﻔﻞ ﺑﻪ ﺩﺭﺧﻮﺍﺳﺖ ﻗﻔﻞ ﺑﺎ ﺍﺭﺳﺎﻝ ﭘﻴﺎﻡ ﻫﺎﻱ ﻣﺠﻮﺯ ﻗﻔﻞ)ﻳﺎ ﻳﻚ ﭘﻴﺎﻡ ﺩﺭﺧﻮﺍﺳﺖ ﺍﺯ ﺗﺮﺍﻛﻨﺶ‬
‫ﺑﺮﮔﺸﺘﻲ‪ ،‬ﺩﺭ ﺻﻮﺭﺕ ﺑﻦ ﺑﺴﺖ(ﭘﺎﺳﺦ ﻣﻲ ﺩﻫﺪ‪.‬‬
‫‪ ‬ﺩﺭﺧﻮﺍﺳﺖ ﺗﺮﺍﻛﻨﺶ ﻣﻨﺘﻈﺮ ﻣﻲ ﻣﺎﻧﺪ ﺗﺎ ﺑﻪ ﺩﺭﺧﻮﺍﺳﺘﺶ ﭘﺎﺳﺦ ﺩﺍﺩﻩ ﺷﻮﺩ‪.‬‬

‫‪74‬‬
 ‫‪ ‬ﻣﺪﻳﺮ ﻗﻔﻞ ﺳﺎﺧﺘﺎﺭ ﺩﺍﺩﻩ ﻫﺎ ﺭﺍ ﺩﺭ ﻳﻚ ﺟﺪﻭﻝ ﻛﻪ ﺟﺪﻭﻝ ﻗﻔﻞ ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﺩ ﺑﺮﺍﻱ ﺿﺒﻂ ﻣﺠﻮﺯ‬
‫ﻗﻔﻞ ﻫﺎ ﻭ ﺩﺭﺧﻮﺍﺳﺖ ﻫﺎﻱ ﺩﺭ ﺣﺎﻝ ﺍﻧﺘﻈﺎﺭ ﺫﺧﻴﺮﻩ ﻣﻲ ﻛﻨﺪ‪.‬‬
‫‪ ‬ﺟﺪﻭﻝ ﻗﻔﻞ ﺍﺳﺖ ﻣﻌﻤﻮﻻ ﺑﻪ ﻋﻨﻮﺍﻥ ﻳﻚ ﺟﺪﻭﻝ ‪ hash‬ﺑﺮﺭﻭﻱ ﻓﻬﺮﺳﺖ ﻧﺎﻡ ﺑﺨﺸﻲ ﺍﺯ ﺩﺍﺩﻩ ﻫﺎﻱ‬
‫ﻗﻔﻞ ﺷﺪﻩ ﺩﺭ ﺣﺎﻓﻈﻪ ﭘﻴﺎﺩﻩ ﺳﺎﺯﻱ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﺟﺪﻭﻝ ﻗﻔﻞ‬

‫‪ ‬ﻣﺴﺘﻄﻴﻞ ﻫﺎﻱ ﺳﻴﺎﻩ ﻗﻔﻞ ﻫﺎﻱ ﻣﺠﻮﺯ ﺩﺍﺩﻩ ﺷﺪﻩ‪ ،‬ﻭ ﺁﻧﻬﺎﻳﻲ ﻛﻪ ﺳﻔﻴﺪ ﻫﺴﺘﻨﺪ ﺩﺭﺧﻮﺍﺳﺖ ﻫﺎﻱ ﺩﺭ‬
‫ﺣﺎﻝ ﺍﻧﺘﻈﺎﺭ ﺭﺍ ﻧﺸﺎﻥ ﻣﻲ ﺩﻫﻨﺪ‪.‬‬
‫‪ ‬ﺟﺪﻭﻝ ﻗﻔﻞ ﻫﻤﭽﻨﻴﻦ ﺳﻮﺍﺑﻖ ﻧﻮﻉ ﻣﺠﻮﺯ ﻗﻔﻞ ﻭ ﻳﺎ ﺩﺭﺧﻮﺍﺳﺖ ﺭﺍ ﻧﺸﺎﻥ ﻣﻲ ﺩﻫﺪ‪.‬‬
‫‪ ‬ﺩﺭﺧﻮﺍﺳﺖ ﻫﺎﻱ ﺟﺪﻳﺪ ﺑﻪ ﺍﻧﺘﻬﺎﻱ ﺻﻒ ﺩﺭﺧﻮﺍﺳﺖ ﻫﺎﻱ ﺑﺨﺶ ﺩﺍﺩﻩ ﺍﺿﺎﻓﻪ ﻣﻲ ﺷﻮﺩ‪ ،‬ﻣﺴﻠﻤﺎً ﺑﻪ‬
‫ﻃﻮﺭ ﭘﻴﺶ ﻓﺮﺽ ﺑﺎﺗﻤﺎﻡ ﻗﻔﻞ ﻫﺎﻱ ﻗﺒﻠﻲ ﺳﺎﺯﮔﺎﺭ ﺍﺳﺖ‪.‬‬
‫‪ ‬ﻗﻔﻞ ﮔﺸﺎﻳﻲ ﻧﺘﻴﺠﻪ ﻱ ﺩﺭﺧﻮﺍﺳﺖ ﻫﺎ ﺩﺭ ﺑﻴﻦ ﺩﺭﺧﻮﺍﺳﺖ ﻫﺎﻱ ﺣﺬﻑ ﺷﺪﻩ ﺍﻧﺠﺎﻡ ﻣﻲ ﺷﻮﺩ‪،‬‬
‫ﺩﺭﺧﻮﺍﺳﺖ ﻫﺎﻱ ﺑﻌﺪﻱ ﺑﺮﺍﻱ ﺁﻧﻜﻪ ﺑﺒﻴﻨﻴﻢ ﻣﺠﻮﺯ ﺩﺍﺭﻧﺪ ﭼﻚ ﻣﻲ ﺷﻮﻧﺪ‪.‬‬
‫‪ ‬ﺍﮔﺮ ﺗﺮﺍﻛﻨﺶ ﺑﻪ ﻧﺘﻴﺠﻪ ﻧﺮﺳﺪ‪ ،‬ﻫﻤﻪ ﻱ ﺩﺭﺧﻮﺍﺳﺖ ﻫﺎﻱ ﺩﺭ ﺣﺎﻝ ﺍﻧﺘﻈﺎﺭ ﻳﺎ ﺩﺭﺧﻮﺍﺳﺖ ﻣﺠﻮﺯ ﺣﺬﻑ‬
‫ﻣﻲ ﺷﻮﻧﺪ‪.‬‬
‫• ﻣﺪﻳﺮ ﻗﻔﻞ ﻣﻤﻜﻦ ﺍﺳﺖ ﻟﻴﺴﺘﻲ ﺍﺯ ﻗﻔﻞ ﻫﺎﻱ ﻧﮕﻪ ﺩﺍﺷﺘﻪ ﺷﺪﻩ ﺗﻮﺳﻂ ﻫﺮ ﺗﺮﺍﻛﻨﺶ ﺭﺍ‬
‫ﻧﮕﻬﺪﺍﺭﻱ ﻛﻨﺪ ﻛﻪ ﺑﺮﺍﻱ ﭘﻴﺎﺩﻩ ﺳﺎﺯﻱ ﻣﺆﺛﺮ ﺍﺳﺖ‪.‬‬

‫‪75‬‬
 ‫ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﻣﺐﺗﻨﻲ ﺑﺮ ﻗﻔﻞ‬

‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﻣﺒﺘﻨﻲ ﺑﺮ ﮔﺮﺍﻑ ﻳﻚ ﺟﺎﻳﮕﺰﻳﻦ ﺑﺮﺍﻱ ﻗﻔﻞ ‪ 2‬ﻓﺎﺯﻫﺴﺘﻨﺪ‬

‫‪ ‬ﺗﺤﻤﻴﻞ ﻛﺮﺩﻥ ﻳﻚ ﺗﺮﺗﻴﺐ ﺑﺨﺸﻲ ﺩﺭ ﻣﺠﻤﻮﻋﻪ }‪ D={d1,d2,…dh‬ﺑﺮﺍﻱ ﺩﺍﺩﻩ ﻫﺎﻱ ﺗﻤﺎﻡ ﺑﺨﺶ‬
‫ﻫﺎ‬
‫• ﺍﮔﺮ ‪ di→dj‬ﺁﻧﮕﺎﻩ ﻫﺮ ﺗﺮﺍﻛﻨﺸﻲ ﺩﺳﺘﺮﺳﻲ ﭘﻴﺪﺍ ﻣﻲ ﻛﻨﺪ ﺑﻪ ﻫﺮ ﺩﻭﻱ ‪ di‬ﻭ ‪ di‬ﻛﻪ‬
‫ﺩﺳﺘﺮﺳﻲ ‪ dj‬ﺑﻴﺸﺘﺮ ﻭ ﻗﺒﻞ ﺍﺯ ‪ dj‬ﺍﺳﺖ ‪.‬‬
‫• ﺩﺭ ﻧﺘﻴﺠﻪ ﻣﺠﻤﻮﻋﻪ ﻱ ‪ D‬ﻣﻲ ﺗﻮﺍﻧﺪ ﺣﺎﻻ ﺑﻪ ﻋﻨﻮﺍﻥ ﻳﻚ ﮔﺮﺍﻑ ﺟﻬﺖ ﺩﺍﺭ‪ ،‬ﭘﺎﻳﮕﺎﻩ ﮔﺮﺍﻑ‬
‫ﻧﺎﻣﻴﺪﻩ ﺷﻮﺩ‪.‬‬
‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﺩﺭﺧﺖ ‪ ،‬ﻳﻚ ﻧﻮﻉ ﺳﺎﺩﻩ ﺍﺯ ﭘﺮﻭﺗﻜﻞ ﮔﺮﺍﻑ ﺍﺳﺖ ‪.‬‬

‫ﭘﺮﻭﺗﻜﻞ ﺩﺭﺧﺖ‬

‫‪76‬‬
 ‫‪ .1‬ﺗﻨﻬﺎ ﻗﻔﻞ ﻫﺎﻱ ﺍﻧﺤﺼﺎﺭﻱ ﺍﻣﻜﺎﻥ ﭘﺬﻳﺮ ﺍﺳﺖ‪.‬‬

‫‪ .2‬ﺍﻭﻟﻴﻦ ﻗﻔﻞ ﻣﻤﻜﻦ ﺍﺳﺖ ﺗﻮﺳﻂ ‪ Ti‬ﺑﻪ ﺗﻤﺎﻡ ﺑﺨﺶ ﻫﺎﻱ ﺩﺍﺩﻩ ﺍﺧﺘﺼﺎﺹ ﺩﺍﺩﻩ ﺷﻮﺩ‪ ،‬ﭘﺲ ﺍﺯ ﺁﻥ‪ ،‬ﻳﻚ ﺩﺍﺩﻩ ‪Q‬‬
‫ﻣﻲ ﺗﻮﺍﻧﺪ ﺗﻮﺳﻂ ‪ Ti‬ﻗﻔﻞ ﺷﻮﺩ ﻓﻘﻂ ﺩ ﺭﺻﻮﺭﺗﻲ ﻛﻪ ﻭﺍﻟﺪ ‪ Q‬ﺑﻪ ﺩﺭﺳﺘﻲ ﺗﻮﺳﻂ ‪ Ti‬ﻗﻔﻞ ﺷﺪﻩ ﺑﺎﺷﺪ‪.‬‬

‫‪ .3‬ﺑﺨﺶ ﻫﺎﻱ ﺩﺍﺩﻩ ﻣﻤﻜﻦ ﺍﺳﺖ ﺩﺭ ﻫﺮ ﺯﻣﺎﻧﻲ ﻗﻔﻞ ﮔﺸﺎﻳﻲ ﺷﻮﻧﺪ‪.‬‬

‫‪ .4‬ﻳﻚ ﺑﺨﺶ ﺩﺍﺩﻩ ﻛﻪ ﺗﻮﺳﻂ ‪ Ti‬ﻗﻔﻞ ﮔﺬﺍﺭﻱ ﻭ ﻗﻔﻞ ﮔﺸﺎﻳﻲ ﺷﺪﻩ ﺍﺳﺖ ﻧﻤﻲ ﺗﻮﺍﻧﺪ ﻣﺘﻌﺎﻗﺒﺎً ﺗﻮﺳﻂ ‪ Ti‬ﻣﺠﺪﺩ‬
‫ﻗﻔﻞ ﮔﺬﺍﺭﻱ ﺷﻮﺩ‪.‬‬

‫ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﻣﺒﺘﻨﻲ ﺑﺮ ﮔﺮﺍﻑ‬

‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﺩﺭﺧﺖ ﺭﻫﺎﻳﻲ ﺍﺯ ﺗﻀﺎﺩﻫﺎﻱ ﺳﺮﻳﺎﻟﻲ ﻭ ﻫﻤﭽﻨﻴﻦ ﺑﻦ ﺑﺴﺖ ﺭﺍ ﺗﻀﻤﻴﻦ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫‪ ‬ﻗﻔﻞ ﮔﺸﺎﻳﻲ ﺩﺭ ﭘﺮﻭﺗﻜﻞ ﺩﺭﺧﺖ ﻣﻲ ﺗﻮﺍﻧﺪ ﺳﺎﺩﻩ ﺗﺮ ﺍﺯ ﭘﺮﻭﺗﻜﻞ ﻓﺎﺯ ‪ 2‬ﺍﻧﺠﺎﻡ ﺷﻮﺩ‪.‬‬
‫• ﻛﻮﺗﺎﻫﺘﺮ ﺷﺪﻥ ﺯﻣﺎﻥ ﺍﻧﺘﻈﺎﺭ ﻭ ﺍﻓﺰﺍﻳﺶ ﻫﻤﺰﻣﺎﻧﻲ‬
‫• ﺩﺭ ﭘﺮﻭﺗﻜﻞ ﺑﻦ ﺑﺴﺖ‪-‬ﺁﺯﺍﺩ ‪ ،‬ﺑﻪ ﻫﻴﭻ ﻋﻘﺐ ﮔﺮﺩﻱ ﻧﻴﺎﺯ ﻧﻴﺴﺖ‪.‬‬
‫‪ ‬ﺍﺷﻜﺎﻻﺕ‬
‫• ﭘﺮﻭﺗﻜﻞ‪ ،‬ﺑﺎﺯﻳﺎﺑﻲ ﻳﺎ ﺁﺯﺍﺩﻱ ﺁﺑﺸﺎﺭﻱ ﺭﺍ ﺗﻀﻤﻴﻦ ﻧﻤﻲ ﻛﻨﺪ‪.‬‬
‫‪ ‬ﻧﻴﺎﺯ ﺑﻪ ﺗﻌﺮﻳﻒ ﻭﺍﺑﺴﺘﮕﻲ ﻫﺎﻱ ﻣﻮﺭﺩ ﭘﺬﻳﺮﺵ ﺑﺮﺍﻱ ﺗﻀﻤﻴﻦ ﺑﺎﺯﻳﺎﺑﻲ‬

‫‪77‬‬
 ‫• ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﻣﻤﻜﻦ ﺍﺳﺖ ﻗﻔﻞ ﺑﺨﺶ ﻫﺎﻱ ﺩﺍﺩﻩ ﺍﻱ ﺭﺍﺩﺍﺷﺘﻪ ﺑﺎﺷﻨﺪ ﻛﻪ ﺑﻪ ﺁﻧﻬﺎ ﺩﺳﺘﺮﺳﻲ‬
‫ﻧﺪﺍﺭﻧﺪ‪.‬‬
‫‪ ‬ﺍﻓﺰﺍﻳﺶ ﺳﺮﺭﻳﺰ ﻗﻔﻞ‪ ،‬ﻭ ﺯﻣﺎﻥ ﺍﻧﺘﻈﺎﺭ ﺍﺿﺎﻓﻲ‬
‫‪ ‬ﭘﺘﺎﻧﺴﻴﻞ ﻛﺎﻫﺶ ﻫﻤﺰﻣﺎﻧﻲ‬
‫‪ ‬ﺑﺮﻧﺎﻣﻪ ﻫﺎ ﺗﺤﺖ ﻗﻔﻞ ﺩﻭ ﻓﺎﺯﺍﻣﻜﺎﻥ ﭘﺬﻳﺮ ﻧﻴﺴﺘﻨﺪ ﻭ ﺑﺎﻟﻌﻜﺲ ﺗﺤﺖ ﭘﺮﻭﺗﻜﻞ ﺩﺭﺧﺖ ﺍﻣﻜﺎﻥ ﭘﺬﻳﺮ‬
‫ﻫﺴﺘﻨﺪ‪.‬‬

‫ﺧﻮﺷﻪ ﺑﻨﺪﻱ ﻫﺎﻱ ﻣﺘﻌﺪﺩ‬

‫‪ ‬ﻣﺠﻮﺯ ﺑﺨﺶ ﻫﺎﻱ ﺩﺍﺩﻩ ﺑﻪ ﺍﻧﺪﺍﺯﻩ ﻫﺎﻱ ﻣﺨﺘﻠﻒ ﻭ ﺗﻌﺮﻳﻒ ﺳﻠﺴﻠﻪ ﻣﺮﺍﺗﺐ ﺑﺨﺶ ﺑﻪ ﺑﺨﺶ ﺁﻧﻬﺎﺳﺖ‪،‬‬
‫ﻗﺴﻤﺘﻬﺎﻱ ﺑﺨﺶ ﺑﻪ ﺑﺨﺶ ﻛﻮﭼﻜﺘﺮ ﺩﺭ ﺩﺭﻭﻥ ﺑﺰﺭﮔﺘﺮﻫﺎ ﻗﺮﺍﺭ ﻣﻲ ﮔﻴﺮﻧﺪ‪.‬‬
‫‪ ‬ﮔﺮﺍﻓﻴﻚ ﻫﺎ ﺭﺍ ﻣﻲ ﺗﻮﺍﻥ ﺑﻪ ﻋﻨﻮﺍﻥ ﻳﻚ ﺩﺭﺧﺖ ﻧﺸﺎﻥ ﺩﺍﺩ )ﺍﻣﺎ ﺑﺎ ﭘﺮﻭﺗﻜﻞ ﻗﻔﻞ ﺩﺭﺧﺖ ﺩﺭﺳﺖ ﻧﻴﺴﺖ(‬
‫‪ ‬ﻭﻗﺘﻲ ﻛﻪ ﺗﺮﺍﻛﻨﺶ ﻗﻔﻞ ﻫﺎﻱ ﻳﻚ ﮔﺮﻩ ﺩﺭ ﺩﺭﺧﺖ ﺑﻪ ﺩﺭﺳﺘﻲ ﻗﺮﺍﺭ ﺩﺍﺭﺩ‪ ،‬ﻣﻄﻤﺌﻨﺎً ﻫﻤﻪ ﻱ ﺑﺨﺶ ﻫﺎﻱ‬
‫ﮔﺮﻩ ﺩﺭ ﻫﻤﺎﻥ ﺣﺎﻟﺖ ﻗﺮﺍﺭ ﻣﻲ ﮔﻴﺮﻧﺪ‪.‬‬
‫‪ ‬ﺧﻮﺷﻪ ﺑﻨﺪﻱ ﻗﻔﻞ ‪) :‬ﺳﻄﺤﻲ ﺩﺭ ﺩﺭﺧﺖ ﻛﻪ ﺩﺭ ﺁﻥ ﻗﻔﻞ ﺍﻧﺠﺎﻡ ﺷﺪﻩ ﺍﺳﺖ(‬
‫• ﺧﻮﺷﻪ ﺑﻨﺪﻱ ﺧﻮﺏ )ﭘﺎﻳﻴﻦ ﺗﺮ ﺩﺭ ﺩﺭﺧﺖ( ‪ :‬ﻫﻤﺰﻣﺎﻧﻲ ﺑﺎﻻ‪ ،‬ﺳﺮﺑﺎﺭ ﻗﻔﻞ ﺑﺎﻻ‬
‫• ﺧﻮﺷﻪ ﺑﻨﺪﻱ ﺩﺭﺷﺖ )ﺩﺭ ﺩﺭﺧﺖ ﺑﺎﻻﺗﺮ( ‪ :‬ﻗﻔﻞ ﺳﺮﺑﺎﺭ ﻛﻢ ‪ ،‬ﻫﻤﺰﻣﺎﻧﻲ ﻛﻢ‬

‫ﻣﺜﺎﻟﻲ ﺍﺯ ﺳﻠﺴﻠﻪ ﻣﺮﺍﺗﺐ ﺧﻮﺷﻪ ﺑﻨﺪﻱ‬

‫‪78‬‬
 ‫ﺳﻄﻮﺡ ﺍﺯ ﺑﺎﻻﺗﺮﻳﻦ ﺳﻄﺢ ﺷﺮﻭﻉ ﻣﻲ ﺷﻮﺩ ﻛﻪ ﻋﺒﺎﺭﺗﻨﺪ ﺍﺯ ‪:‬‬

‫• ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ‬
‫• ﻧﺎﺣﻴﻪ‬
‫• ﭘﻮﺷﻪ‬
‫• ﺳﻮﺍﺑﻖ‬

‫ﺣﺎﻻﺕ ﻗﻔﻞ ﻣﻘﺼﺪ‬

‫‪ ‬ﻋﻼﻭﻩ ﺑﺮ ﺣﺎﻟﺖ ﻫﺎﻱ ﻗﻔﻞ ‪ S‬ﻭ ‪ ،X‬ﺳﻪ ﺣﺎﻟﺖ ﻗﻔﻞ ﻫﺎﻱ ﺍﺿﺎﻓﻲ ﺑﺎ ﺧﻮﺷﻪ ﺑﻨﺪﻱ ﻣﺘﻌﺪﺩ ﻭﺟﻮﺩ ﺩﺍﺭﺩ‪:‬‬
‫• ﻣﻔﻬﻮﻡ ﺍﺷﺘﺮﺍﻙ)‪ : (IS‬ﻗﻔﻞ ﺻﺤﻴﺢ ﺑﺨﺶ ﭘﺎﻳﻴﻨﻲ ﺩﺭﺧﺖ ﺭﺍ ﻧﺸﺎﻥ ﻣﻲ ﺩﻫﺪ ﺍﻣﺎ ﻓﻘﻂ ﺑﺮﺍﻱ‬
‫ﻗﻔﻞ ﻫﺎﻱ ﺑﻪ ﺍﺷﺘﺮﺍﻙ ﮔﺬﺍﺷﺘﻪ ﺷﺪﻩ ‪.‬‬
‫• ﻣﻔﻬﻮﻡ ﺍﻧﺤﺼﺎﺭﻱ )‪ : (IX‬ﻗﻔﻞ ﺻﺤﻴﺢ ﺑﺨﺶ ﭘﺎﻳﻴﻨﻲ ﺩﺭﺧﺖ ﺑﺎ ﻗﻔﻞ ﻫﺎﻱ ﺍﻧﺤﺼﺎﺭﻱ ﻳﺎ‬
‫ﺍﺷﺘﺮﺍﻛﻲ ﺭﺍ ﻧﺸﺎﻥ ﻣﻲ ﺩﻫﺪ‪.‬‬
‫• ﻣﻔﻬﻮﻡ ﺍﻧﺤﺼﺎﺭﻱ ﻭ ﺍﺷﺘﺮﺍﻛﻲ )‪ : (SIX‬ﺭﻳﺸﻪ ﻫﺎﻱ ﺯﻳﺮﻳﻦ ﺩﺭﺧﺖ ﺗﻮﺳﻂ ﺑﺨﺶ ﻫﺎﻳﻲ ﻛﻪ ﺑﻪ‬
‫ﺩﺭﺳﺘﻲ ﻗﻔﻞ ﺷﺪﻩ ﺍﻧﺪ ﺩﺭ ﺣﺎﻟﺖ ﺍﺷﺘﺮﺍﻛﻲ ﻭ ﻗﻔﻞ ﻫﺎﻱ ﺩﺭﺳﺖ ﺍﻣﺎ ﺩﺭ ﻳﻚ ﺳﻄﺢ ﭘﺎﻳﻴﻦ ﺗﺮ‬
‫ﺩﺭ ﺣﺎﻟﺖ ﻗﻔﻞ ﻫﺎﻱ ﺍﻧﺤﺼﺎﺭﻱ ﺍﻧﺠﺎﻡ ﻣﻲ ﺷﻮﻧﺪ‪.‬‬
‫‪ ‬ﻣﻔﻬﻮﻡ ﻣﺠﻮﺯ ﻗﻔﻞ ﻫﺎ ﺩﺭ ﮔﺮﻩ ﻫﺎﻱ ﺳﻄﺢ ﺑﺎﻻﺗﺮ ﻳﻌﻨﻲ ﺩﺭ ﺣﺎﻟﺖ ‪ S‬ﻳﺎ ‪ X‬ﺑﺪﻭﻥ ﻧﻴﺎﺯ ﺑﻪ ﺑﺮﺭﺳﻲ ﻫﻤﻪ‬
‫ﻱ ﮔﺮﻩ ﻫﺎﻱ ﺑﻌﺪﻱ ﻗﻔﻞ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﻣﺎﺗﺮﻳﺲ ﺳﺎﺯﮔﺎﺭﻱ ﺑﺎ ﺣﺎﻻﺕ ﻗﻔﻞ ﻣﻘﺼﺪ‬

‫‪ ‬ﻣﺎﺗﺮﻳﺲ ﺳﺎﺯﮔﺎﺭﻱ ﺑﺮﺍﻱ ﺗﻤﺎﻡ ﺣﺎﻻﺕ ﻗﻔﻞ ﻋﺒﺎﺭﺗﺴﺖ ﺍﺯ ‪:‬‬

‫‪IS‬‬ ‫‪IX‬‬ ‫‪S‬‬ ‫‪SIX‬‬ ‫‪X‬‬

‫‪IS‬‬ ‫√‬ ‫√‬ ‫√‬ ‫√‬ ‫×‬

‫‪IX‬‬ ‫√‬ ‫√‬ ‫×‬ ‫×‬ ‫×‬

‫‪S‬‬ ‫√‬ ‫×‬ ‫√‬ ‫×‬ ‫×‬

‫‪79‬‬
 ‫‪SIX‬‬ ‫√‬ ‫×‬ ‫×‬ ‫×‬ ‫×‬

‫‪X‬‬ ‫×‬ ‫×‬ ‫×‬ ‫×‬ ‫×‬

‫ﻃﺮﺡ ﺧﻮﺷﻪ ﺑﻨﺪﻱ ﻣﺘﻌﺪﺩ ﻗﻔﻞ‬

‫‪ ‬ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﻣﻲ ﺗﻮﺍﻥ ﻳﻚ ﮔﺮﻩ ‪ Q‬ﺭﺍ ﻗﻔﻞ ﻛﻨﺪ ‪ ،‬ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻗﻮﺍﻧﻴﻦ ﺯﻳﺮ‪:‬‬

‫‪ .1‬ﻣﺎﺗﺮﻳﺲ ﺳﺎﺯﮔﺎﺭﻱ ﻗﻔﻞ ﻫﺎ ﺑﺎﻳﺪ ﺭﻋﺎﻳﺖ ﺷﻮﺩ‪.‬‬

‫‪ .2‬ﺭﻳﺸﻪ ﺩﺭﺧﺖ ﺑﺎﻳﺪ ﺩﺭ ﺍﺑﺘﺪﺍ ﻗﻔﻞ ﺷﻮﺩ ‪ ،‬ﻣﻤﻜﻦ ﺍﺳﺖ ﺩﺭ ﻫﺮ ﺣﺎﻟﺖ ﻗﻔﻞ ﺷﻮﺩ‪.‬‬

‫‪ .3‬ﮔﺮﻩ ‪ Q‬ﻣﻲ ﺗﻮﺍﻧﺪ ﺗﻮﺳﻂ ‪ Ti‬ﺩﺭﺣﺎﻟﺖ ‪ s‬ﻳﺎ ‪ Is‬ﻗﻔﻞ ﺷﻮﺩ‪ ،‬ﻓﻘﻂ ﺩﺭ ﺻﻮﺭﺗﻲ ﻛﻪ ﻭﺍﻟﺪ ‪ Q‬ﺑﻪ ﺩﺭﺳﺘﻲ‬
‫ﺗﻮﺳﻂ ‪ Ti‬ﺩﺭ ﻫﺮ ﻳﻚ ﺍﺯ ﺣﺎﻻﺕ ‪ IX‬ﻳﺎ ‪ IS‬ﻗﻔﻞ ﺷﺪﻩ ﺑﺎﺷﺪ‪.‬‬

‫‪ .4‬ﻳﻚ ﮔﺮﻩ ‪ Q‬ﻣﻲ ﺗﻮﺍﻧﺪ ﺗﻮﺳﻂ ‪ Ti‬ﺩﺭ ﺣﺎﻟﺖ ‪ SIX ،x‬ﻳﺎ ‪ IX‬ﻗﻔﻞ ﺷﻮﺩ ﻓﻘﻂ ﺍﮔﺮ ﻭﺍﻟﺪ ‪ Q‬ﺑﻪ ﺩﺭﺳﺘﻲ‬
‫ﺗﻮﺳﻂ ‪ Ti‬ﺩﺭ ﺣﺎﻟﺖ ﻫﺎﻱ ‪ IX‬ﻳﺎ ‪ SIX‬ﻗﻔﻞ ﺷﺪﻩ ﺑﺎﺷﺪ‪.‬‬

‫ﻝ ﺍﺯ ﺁﻥ ﻫﻴﭻ ﮔﺮﻩ ﺍﻱ ﺭﺍ ﺑﺎﺯ ﻧﻜﺮﺩﻩ ﺑﺎﺷﺪ)ﻛﻪ‬

‫‪ .5‬ﮔﺮﻩ ‪ Ti‬ﻣﻲ ﺗﻮﺍﻧﺪ ﻗﻔﻞ ﺷﻮﺩ ﻓﻘﻂ ﺩﺭ ﺻﻮﺭﺗﻲ ﻛﻪ ﻗﺐ‬
‫ﻳﻌﻨﻲ ‪ 2 Ti‬ﻓﺎﺯ ﺍﺳﺖ(‪.‬‬

‫‪ TI .6‬ﻣﻲ ﺗﻮﺍﻧﺪ ﮔﺮﻩ ‪ Q‬ﺭﺍ ﺑﺎﺯ ﻛﻨﺪ ﺗﻨﻬﺎ ﺩﺭ ﺻﻮﺗﻲ ﻛﻪ ﻫﻴﭻ ﻳﻚ ﺍﺯ ﻓﺮﺯﻧﺪﺍﻥ ‪ Q‬ﺩﺭ ﺣﺎﻝ ﺣﺎﺿﺮ ﺗﻮﺳﻂ‬
‫‪ TI‬ﻗﻔﻞ ﻧﺸﺪﻩ ﺑﺎﺷﻨﺪ‪.‬‬

‫‪ ‬ﻫﻤﺎﻥ ﻃﻮﺭ ﻛﻪ ﻣﺸﺎﻫﺪﻩ ﻣﻲ ﻛﻨﻴﺪ ﻗﻔﻞ ﻫﺎ ﺍﺯ ﺭﻳﺸﻪ ﺑﻪ ﺑﺮگ ﺑﻪ ﺩﺳﺖ ﻣﻲ ﺁﻳﻨﺪ‪ ،‬ﺩﺭ ﺣﺎﻟﻲ ﻛﻪ ﺁﻧﻬﺎ ﺍﺯ‬
‫ﺑﺮگ ﺑﻪ ﺭﻳﺸﻪ ﺁﺯﺍﺩ ﻣﻲ ﺷﻮﻧﺪ‪.‬‬

‫ﺟﺎﺑﻪ ﺟﺎﻳﻲ ﺑﻦ ﺑﺴﺖ‬

‫‪ ‬ﺩﻭ ﺗﺮﺍﻛﻨﺶ ﺯﻳﺮ ﺭﺍ ﺩﺭ ﻧﻈﺮ ﺑﮕﻴﺮﻳﺪ ‪:‬‬

‫‪T1:‬‬ ‫)‪write (X‬‬ ‫‪T2: write(Y) ‬‬

‫)‪write(Y‬‬ ‫)‪write(X‬‬

‫‪80‬‬
 ‫‪ ‬ﺑﺮﻧﺎﻣﻪ ﺑﺎ ﺑﻦ ﺑﺴﺖ‬

‫‪T1‬‬ ‫‪T2‬‬

‫‪lock-X on X‬‬
‫)‪write (X‬‬
‫‪lock-X on Y‬‬
‫)‪write (Y‬‬
‫‪wait for lock-X on X‬‬

‫‪wait for lock-X on Y‬‬

‫‪ ‬ﺳﻴﺴﺘﻢ ﺑﻪ ﺑﻦ ﺑﺴﺖ ﻣﻲ ﺭﺳﺪ ﺍﮔﺮ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﻃﻮﺭﻱ ﻗﺮﺍﺭ ﺑﮕﻴﺮﻧﺪ ﻛﻪ ﻫﺮ ﺗﺮﺍﻛﻨﺶ‬
‫ﺩﺭ ﻣﺠﻤﻮﻋﻪ ﻣﻨﺘﻈﺮ ﺗﺮﺍﻛﻨﺶ ﺩﻳﮕﺮ ﺁﻥ ﻣﺠﻤﻮﻋﻪ ﺑﺎﺷﺪ‪.‬‬
‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﭘﻴﺸﮕﻴﺮﻱ ﺍﺯ ﺑﻦ ﺑﺴﺖ ﺍﻳﻦ ﺍﻃﻤﻴﻨﺎﻥ ﺭﺍ ﺣﺎﺻﻞ ﻣﻲ ﻛﻨﻨﺪ ﻛﻪ ﺳﻴﺴﺘﻢ ﻫﺮﮔﺰ ﺑﻪ ﺣﺎﻟﺖ‬
‫ﺑﻦ ﺑﺴﺖ ﻭﺍﺭﺩ ﻧﻤﻲ ﺷﻮﺩ‪ .‬ﺑﻌﻀﻲ ﺍﺯ ﺭﺍﻫﺒﺮﺩﻫﺎﻱ ﭘﻴﺸﮕﻴﺮﻱ ‪:‬‬
‫• ﻧﻴﺎﺯ ﺍﺳﺖ ﻛﻪ ﻫﻤﻪ ﻱ ﻗﻔﻞ ﻫﺎﻱ ﻫﺮ ﺗﺮﺍﻛﻨﺶ ﺩﺭ ﺑﺨﺶ ﻫﺎﻱ ﺩﺍﺩﻩ ﻗﺒﻞ ﺍﺯ ﺁﻏﺎﺯ ﺑﺮﺭﺳﻲ‬
‫ﺷﻮﻧﺪ)ﺍﻃﻼﻉ ﺭﺳﺎﻧﻲ(‬
‫• ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﺑﺨﺸﻲ ﺍﺯ ﺗﻤﺎﻡ ﺑﺨﺶ ﻫﺎﻱ ﺍﻃﻼﻋﺎﺕ ﻭ ﻟﺰﻭﻡ ﺍﻳﻦ ﻛﻪ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﻓﻘﻂ ﺩﺭ‬
‫ﺑﺨﺶ ﻫﺎﻱ ﻣﺸﺨﺺ ﺷﺪﻩ ﺑﺘﻮﺍﻧﺪ ﺑﺨﺶ ﺭﻭﻱ ﺩﺍﺩﻩ ﺭﺍ ﻗﻔﻞ ﻛﻨﺪ)ﭘﺮﻭﺗﻜﻞ ﺑﺮ ﭘﺎﻳﻪ ﻱ ﮔﺮﺍﻑ(‬

‫ﺍﺳﺘﺮﺍﺗﮋﻱ ﻫﺎﻱ ﭘﻴﺸﮕﻴﺮﻱ ﺍﺯ ﺑﻦ ﺑﺴﺖ ﻫﺎﻱ ﺑﻴﺸﺘﺮ‬

‫‪81‬‬
‫‪ ‬ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﺍﺷﺎﺭﻩ ﺷﺪﻩ ﺗﻨﻬﺎ ﺑﺮﺍﻱ ﭘﻴﺸﮕﻴﺮﻱ ﺍﺯ ﺑﻦ ﺑﺴﺖ ﺍﺯ ﺑﺮﭼﺴﺐ ﻫﺎﻱ ﺯﻣﺎﻧﻲ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﻛﻨﺪ‪.‬‬
‫‪ ‬ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﻣﺮگ ‪ -‬ﺍﻧﺘﻈﺎﺭ ‪ -‬ﻏﻴﺮﭘﻴﺸﮕﻴﺮﺍﻧﻪ‬
‫• ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﻗﺪﻳﻤﻲ ﺗﺮ ﻣﻤﻜﻦ ﺍﺳﺖ ﺑﺨﺸﻲ ﺍﺯ ﺩﺍﺩﻩ ﺭﺍ ﺑﺮﺍﻱ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺟﺪﻳﺪﺗﺮ ﺁﺯﺍﺩ‬
‫ﻛﻨﻨﺪ ﺍﻣﺎ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﺟﺪﻳﺪﺗﺮ ﻫﺮﮔﺰ ﺑﺮﺍﻱ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﻗﺪﻳﻤﻲ ﺗﺮ ﺑﻪ ﺣﺎﻟﺖ ﺍﻧﺘﻈﺎﺭ‬
‫ﻧﻤﻲ ﺭﻭﻧﺪ ﺑﻠﻜﻪ ﺑﻪ ﺟﺎﻱ ﺁﻥ ﻋﻘﺒﮕﺮﺩ ﻣﻲ ﻛﻨﻨﺪ‪.‬‬
‫• ﺗﺮﺍﻛﻨﺶ ﻣﻤﻜﻦ ﺍﺳﺖ ﭼﻨﺪﻳﻦ ﺑﺎﺭ ﻗﺒﻞ ﺍﺯ ﺩﺳﺘﻴﺎﺑﻲ ﺑﻪ ﺑﺨﺶ ﺩﺍﺩﻩ ﻱ ﻣﻮﺭﺩ ﻧﻴﺎﺯﺵ ﺍﺯ ﺑﻴﻦ‬
‫ﺑﺮﻭﺩ‪.‬‬
‫‪ ‬ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﺟﺮﺍﺣﺖ – ﺍﻧﺘﻈﺎﺭ ‪ -‬ﭘﻴﺸﮕﻴﺮﺍﻧﻪ‬
‫• ﺕﺭﺍﻛﻨﺶ ﻫﺎﻱ ﻗﺪﻳﻤﻲ ﺗﺮﺑﻪ ﺟﺎﻱ ﺍﻳﻦ ﻛﻪ ﻣﻨﺘﻈﺮ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﺟﺪﻳﺪ ﺑﺎﺷﻨﺪ ﺩﭼﺎﺭ‬
‫ﺟﺮﺍﺣﺖ )ﺑﺮﮔﺸﺖ ﺗﺤﻤﻴﻠﻲ ( ﻣﻲ ﺷﻮﻧﺪ‪.‬‬
‫• ﻣﻤﻜﻦ ﺍﺳﺖ ﻋﻘﺐ ﮔﺮﺩﻫﺎﻱ ﻛﻤﺘﺮﻱ ﺩﺭ ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﻣﺮگ‪-‬ﺍﻧﺘﻈﺎﺭ ﻧﻘﺶ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻨﺪ‪.‬‬

‫ﭘﻴﺸﮕﻴﺮﻱ ﺍﺯ ﺑﻦ ﺑﺴﺖ‬

‫‪ ‬ﺩﺭ ﻫﺮﺩﻭ ﺣﺎﻟﺖ ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﻣﺮگ –ﺍﻧﺘﻈﺎﺭ ﻭ ﺟﺮﺍﺣﺖ‪-‬ﺍﻧﺘﻈﺎﺭ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﺑﺮﮔﺸﺖ ﺧﻮﺭﺩﻩ ﺑﺎ‬
‫ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﺍﺻﻠﻲ ﺷﺎﻥ ﻣﺠﺪﺩﺍ ﺭﺍ ﻩ ﺍﻧﺪﺍﺯﻱ ﻣﻲ ﺷﻮﻧﺪ‪ .‬ﺩﺭ ﻧﺘﻴﺠﻪ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﻗﺪﻳﻤﻲ ﺗﺮ ﺩﺍﺭﺍﻱ‬
‫ﺍﻭﻝﻭﻳﺖ ﺑﻴﺸﺘﺮﻱ ﺍﺯ ﺟﺪﻳﺪﺗﺮﻫﺎ ﻣﻲ ﺷﻮﻧﺪ ﺍﺯ ﺍﻳﻦ ﺭﻭ ﻗﺤﻄﻲ ﺍﺟﺘﻨﺎﺏ ﻧﺎﭘﺬﻳﺮ ﺍﺳﺖ‪.‬‬
‫‪ ‬ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﻣﺒﺘﻨﻲ ﺑﺮ ﺑﺮﺟﺴﺐ ﺯﻣﺎﻥ ‪:‬‬
‫• ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﻓﻘﻂ ﺑﺮﺍﻱ ﻣﺪﺕ ﺯﻣﺎﻥ ﻣﻌﻴﻨﻲ ﺑﺮﺍﻱ ﻳﻚ ﻗﻔﻞ ﻣﻨﺘﻈﺮ ﻣﻲ ﻣﺎﻧﺪ ﭘﺲ ﺍﺯ ﺁﻥ ﺍﺯ‬
‫ﺻﻒ ﺍﻧﺘﻈﺎﺭ ﺧﺎﺭﺝ ﻭ ﺗﺮﺍﻛﻨﺶ ﻋﻤﻞ ﻋﻘﺒﮕﺮﺩ ﺭﺍ ﺍﻧﺠﺎﻡ ﻣﻲ ﺩﻫﺪ‪.‬‬
‫• ﺩﺭ ﻧﺘﻴﺠﻪ ﺑﻦ ﺑﺴﺖ ﺍﻣﻜﺎﻥ ﭘﺬﻳﺮ ﻧﻴﺴﺖ‪.‬‬
‫• ﭘﻴﺎﺩﻩ ﺳﺎﺯﻱ ﺍﻳﻦ ﺭﻭﺵ ﺳﺎﺩﻩ ﺍﺳﺖ ‪ ،‬ﺍﻣﺎ ﻗﺤﻄﻲ ﺷﺎﻥ ﺍﻣﻜﺎﻥ ﭘﺬﻳﺮ ﺍﺳﺖ‪ .‬ﻫﻤﭽﻨﻴﻦ ﺗﻌﻴﻴﻦ‬
‫ﻣﻘﺪﺍﺭ ﺁﻥ ﺑﺮﺍﻱ ﻳﻚ ﻓﺎﺻﻠﻪ ﺯﻣﺎﻧﻲ ﺩﺷﻮﺍﺭ ﺍﺳﺖ‪.‬‬

‫ﺗﺸﺨﻴﺺ ﺑﻦ ﺑﺴﺖ‬

‫‪ ‬ﺑﻦ ﺑﺴﺖ ﺭﺍ ﻣﻲ ﺗﻮﺍﻥ ﺑﻪ ﻋﻨﻮﺍﻥ ﻳﻚ ﮔﺮﺍﻑ ‪ ،wait-for‬ﻣﺘﺸﻜﻞ ﺍﺯ ﻳﻚ ﺟﻔﺖ )‪ G=(V,E‬ﺩﺭ ﻧﻈﺮ ﮔﺮﻓﺖ‬

‫• ‪ V‬ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺭﺋﻮﺱ ﺍﺳﺖ )ﺗﻤﺎﻣﻲ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺩﺭ ﺳﻴﺴﺘﻢ(‬

‫‪82‬‬
 ‫• ‪ E‬ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﻟﺒﻪ ﻫﺎ ﺍﺳﺖ‪ ،‬ﻫﺮ ﻋﻨﺼﺮﻳﻚ ﺯﻭﺝ ﻣﺮﺗﺐ ‪ Ti→Tj‬ﺍﺳﺖ‪.‬‬

‫‪ ‬ﺍﮔﺮ ‪ Ti→Tj‬ﺩﺭ ‪ ،E‬ﺁﻧﮕﺎﻩ ﻳﻚ ﻟﺒﻪ ﻣﺴﺘﻘﻴﻢ ﺍﺯ ‪ Ti‬ﺑﻪ ‪ Tj‬ﻭﺟﻮﺩ ﺩﺍﺭﺩ‪ ،‬ﻭ ﺑﻪ ﻋﺒﺎﺭﺗﻲ ‪ Ti‬ﺑﺮﺍﻱ ﺁﺯﺍﺩﻛﺮﺩﻥ ﻳﻚ‬
‫ﺑﺨﺶ ﺍﺯ ﺩﺍﺩﻩ ﻣﻨﺘﻈﺮ ‪ Tj‬ﺍﺳﺖ‪.‬‬
‫‪ ‬ﻫﻨﮕﺎﻣﻲ ﻛﻪ ‪ Ti‬ﺩﺭﺧﻮﺍﺳﺖ ﻳﻚ ﺑﺨﺶ ﺍﺯ ﺩﺍﺩﻩ ﻛﻪ ﺩﺭ ﺣﺎﻝ ﺣﺎﺿﺮ ﺗﻮﺳﻂ ‪ Tj‬ﻧﮕﻬﺪﺍﺭﻱ ﻣﻲ ﺷﻮﺩ ﺭﺍﻣﻲ ﻛﻨﺪ‪،‬‬
‫ﺳﭙﺲ ﻟﺒﻪ ‪ TiTj‬ﺩﺭ ﮔﺮﺍﻑ ‪ wait-for‬ﺩﺭﺝ ﻣﻲ ﺷﻮﺩ‪ .‬ﺍﻳﻦ ﻟﺒﻪ ﻓﻘﻂ ﺯﻣﺎﻧﻲ ﺣﺬﻑ ﻣﻲ ﺷﻮﺩ ﻛﻪ ‪ Tj‬ﻳﻚ ﺑﺨﺶ‬
‫ﺩﺍﺩﻩ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﺗﻮﺳﻂ ‪ Ti‬ﺭﺍ ﻃﻮﻻﻧﻲ ﻧﮕﻪ ﻧﺪﺍﺭﺩ‪.‬‬
‫‪ ‬ﺍﻳﻦ ﺳﻴﺴﺘﻢ ﺩﺭ ﺣﺎﻟﺖ ﺑﻦ ﺑﺴﺖ ﺍﺳﺖ ﺍﮔﺮ ﻭ ﺗﻨﻬﺎ ﺍﮔﺮ ﮔﺮﺍﻑ ‪ wait-for‬ﺩﺍﺭﺍﻱ ﻳﻚ ﭼﺮﺧﻪ ﺑﺎﺷﺪ‪ .‬ﺑﺎﻳﺪ ﺑﻪ‬
‫ﻳﻚ ﺍﻟﮕﻮﺭﻳﺘﻢ ﺗﺸﺨﻴﺺ ﺑﻦ ﺑﺴﺖ ﺑﻪ ﺻﻮﺭﺕ ﺩﻭﺭﻩ ﺍﻱ ﺑﺮﺍﻱ ﺍﺷﺎﺭﻩ ﺑﻪ ﭼﺮﺧﻪ ﻫﺎ ﺍﺳﺘﻨﺎﺩ ﻛﺮﺩ‪.‬‬

‫ﮔﺭﺍﻑ‪wait-for‬ﺑﺩﻭﻥ ﭼﺭﺧﻪ‬ ‫ﮔﺭﺍﻑ ‪ wait-for‬ﺑﺎ ﭼﺭﺧﻪ‬

‫ﺑﺎﺯﻳﺎﺑﻲ ﺑﻦ ﺑﺴﺖ‬

‫‪ ‬ﭼﻪ ﻭﻗﺖ ﺑﻦ ﺑﺴﺖ ﺷﻨﺎﺳﺎﻳﻲ ﻣﻲ ﺷﻮﺩ‪:‬‬

‫• ﺑﺮﺧﻲ ﺍﺯ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺑﻪ ﻋﻘﺐ ﺑﺮﻣﻲ ﮔﺮﺩﻧﺪ )ﻗﺮﺑﺎﻧﻲ( ﺑﺮﺍﻱ ﺷﻜﺴﺘﻦ ﺑﻦ ﺑﺴﺖ‪ .‬ﺗﺮﺍﻛﻨﺸﻲ‬
‫ﻛﻪ ﺑﻪ ﻋﻨﻮﺍﻥ ﻗﺮﺑﺎﻧﻲ ﻛﻤﺘﺮﻳﻦ ﻫﺰﻳﻨﻪ ﺭﺍ ﺩﺍﺭﺩ ﺍﻧﺘﺨﺎﺏ ﻛﻨﻴﺪ‪.‬‬
‫• ﻋﻘﺒﮕﺮﺩ‪ -‬ﺗﻌﻴﻴﻦ ﺍﻳﻦ ﻛﻪ ﺗﺎ ﭼﻪ ﺣﺪ ﺑﺎﻳﺪ ﺑﻪ ﻋﻘﺐ ﺑﺮﮔﺸﺖ‪.‬‬
‫‪ ‬ﺗﻌﺪﺍﺩ ﻋﻘﺒﮕﺮﺩ ‪ :‬ﺑﻲ ﻧﺘﻴﺠﻪ ﻣﺎﻧﺪﻥ ﺗﺮﺍﻛﻨﺶ ﻭ ﺳﭙﺲ ﺭﺍﻩ ﺍﻧﺪﺍﺯﻱ ﻣﺠﺪﺩ ﺁﻥ‬

‫‪83‬‬
 ‫‪ ‬ﺗﺄﺛﻴﺮ ﺑﻴﺸﺘﺮ ﺗﺮﺍﻛﻨﺶ ﻋﻘﺒﮕﺮﺩ ﻓﻘﻂ ﺑﺮﺍﻱ ﺷﻜﺴﺘﻦ ﺑﻦ ﺑﺴﺖ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﺍﺳﺖ‪.‬‬
‫• ﻗﺤﻄﻲ ﺍﺗﻔﺎﻕ ﻣﻲ ﺍﻓﺘﺪ ﺍﮔﺮ ﻫﻤﺎﻥ ﺗﺮﺍﻛﻨﺶ ﺑﺎﺷﺪ ﻛﻪ ﺑﻪ ﻋﻨﻮﺍﻥ ﻗﺮﺑﺎﻧﻲ ﺍﻧﺘﺨﺎﺏ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬
‫ﺗﻌﺪﺍﺩ ﻋﻘﺒﮕﺮﺩﻫﺎ ﺩﺭ ﻋﺎﻣﻞ ﻫﺰﻳﻨﻪ ﺑﺮﺍﻱ ﺟﻠﻮﮔﻴﺮﻱ ﺍﺯ ﮔﺮﺳﻨﮕﻲ ﺭﺍ ﺷﺎﻣﻞ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﺭﻭﻳﻜﺮﺩﻫﺎﻱ ﺩﻳﮕﺮ ﺑﺮﺍﻱ ﻛﻨﺘﺮﻝ ﻫﻤﺰﻣﺎﻧﻲ‬

‫ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﺑﺮ ﭘﺎﻳﻪ ﺯﻣﺎﻥ‬

‫‪ ‬ﻫﺮ ﺗﺮﺍﻛﻨﺶ ﺻﺎﺩﺭﻛﻨﻨﺪﻩ ﻳﻚ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﺍﺳﺖ ﻛﻪ ﻭﺍﺭﺩ ﺳﻴﺴﺘﻢ ﻣﻲ ﺷﻮﺩ‪ .‬ﺍﮔﺮ ﻳﻚ ﺗﺮﺍﻛﻨﺶ‬
‫ﻗﺪﻳﻤﻲ ‪ Ti‬ﺩﺍﺭﺍﻱ ﺯﻣﺎﻥ‪-‬ﭼﺎپ )‪ TS(Ti‬ﺑﺎﺷﺪ‪ ،‬ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺟﺪﻳﺪ ‪ Tj‬ﺯﻣﺎﻥ‪-‬ﭼﺎپ )‪ TS(Tj‬ﺭﺍ‬
‫ﺗﺨﺼﻴﺺ ﻣﻲ ﺩﻫﺪ ﺑﻪ ﻃﻮﺭﻱ ﻛﻪ )‪TS(Ti)<TS(Tj‬‬
‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﻣﺪﻳﺮﻳﺖ ﺍﺟﺮﺍﻱ ﻫﻤﺰﻣﺎﻥ ﺑﻪ ﻃﻮﺭﻱ ﻛﻪ ﺯﻣﺎﻥ – ﭼﺎپ ﻫﺎﻱ ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﺳﺮﻳﺎﻟﻲ ﺭﺍ ﺗﻌﻴﻴﻦ‬
‫ﻡﻱ ﻛﻨﺪ‪.‬‬
‫‪ ‬ﺑﺮﺍﻱ ﺍﻃﻤﻴﻨﺎﻥ ﺍﺯ ﭼﻨﻴﻦ ﺭﻓﺘﺎﺭﻱ‪ ،‬ﭘﺮﻭﺗﻜﻞ ﺑﺮﺍﻱ ﻫﺮ ﺩﺍﺩﻩ ‪ 2 Q‬ﻣﻘﺪﺍﺭ ﺍﺯ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﺭﺍ ﻧﮕﻪ ﻣﻲ‬
‫ﺩﺍﺭﺩ‪:‬‬

‫• )‪ W-timestamp(Q‬ﺑﺰﺭﮔﺘﺮﻳﻦ ﺯﻣﺎﻥ‪-‬ﭼﺎپ ﻫﺮ ﺗﺮﺍﻛﻨﺶ ﻛﻪ )‪ write(Q‬ﺭﺍ ﺑﻪ ﻃﻮﺭ ﻣﻮﻓﻘﻴﺖ‬

‫ﺁﻣﻴﺰ ﺍﺟﺮﺍ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫• )‪ R-timestamp(Q‬ﺑﺰﺭﮔﺘﺮﻳﻦ ﺯﻣﺎﻥ‪-‬ﭼﺎپ ﻫﺮ ﺗﺮﺍﻛﻨﺶ ﻛﻪ )‪ read(Q‬ﺭﺍ ﺑﻪ ﻃﻮﺭ ﻣﻮﻓﻘﻴﺖ ﺁﻣﻴﺰ‬

‫ﺍﺟﺮﺍ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﺩﺳﺘﻮﺭ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﺗﻀﻤﻴﻦ ﻣﻲ ﻛﻨﺪ ﻛﻪ ﻫﻴﭻ ﺿﺪ ﻭ ﻧﻘﻴﻀﻲ ﺩﺭ ﻋﻤﻠﻴﺎﺕ ﺧﻮﺍﻧﺪﻥ ﻭ‬
‫ﻧﻮﺷﺘﻦ ﺩﺭ ﺩﺳﺘﻮﺭ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﺍﺟﺮﺍ ﻧﻤﻲ ﺷﻮﺩ‪.‬‬
‫‪ ‬ﻳﻚ ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﻣﺮﺑﻮﻁ ﺑﻪ ﻣﺴﺎﺋﻞ ﻳﻚ )‪ read(Q‬ﺭﺍ ﻓﺮﺽ ﻛﻨﻴﺪ‪.‬‬
‫‪ .1‬ﺍﮔﺮ )‪ ،TS(Ti)≤ w-timestamp(Q‬ﺳﭙﺲ ‪ Ti‬ﺑﻪ ﺧﻮﺍﻧﺪﻥ ﻱﻙ ﻣﻘﺪﺍﺭ ﺍﺯ ‪ Q‬ﻛﻪ ﺩﺭ ﺣﺎﻝ‬
‫ﺣﺎﺿﺮ ﺭﻭﻧﻮﻳﺴﻲ ﻣﻲ ﺷﻮﺩ ﻧﻴﺎﺯ ﺩﺍﺭﺩ‪.‬‬
‫‪ ‬ﺍﺯ ﺍﻳﻦ ﺭﻭ‪ ،‬ﻋﻤﻞ ﺧﻮﺍﻧﺪﻥ ﺭﺩ ﻣﻲ ﺷﻮﺩ‪ ،‬ﻭ ‪ Ti‬ﺑﻪ ﻋﻘﺐ ﺑﺮ ﻣﻲ ﮔﺮﺩﺩ‪.‬‬
‫‪ .2‬ﺍﮔﺮ )‪ TS(Ti)≥w-timestamp(Q‬ﺑﺎﺷﺪ‪ ،‬ﺳﭙﺲ ﻋﻤﻠﻴﺎﺕ ﺧﻮﺍﻧﺪﻥ ﺍﺟﺮﺍ ﻣﻲ ﺷﻮﺩ‪ ،‬ﻭ ‪R-‬‬
‫)‪ timestamp(Q‬ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺑﺮﺍﻱ )‪ max(R-timestamp(Q‬ﻭ )‪ TS(Ti‬ﺍﺳﺖ‪.‬‬

‫‪84‬‬
 ‫‪ ‬ﻓﺮﺽ ﻛﻨﻴﺪ ﻛﻪ ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﺩﺍﺭﺍﻱ ﻣﺴﺎﺋﻞ ﻣﺮﺑﻮﻁ ﺑﻪ )‪ write(Q‬ﺍﺳﺖ‪.‬‬
‫‪ .1‬ﺍﮔﺮ )‪ TS(Ti)<R-timestamp(Q‬ﺑﺎﺷﺪ‪ ،‬ﺳﭙﺲ ﻣﻘﺪﺍﺭ ‪ Q‬ﻛﻪ ‪ Ti‬ﺗﻮﻟﻴﺪ ﻣﻲ ﻛﻨﺪ ﻗﺒﻼ ﻣﻮﺭﺩ‬
‫ﻧﻴﺎﺯ ﺑﻮﺩﻩ‪ ،‬ﻭ ﺳﻴﺴﺘﻢ ﺁﻥ ﺭﺍ ﺑﺮﻋﻬﺪﻩ ﮔﺮﻓﺘﻪ ﻛﻪ ﺁﻥ ﻣﻘﺪﺍﺭ ﻫﺮﮔﺰ ﺗﻮﻟﻴﺪ ﻧﺸﻮﺩ‪.‬‬
‫‪ ‬ﺍﺯ ﺍﻳﻦ ﺭﻭ‪ ،‬ﺍﻳﻦ ﻋﻤﻠﻴﺎﺕ ﻧﻮﺷﺘﻦ ﺭﺩ ﻣﻲ ﺷﻮﺩ‪ ،‬ﻭ ‪ Ti‬ﺑﻪ ﻋﻘﺐ ﺑﺮ ﻣﻲ ﮔﺮﺩﺩ‪.‬‬
‫‪ .2‬ﺍﮔﺮ )‪ TS(Ti)<w-timestamp(Q‬ﺑﺎﺷﺪ‪ ،‬ﺳﭙﺲ ‪ Ti‬ﺑﺮﺍﻱ ﺟﺴﺘﺠﻮﻱ ﻳﻚ ﻣﻘﺪﺍﺭ ﻏﻴﺮﻗﺎﺑﻞ‬
‫ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ‪ Q‬ﺗﻼﺵ ﻣﻲ ﻛﻨﺪ‪.‬‬
‫‪ ‬ﺍﺯ ﺍﻳﻦ ﺭﻭ‪ ،‬ﺍﻳﻦ ﻋﻤﻠﻴﺎﺕ ﻧﻮﺷﺘﻦ ﺭﺩ ﻣﻲ ﺷﻮﺩ‪ ،‬ﻭ ‪ Ti‬ﺑﻪ ﻋﻘﺐ ﺑﺮ ﻣﻲ ﮔﺮﺩﺩ‪.‬‬
‫‪ .3‬ﺩﺭ ﻏﻴﺮ ﺍﻳﻦ ﺻﻮﺭﺕ‪ ،‬ﻋﻤﻠﻴﺎﺕ ﻧﻮﺷﺘﻦ ﺍﺟﺮﺍ ﻣﻲ ﺷﻮﺩ‪ ،‬ﻭ )‪ w-timestamp(Q‬ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ‬
‫)‪ TS(Ti‬ﺍﺳﺖ‪.‬‬

‫ﻣﺜﺎﻝ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﭘﺮﻭﺗﻜﻞ‬

‫ﺑﺮﻧﺎﻣﻪ ﺯﻳﺮ ﺑﺮﺍﻱ ﭼﻨﺪﻳﻦ ﺁﻳﺘﻢ ﺩﺍﺩﻩ ﺑﺮﺍﻱ ﺍﻧﺠﺎﻡ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻳﻲ ﺑﺎ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻧﻲ ‪ ،4 ،3 ،2 ،1‬ﺍﺳﺖ‬

‫‪T1‬‬ ‫‪T2‬‬ ‫‪T3‬‬ ‫‪T4‬‬ ‫‪T5‬‬

‫)‪read(X‬‬
‫)‪read(Y‬‬
‫)‪read(Y‬‬
‫)‪write(Y‬‬
‫)‪write(Z‬‬
‫)‪read(Z‬‬
‫)‪read(X‬‬
‫‪abort‬‬
‫)‪read(X‬‬
‫)‪write(Z‬‬
‫‪abort‬‬
‫)‪write(Y‬‬
‫)‪write(Z‬‬

‫ﺻﺤﺖ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﺩﺳﺘﻮﺭ ﭘﺮﻭﺗﻜﻞ‬

‫‪85‬‬
 ‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﺩﺳﺘﻮﺭ‪ -‬ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﺗﻤﺎﻡ ﻛﻤﺎﻥ ﻫﺎﻱ ﮔﺮﺍﻑ ﺍﻭﻟﻮﻳﺖ ﺍﺯ ﻓﺮﻡ ﺭﺍ ﺑﻪ ﺻﻮﺭﺕ ﺳﺮﻳﺎﻟﻲ‬
‫ﺗﻀﻤﻴﻦ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫‪transaction‬‬ ‫‪transaction‬‬
‫‪with smaller‬‬ ‫‪with larger‬‬
‫‪timestamp‬‬ ‫‪timestamp‬‬

‫ﺑﻨﺎﺑﺮ ﺍﻳﻦ‪ ،‬ﻫﻴﭻ ﭼﺮﺧﻪ ﺍﻱ ﺩﺭ ﮔﺮﺍﻑ ﺍﻭﻟﻮﻳﺖ ﻭﺟﻮﺩ ﻧﺪﺍﺭﺩ‪.‬‬

‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﺭﻫﺎﻳﻲ ﺍﺯ ﺑﻦ ﺑﺴﺖ ﺭﺍ ﺗﻀﻤﻴﻦ ﻣﻲ ﻛﻨﺪ ﻛﻪ ﻫﻴﭻ ﺗﺮﺍﻛﻨﺸﻲ ﻣﻨﺘﻈﺮ ﻧﻤﻲ ﻣﺎﻧﺪ‪.‬‬
‫‪ ‬ﺑﺮﻧﺎﻣﻪ ﻣﻤﻜﻦ ﻧﻴﺴﺖ ﺁﺑﺸﺎﺭﻱ‪-‬ﺁﺯﺍﺩ ﺑﺎﺷﺪ‪ ،‬ﻭ ﺣﺘﻲ ﻣﻤﻜﻦ ﺍﺳﺖ ﺑﺎﺯﻳﺎﺑﻲ ﻧﺸﻮﺩ‪.‬‬

‫ﺑﺎﺯﻳﺎﺑﻲ ﻭ ﺍﺳﺘﻘﻼﻝ ﺁﺑﺸﺎﺭﻱ‬

‫‪ ‬ﻣﺸﻜﻼﺕ ﭘﺮﻭﺗﻜﻞ ﺩﺳﺘﻮﺭ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ‪:‬‬

‫• ﻓﺮﺽ ﺷﺪﻩ ﺑﺮ ﺑﻲ ﻧﺘﻴﺠﻪ ﻣﺎﻧﺪﻥ‪ ،‬ﺍﻣﺎ ‪ Tj‬ﻳﻚ ﺑﺨﺶ ﺍﺯ ﺩﺍﺩﻩ ﻧﻮﺷﺘﻪ ﺷﺪﻩ ﺗﻮﺳﻂ ‪ Ti‬ﺭﺍ ﻣﻲ‬
‫ﺧﻮﺍﻧﺪ‪.‬‬
‫• ﺑﻨﺎﺑﺮﺍﻳﻦ ‪ Tj‬ﺑﺎﻳﺪ ﺑﻲ ﻧﺘﻴﺠﻪ ﺑﻤﺎﻧﺪ‪ ،‬ﺍﮔﺮ ‪ Tj‬ﺑﻪ ﺗﺄﻳﻴﺪ ﺯﻭﺩﺗﺮ ﺍﺟﺎﺯﻩ ﺩﻫﺪ‪ ،‬ﺑﺮﻧﺎﻣﻪ ﻏﻴﺮﻗﺎﺑﻞ‬
‫ﺑﺎﺯﻳﺎﺑﻲ ﺍﺳﺖ‪.‬‬
‫• ﻋﻼﻭﻩ ﺑﺮ ﺍﻳﻦ‪ ،‬ﻫﺮ ﮔﻮﻧﻪ ﺗﺮﺍﻛﻨﺸﻲ ﻛﻪ ﻳﻚ ﺑﺨﺶ ﺍﺯ ﺩﺍﺩﻩ ﻧﻮﺷﺘﻪ ﺵﺩﻩ ﺗﻮﺳﻂ ‪ Tj‬ﺭﺍ ﺑﺨﻮﺍﻧﺪ‬
‫ﺑﺎﻳﺪ ﺑﻲ ﻧﺘﻴﺠﻪ ﺑﻤﺎﻧﺪ‪.‬‬
‫• ﻣﻲ ﺗﻮﺍﻧﺪ ﻣﻨﺠﺮ ﺑﻪ ﻋﻘﺐ ﮔﺮﺩ ﺁﺑﺸﺎﺭﻱ ﺷﻮﺩ‪ ،‬ﻛﻪ ﻳﻚ ﻋﻘﺐ ﮔﺮﺩ ﺯﻧﺠﻴﺮﻩ ﺍﻱ ﺍﺳﺖ‪.‬‬
‫‪ ‬ﺭﺍﻩ ﺣﻞ ‪: 1‬‬
‫• ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺩﺭ ﻣﻮﺭﺩ ﺳﺎﺧﺘﺎﺭﺵ ﭼﻨﻴﻦ ﻣﻲ ﻧﻮﻳﺴﺪ ﻛﻪ ﻫﻤﻪ ﺭﺍ ﺩﺭ ﭘﺎﻳﺎﻥ ﭘﺮﺩﺍﺯﺵ ﺍﻧﺠﺎﻡ‬
‫ﻣﻲ ﺩﻫﺪ‪.‬‬
‫• ﻫﻤﻪ ﺍﺯ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺍﺯ ﻳﻚ ﻋﻤﻞ ﺍﺗﻤﻴﻚ ﻣﻲ ﻧﻮﻳﺴﻨﺪ‪ ،‬ﻫﻴﭻ ﺗﺮﺍﻛﻨﺸﻲ ﺍﻣﻜﺎﻥ ﻧﺪﺍﺭﺩ ﺍﺟﺮﺍ‬
‫ﺷﻮﺩ ﺩﺭ ﺣﺎﻟﻲ ﻛﻪ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﻧﻮﺷﺘﻪ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬

‫‪86‬‬
‫• ﻳﻚ ﺗﺮﺍﻛﻨﺸﻲ ﻛﻪ ﺑﻲ ﻧﺘﻴﺠﻪ ﻣﻲ ﻣﺎﻧﺪ ﻣﺠﺪﺩﺍ ﺑﺎ ﻳﻚ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻧﻲ ﺟﺪﻳﺪ ﺭﺍﻩ ﺍﻧﺪﺍﺯﻱ ﻣﻲ‬
‫ﺷﻮﺩ‪.‬‬
‫‪ ‬ﺭﺍﻩ ﺣﻞ ‪ : 2‬ﻣﺤﺪﻭﺩ ﻛﺮﺩﻥ ﻓﺮﻡ ﻗﻔﻞ‪ ،‬ﺍﻧﺘﻈﺎﺭ ﺑﺮﺍﻱ ﺩﺍﺩﻩ ﻛﻪ ﻗﺒﻞ ﺍﺯ ﺧﻮﺍﻧﺪﻥ ﺁﻥ ﺗﺄﻳﻴﺪ ﻣﻲ ﺷﻮﺩ‪.‬‬
‫‪ ‬ﺭﺍﻩ ﺣﻞ ‪ : 3‬ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻭﺍﺑﺴﺘﮕﻲ ﻫﺎﻱ ﻣﻮﺭﺩ ﺗﺄﻳﻴﺪ ﺑﺮﺍﻱ ﺍﻃﻤﻴﻨﺎﻥ ﺍﺯ ﺑﺎﺯﻳﺎﺑﻲ‬

‫ﻧﻮﺷﺖ ﺷﺪﻩ ﻱ ﺗﻮﻣﺎﺱ‬

‫ﻩ‬ ‫ﻗﺎﻧﻮﻥ‬

‫‪ ‬ﻧﺴﺨﻪ ﺍﺻﻼﺡ ﺷﺪﻩ ﭘﺮﻭﺗﻜﻞ ﺩﺳﺘﻮﺭ‪ -‬ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﻛﻪ ﺩﺭ ﺁﻥ ﻋﻤﻠﻴﺎﺕ ﻣﻨﺴﻮﺥ ﺷﺪﻩ ﻣﻤﻜﻦ ﺍﺳﺖ‬
‫ﺗﺤﺖ ﺷﺮﺍﻳﻂ ﺧﺎﺹ ﻧﺎﺩﻳﺪﻩ ﮔﺮﻓﺘﻪ ﺷﻮﺩ‪.‬‬
‫‪ ‬ﻫﻨﮕﺎﻣﻲ ﻛﻪ ‪Ti‬ﺑﺮﺍﻱ ﻧﻮﺷﺘﻦ ﺁﻳﺘﻢ ﺩﺍﺩﻩ ‪Q‬ﺗﻼﺵ ﻣﻲ ﻛﻨﺪ‪ ،‬ﺍﮔﺮ )‪TS(Ti)<w-timestamp(Q‬‬
‫ﺑﺎﺷﺪ‪ ،‬ﺳﭙﺲ ‪ Ti‬ﺑﺮﺍﻱ ﻧﻮﺷﺘﻦ ﻳﻚ ﻣﻘﺪﺍﺭ ﻏﻴﺮﻗﺎﺑﻞ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ}‪ {Q‬ﺗﻼﺵ ﻣﻲ ﻛﻨﺪ‪.‬‬
‫• ﺑﻪ ﺟﺎﻱ ﻋﻘﺐ ﮔﺮﺩ ‪ Ti‬ﺑﻪ ﻋﻨﻮﺍﻥ ﭘﺮﻭﺗﻜﻞ ﺩﺳﺘﻮﺭ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ‪ ،‬ﻋﻤﻠﻴﺎﺕ }‪ {write‬ﺭﺍ‬
‫ﻣﻲ ﺗﻮﺍﻥ ﻧﺎﺩﻳﺪﻩ ﮔﺮﻓﺖ ‪.‬‬
‫‪ ‬ﺩﺭ ﻏﻴﺮ ﺍﻳﻦ ﺻﻮﺭﺕ ﺍﻳﻦ ﭘﺮﻭﺗﻜﻞ ﻫﻤﺎﻥ ﭘﺮﻭﺗﻜﻞ ﺩﺳﺘﻮﺭ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﺍﺳﺖ‪.‬‬
‫‪ ‬ﻗﺎﻧﻮﻥ ﻧﻮﺷﺘﻪ ﺷﺪﻩ ﻱ ﺗﻮﻣﺎﺱ ﺍﺟﺎﺯﻩ ﺑﻪ ﭘﺘﺎﻧﺴﻴﻞ ﻫﻤﺰﻣﺎﻧﻲ ﺑﻴﺸﺘﺮ ﻣﻲ ﺩﻫﺪ‪.‬‬
‫• ﺍﺟﺎﺯﻩ ﻣﻲ ﺩﻫﺪ ﺗﺎ ﺑﺮﺧﻲ ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﻗﺎﺑﻞ ﺩﻳﺪﻥ ﺳﺮﻳﺎﻟﻲ ﻣﺘﻀﺎﺩ ﺳﺮﻳﺎﻟﻲ ﻧﺒﺎﺷﻨﺪ‪.‬‬

‫ﭘﺮﻭﺗﻜﻞ ﻣﺒﺘﻨﻲ ﺑﺮ ﺍﻋﺘﺒﺎﺭ ﺳﻨﺠﻲ‬

‫‪ ‬ﺍﺟﺮﺍﻱ ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﺩﺭ ﺳﻪ ﻓﺎﺯ ﺍﻧﺠﺎﻡ ﻣﻲ ﺷﻮﺩ‪:‬‬

‫‪ .1‬ﻓﺎﺯ ﺧﻮﺍﻧﺪﻥ ﻭ ﺍﺟﺮﺍ ‪ :‬ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﻓﻘﻂ ﻣﺘﻐﻴﺮﻫﺎﻱ ﻣﺤﻠﻲ ﺭﺍ ﺑﻪ ﻃﻮﺭ ﻣﻮﻗﺖ ﻣﻲ ﻧﻮﻳﺴﺪ‪.‬‬

‫‪ .2‬ﻓﺎﺯ ﺍﻋﺘﺒﺎﺭ ‪ :‬ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﺁﺯﻣﻮﻥ ﺍﻋﺘﺒﺎﺭ ﺑﺮﺍﻱ ﺗﻌﻴﻴﻦ ﺍﻳﻨﻜﻪ ﺁﻳﺎ ﻣﺘﻐﻴﺮﻫﺎﻱ ﻣﺤﻠﻲ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﺑﺪﻭﻥ ﻧﻘﺾ ﺳﺮﻳﺎﻟﻲ‬
‫ﻧﻮﺷﺘﻪ ﺷﻮﻧﺪ ﺭﺍ ﺍﻧﺠﺎﻡ ﻣﻲ ﺩﻫﺪ‪.‬‬

‫‪ .3‬ﻓﺎﺯ ﻧﻮﺷﺘﻦ ‪ :‬ﺍﮔﺮ ‪ Ti‬ﺩﺍﺭﺍﻱ ﺍﻋﺘﺒﺎﺭ ﺍﺳﺖ‪ ،‬ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ ﻫﺎ ﻛﺎﺭﺑﺮﺩﻱ ﺍﺯ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻫﺴﺘﻨﺪ ﺩﺭ ﻏﻴﺮ ﺍﻳﻦ ﺻﻮﺭﺕ ‪،‬‬
‫‪ Ti‬ﺑﻪ ﻋﻘﺐ ﺑﺮ ﻣﻲ ﮔﺮﺩﺩ‪.‬‬

‫‪ 3 ‬ﻓﺎﺯ ﺍﺯ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﺍﺟﺮﺍﻱ ﻫﻤﺰﻣﺎﻥ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﺑﻪ ﺻﻮﺭﺕ ﺟﺪﺍ ﺷﺪﻩ ﺑﺎﺷﻨﺪ‪ ،‬ﺍﻣﺎ ﻫﺮ ﺗﺮﺍﻛﻨﺸﻲ ﺑﺎﻳﺪ ﺍﺯ‬
‫ﻃﺮﻳﻖ ‪ 3‬ﻓﺎﺯ ﺩﺭ ﺩﺳﺘﻮﺭ ﺍﺟﺮﺍ ﺷﻮﺩ‪.‬‬

‫‪87‬‬
‫• ﺑﺮﺍﻱ ﺳﺎﺩﮔﻲ ﻓﺮﺽ ﻛﻨﻴﺪ ﻛﻪ ﻣﺮﺣﻠﻪ ﻱ ﺍﻋﺘﺒﺎﺭﺳﻨﺠﻲ ﻭ ﻧﻮﺷﺘﻦ ﺑﺎ ﻫﻢ ﺍﺯ ﻃﺮﻳﻖ ﺍﺗﻤﻴﻚ ﻭ ﺳﺮﻳﺎﻟﻲ ﺭﺥ‬
‫ﻣﻲ ﺩﻫﺪ‪.‬‬

‫‪ ‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ‪ ،‬ﺗﻨﻬﺎ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺍﻋﺘﺒﺎﺭ‪/‬ﻧﻮﺷﺘﻦ ﺩﺭ ﻳﻚ ﺯﻣﺎﻥ ﺍﺝﺭﺍ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫‪ ‬ﻫﻤﭽﻨﻴﻦ ﺑﻪ ﻋﻨﻮﺍﻥ ﻛﻨﺘﺮﻝ ﻫﻤﺰﻣﺎﻧﻲ ﺧﻮﺷﺒﻴﻨﺎﻧﻪ ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﺩ‪ ،‬ﺍﺯ ﺁﻧﺠﺎ ﻛﻪ ﺗﺮﺍﻛﻨﺶ ﺑﻪ ﻃﻮﺭ ﻛﺎﻣﻞ ﺑﻪ ﺍﻳﻦ‬
‫ﺍﻣﻴﺪ ﻛﻪ ﻫﻤﻪ ﺑﻪ ﺧﻮﺑﻲ ﺩﺭ ﻃﻮﻝ ﺍﻋﺘﺒﺎﺭ ﺍﺟﺮﺍ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪ ‬ﻫﺮ ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﺩﺍﺭﺍﻱ ‪ 3‬ﺑﺮﭼﺴﺐ ﺯﻣﺎﻧﻲ ﺍﺳﺖ‪.‬‬

‫• ﺷﺮﻭﻉ )‪ :(Ti‬ﺯﻣﺎﻧﻲ ﻛﻪ ﺍﺟﺮﺍﻱ ‪ Ti‬ﺷﺮﻭﻉ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﻙ ‪ Ti‬ﻭﺍﺭﺩ ﻓﺎﺯ ﺍﻋﺘﺒﺎﺭﺳﻨﺠﻲ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫• ﺍﻋﺘﺒﺎﺭ ﺳﻨﺠﻲ )‪ : (Ti‬ﺯﻣﺎﻧﻲ ﻩ‬

‫• ﭘﺎﻳﺎﻥ )‪ : (Ti‬ﺯﻣﺎﻧﻲ ﻛﻪ ﻓﺎﺯ ﻧﻮﺷﺘﻦ ‪ Ti‬ﺑﻪ ﭘﺎﻳﺎﻥ ﻣﻲ ﺭﺳﺪ‪.‬‬

‫‪ ‬ﺩﺳﺘﻮﺭ ﺳﺮﻳﺎﻟﻲ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﺩﺍﺩﻩ ﺷﺪﻩ ﺩﺭ ﺯﻣﺎﻥ ﺍﻋﺘﺒﺎﺭ ﺑﺮﺍﻱ ﺍﻓﺰﺍﻳﺶ ﻫﻤﺰﻣﺎﻧﻲ ﺗﻌﻴﻴﻦ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫• ﺑﻨﺎﺑﺮﺍﻳﻦ )‪ TS(Ti‬ﻣﻘﺪﺍﺭﻱ ﺍﺯ ﺍﻋﺘﺒﺎﺭﺳﻨﺠﻲ )‪ (Ti‬ﺭﺍ ﻣﻲ ﺩﻫﺪ‪.‬‬

‫‪ ‬ﺍﻳﻦ ﭘﺮﻭﺗﻜﻞ ﺑﺴﻴﺎﺭ ﻣﻔﻴﺪ ﺍﺳﺖ ﻭ ﻣﻴﺰﺍﻥ ﺑﺎﻻﻳﻲ ﺍﺯ ﻫﻤﺰﻣﺎﻧﻲ ﻛﻪ ﺍﺣﺘﻤﺎﻝ ﺩﺭﮔﻴﺮﻱ ﺁﻥ ﻛﻢ ﺍﺳﺖ ﺭﺍ ﻣﻲ ﺩﻫﺪ‪.‬‬

‫• ﺯﻳﺮﺍ ﺩﺳﺘﻮﺭ ﺳﺮﻳﺎﻟﻲ ﻗﺎﺑﻞ ﺗﺼﻤﻴﻢ ﮔﻴﺮﻱ ﻧﻴﺴﺖ‪ ،‬ﻭ‬

‫• ﺗﻌﺪﺍﺩ ﻛﻤﻲ ﺍﺯ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺑﺎﻳﺪ ﺑﻪ ﻋﻘﺐ ﺑﺮﮔﺮﺩﺩ‪.‬‬

‫ﺁﺯﻣﻮﻥ ﺍﻋﺘﺒﺎﺭ ﺑﺮﺍﻱ ﺗﺮﺍﻛﻨﺶ ‪Tj‬‬

‫‪ ‬ﺍﮔﺮ ‪ Ti‬ﺑﺮﺍﻱ ﻫﻤﻪ )‪ TS(Ti)<TS(Tj‬ﺑﺎﺷﺪ ﻳﻜﻲ ﺍﺯ ﺷﺮﺍﻳﻂ ﺯﻳﺮ ﺭﺍ ﻧﮕﻪ ﻣﻲ ﺩﺍﺭﺩ ‪:‬‬

‫• )‪finish(Ti) < start(Tj‬‬

‫• )‪ Start(Tj)<finish(Ti)<validation(Tj‬ﻭ ﻣﺠﻤﻮﻋﻪ ﺁﻳﺘﻢ ﻫﺎﻱ ﺩﺍﺩﻩ ﻱ ﻧﻮﺷﺘﻪ ﺷﺪﻩ ﺗﻮﺳﻂ‬
‫‪ Ti‬ﻛﻪ ﺑﺎ ﻣﺠﻤﻮﻋﻪ ﺁﻳﺘﻢ ﻫﺎﻱ ﺩﺍﺩﻩ ﺧﻮﺍﻧﺪﻩ ﺷﺪﻩ ﺗﻮﺳﻂ ‪ Tj‬ﻋﻤﻞ ﺗﻘﺴﻴﻢ ﻛﺮﺩﻥ ﺭﺍ ﺍﻧﺠﺎﻡ ﻧﻤﻲ ﺩﻫﻨﺪ‪.‬‬
‫ﺁﻧﮕﺎﻩ ﺍﻋﺘﺒﺎﺭﺳﻨﺠﻲ ﺑﺎ ﻣﻮﻓﻘﻴﺖ ﺍﻧﺠﺎﻡ ﻣﻲ ﺷﻮﺩ ﻭ ‪ Tj‬ﻣﻲ ﺗﻮﺍﻧﺪ ﭘﺬﻳﺮﻓﺘﻪ ﺷﻮﺩ‪ ،‬ﺩﺭﻏﻴﺮﺍﻳﻦ ﺻﻮﺭﺕ‬
‫ﺍﻋﺘﺒﺎﺭﺳﻨﺠﻲ ﺑﺎ ﺷﻜﺴﺖ ﻣﻮﺍﺟﻪ ﻣﻲ ﺷﻮﺩ ﻭ ‪ Tj‬ﺑﻲ ﻧﺘﻴﺠﻪ ﻣﻲ ﻣﺎﻧﺪ‪.‬‬

‫‪88‬‬
 ‫‪ ‬ﺗﻮﺟﻴﻪ ‪ :‬ﻳﻜﻲ ﺍﺯ ‪ 2‬ﺷﺮﻁ ﺍﻭﻝ ﺭﺿﺎﻳﺖ ﺑﺨﺶ ﺍﺳﺖ‪ ،‬ﻭ ﻫﻴﭻ ﺍﺟﺮﺍﻱ ﻫﻤﭙﻮﺷﺎﻧﻲ ﻭﺟﻮﺩ ﻧﺪﺍﺭﺩ‪ ،‬ﻳﺎ ﺍﻳﻨﻜﻪ ﺷﺮﻁ‬
‫ﺩﻭﻡ ﺭﺿﺎﻳﺖ ﺑﺨﺶ ﺍﺳﺖ ﻭ‬
‫‪ ‬ﻧﻮﺷﺘﻪ ﻫﺎﻱ ‪ Tj‬ﺑﺮﺭﻭﻱ ﺧﻮﺍﻧﺪﻥ ‪Ti‬ﺗﺄﺛﻴﺮ ﻧﻤﻲ ﮔﺬﺍﺭﻧﺪ ﺍﺯ ﺁﻧﺠﺎﻳﺌﻜﻪ ﺁﻧﻬﺎ ﺑﻌﺪ ﺍﺯ ﺍﺗﻤﺎﻡ ‪ Ti‬ﺭﺥ ﻣﻲ‬
‫ﺩﻫﻨﺪ ﻭ ﺁﻥ ﺭﺍ ﻣﻲ ﺧﻮﺍﻧﻨﺪ‪.‬‬
‫‪ ‬ﻧﻮﺷﺘﻪ ﻫﺎﻱ ‪Ti‬ﺑﺮﺭﻭﻱ ﺧﻮﺍﻧﺪﻥ ‪ Tj‬ﺗﺄﺛﻴﺮﻧﻤﻲ ﮔﺬﺍﺭﻧﺪ ﭼﻮﻥ ﻛﻪ ‪ Tj‬ﻫﺮ ﺁﻳﺘﻢ ﻧﻮﺷﺘﻪ ﺷﺪﻩ ﺗﻮﺳﻂ‬
‫‪ Ti‬ﺭﺍ ﻧﻤﻲ ﺧﻮﺍﻧﺪ‪.‬‬

‫ﺑﺮﻧﺎﻣﻪ ﺯﻣﺎﻧﺒﻨﺪﻱ ﺗﻮﻟﻴﺪ ﺷﺪﻩ ﺗﻮﺳﻂ ﺍﻋﺘﺒﺎﺭ ﺳﻨﺠﻲ‬

‫‪ ‬ﻣﺜﺎﻟﻲ ﺍﺯ ﺑﺮﻧﺎﻣﻪ ﺗﻮﻟﻴﺪ ﺷﺪﻩ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺍﻋﺘﺒﺎﺭﺳﻨﺠﻲ‬

‫‪T14‬‬ ‫‪T15‬‬

‫)‪read(B‬‬
‫)‪read(B‬‬
‫‪B:= B-50‬‬
‫)‪read(A‬‬
‫‪A:= A+50‬‬

‫)‪read(A‬‬
‫)‪(validate‬‬
‫)‪display (A+B‬‬

‫)‪(validate‬‬
‫)‪write (B‬‬
‫)‪write (A‬‬

‫ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﭼﻨﺪ ﻧﺴﺨﻪ ﺍﻱ‬

‫• ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﭼﻨﺪ ﻧﺴﺨﻪ ﺍﻱ ﻧﺴﺨﻪ ﻫﺎﻱ ﻗﺪﻳﻤﻲ ﺍﺯ ﺁﻳﺘﻢ ﺩﺍﺩﻩ ﺭﺍ ﺑﺮﺍﻱ ﺍﻓﺰﺍﻳﺶ ﻫﻤﺰﻣﺎﻧﻲ ﻧﮕﻪ ﻣﻲ‬
‫ﺩﺍﺭﻧﺪ‬

‫‪89‬‬
 ‫• ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﺑﺮ ﺣﺴﺐ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﭼﻨﺪ ﻧﺴﺨﻪ ﺍﻱ‬

‫• ﻗﻔﻞ ‪ 2‬ﻓﺎﺯ ﭼﻨﺪ ﻧﺴﺨﻪ ﺍﻱ‬

‫‪ ‬ﻫﺮ ﻳﻚ ﺍﺯ ﻧﺘﺎﻳﺞ ﻧﻮﺷﺘﻦ ﺑﻪ ﻃﻮﺭ ﻣﻮﻓﻘﻴﺖ ﺁﻣﻴﺰ ﺩﺭ ﺍﻳﺠﺎﺩ ﻳﻚ ﻧﺴﺨﻪ ﺟﺪﻳﺪ ﺍﺯ ﺁﻳﺘﻢ ﺩﺍﺩﻩ ﻧﻮﺷﺘﻪ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪ ‬ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻧﻲ ﺑﺮ ﻃﺒﻖ ﻧﺴﺨﻪ ﻫﺎﻱ ﺑﺮﭼﺴﺐ‬

‫‪ ‬ﻫﻨﮕﺎﻣﻲ ﻛﻪ ﻋﻤﻞ ﺧﻮﺍﻧﺪﻥ )‪ (Q‬ﺻﺎﺩﺭ ﻣﻲ ﺷﻮﺩ‪ ،‬ﻳﻚ ﻧﺴﺨﻪ ﻣﻨﺎﺳﺐ ﺍﺯ ‪ Q‬ﺭﺍ ﺑﺮ ﺍﺳﺎﺱ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ‬
‫ﺗﺮﺍﻛﻨﺶ ﺍﻧﺘﺨﺎﺏ ﻛﻨﻴﺪ ﻭ ﻣﻘﺪﺍﺭ ﺁﻥ ﺭﺍ ﺑﻪ ﻧﺴﺨﻪ ﻱ ﺍﻧﺘﺨﺎﺏ ﺷﺪﻩ ﺑﺮﮔﺮﺩﺍﻧﻴﺪ‪.‬‬

‫‪ Read ‬ﻫﺎ ﻫﺮﮔﺰ ﺑﻪ ﺍﻧﺘﻈﺎﺭ ﻳﻚ ﻧﺴﺨﻪ ﻣﻨﺎﺳﺐ ﺑﺮﺍﻱ ﺑﺎﺯﮔﺸﺖ ﻓﻮﺭﻱ ﻧﻴﺎﺯ ﻧﺪﺍﺭﻧﺪ‪.‬‬

‫ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﺑﺮ ﺍﺳﺎﺱ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ‪Multiversion‬‬

‫‪ ‬ﻫﺮ ﻛﺪﺍﻡ ﺍﺯ ﺁﻳﺘﻢ ﻫﺎﻱ ﺩﺍﺩﻩ ‪ Q‬ﻳﻚ ﺩﻧﺒﺎﻟﻪ ﺍﺯ ﻧﺴﺨﻪ ﻫﺎﻱ >‪ <Q1,Q2,…Qm‬ﺭﺍ ﺩﺍﺭﺩ‪ .‬ﻫﺮ ﻧﺴﺨﻪ ‪ Qk‬ﺷﺎﻣﻞ‬
‫‪ 3‬ﺯﻣﻴﻨﻪ ﺩﺍﺩﻩ ﺍﺳﺖ ‪:‬‬

‫‪ .1‬ﻣﻄﺎﻟﺐ ﻭ ﻣﺤﺘﻮﺍ ‪ -‬ﺍﺭﺯﺵ ﻧﺴﺨﻪ ‪Qk‬‬

‫‪ – W-timestamp(Qk) .2‬ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﺗﺮﺍﻛﻨﺸﻲ ﻛﻪ ﻧﺴﺨﻪ ‪ Qk‬ﺭﺍ ﺍﻳﺠﺎﺩ ﻣﻲ ﻛﻨﺪ)ﻧﻮﺷﺘﻦ(‬

‫‪ -- R-timestamp(Qk) .3‬ﺑﺰﺭﮔﺘﺮﻳﻦ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﺗﺮﺍﻛﻨﺸﻲ ﻛﻪ ﺑﺎ ﻣﻮﻓﻘﻴﺖ ﺑﻪ ﻋﻨﻮﺍﻥ ﻧﺴﺨﻪ‬

‫‪ Qk‬ﺧﻮﺍﻧﺪﻩ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬

‫‪ ‬ﻫﻨﮕﺎﻣﻲ ﻛﻪ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﻳﻚ ﻧﺴﺨﻪ ﺟﺪﻳﺪ ‪ Qk‬ﺍﺯ ‪ Q‬ﺭﺍ ﺍﻳﺠﺎﺩ ﻣﻲ ﻛﻨﺪ‪ Qks w-timestamp ،‬ﻭ ‪R-‬‬
‫‪ timestamp‬ﺑﺎ )‪ TS(Ti‬ﻣﻘﺪﺍﺭ ﺩﻫﻲ ﺍﻭﻟﻴﻪ ﺷﺪﻩ ﺍﻧﺪ‪.‬‬

‫‪ R-timestamp ‬ﺍﺯ ‪ Qk‬ﺑﻪ ﺭﻭﺯ ﺷﺪﻩ ﺍﺳﺖ ﻫﺮﺯﻣﺎﻥ ﻛﻪ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ‪ Tj‬ﺑﺨﻮﺍﻧﺪ ‪ ، Qk‬ﻭ ‪Ts(Tj)>R-‬‬
‫)‪timestamp(Qk‬‬

‫‪ ‬ﻓﺮﺽ ﻛﻨﻴﺪ ﻛﻪ ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﻋﻤﻠﻴﺎﺕ ﻣﺮﺑﻮﻁ ﺑﻪ )‪ read(Q‬ﻳﺎ )‪ write(Q‬ﺭﺍ ﺑﻔﺮﺳﺘﺪ‪ .‬ﺍﺟﺎﺯﻩ ﺩﻫﻴﺪ ﺑﻪ ‪Qk‬‬
‫ﻛﻪ ﺩﻻﻟﺖ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ ﺑﺮ ﻧﺴﺨﻪ ﺍﻱ ﺍﺯ ‪ Q‬ﻛﻪ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﻧﻮﺷﺘﻦ ﺁﻥ ﺑﺰﺭﮔﺘﺮﻳﻦ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﻧﻮﺷﺘﻦ‬
‫ﻛﻮﭼﻜﺘﺮ ﻳﺎ ﻣﺴﺎﻭﻱ )‪ TS(Ti‬ﺍﺳﺖ‪.‬‬

‫‪90‬‬
‫‪ .1‬ﺍﮔﺮ ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﻣﺴﺎﺋﻞ ﻣﺮﺑﻮﻁ ﺑﻪ )‪ read(Q‬ﺭﺍ ﻣﻄﺮﺡ ﻛﻨﺪ‪ ،‬ﺁﻧﮕﺎﻩ ﻣﻘﺪﺍﺭ ﺑﺮﮔﺸﺘﻲ‪ ،‬ﻣﺤﺘﻮﺍﻱ ﻧﺴﺨﻪ‬
‫‪ Qk‬ﺍﺳﺖ‪.‬‬

‫‪ .2‬ﺍﮔﺮ ﺗﺮﺍﻛﻨﺶ ‪Ti‬ﻣﺴﺎﺋﻞ ﻣﺮﺑﻮﻁ ﺑﻪ )‪ Write(Q‬ﺭﺍ ﻣﻄﺮﺡ ﻛﻨﺪ‪.‬‬

‫‪ .1‬ﺍﮔﺮ )‪ TS(Ti)<R-timestamp(Qk‬ﺑﺎﺷﺪ‪ ،‬ﺁﻥ ﮔﺎﻩ ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﺑﻪ ﻋﻘﺐ ﺑﺮ ﻣﻲ ﮔﺮﺩﺩ‪.‬‬

‫‪ .2‬ﺍﮔﺮ )‪ TS(Ti)=W-timestamp(Qk‬ﺑﺎﺷﺪ‪ ،‬ﻣﺤﺘﻮﻳﺎﺕ ‪ Qk‬ﺑﺎﻳﺪ ﺑﺎﺯﻧﻮﻳﺴﻲ ﺷﻮﺩ‪.‬‬

‫‪ .3‬ﺩﺭ ﻏﻴﺮ ﺍﻳﻦ ﺻﻮﺭﺕ ﻳﻚ ﻧﺴﺨﻪ ﺟﺪﻳﺪ ﺍﺯ ‪ Q‬ﺭﺍ ﺑﻪ ﻭﺟﻮﺩ ﺁﻭﺭﺩﻩ ﺍﺳﺖ‪.‬‬

‫‪ ‬ﻣﺸﺎﻫﺪﻩ ﻣﻲ ﻛﻨﻴﻢ ﻛﻪ‬

‫• ﺑﺎﺭ ﺧﻮﺍﻧﺪﻩ ﺷﺪﻩ ﻩ ﻣﻴﺸﻪ ﻣﻮﻓﻘﻴﺖ ﺁﻣﻴﺰ‬

‫• ﻳﻚ ﻧﻮﺷﺘﻪ ﺗﻮﺳﻂ ‪ Ti‬ﺭﺩ ﻣﻲ ﺷﻮﺩ ﺍﮔﺮ ﺗﺮﺍﻛﻨﺶ ‪ Tj‬ﺩﻳﮕﺮﻱ ) ﺩﺭ ﺩﺳﺘﻮﺭ ﺳﺮﻳﺎﻟﻲ ﺗﻮﺳﻂ‬
‫ﻣﻘﺎﺩﻳﺮ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﺗﻌﺮﻳﻒ ﻣﻲ ﺷﻮﻧﺪ( ﻧﻮﺷﺘﻪ ﻱ ‪ Tis‬ﺭﺍ ﺑﺨﻮﺍﻧﺪ‪ ،‬ﻛﻪ ﺩﺭ ﺣﺎﻝ ﺣﺎﺿﺮ ﻳﻚ‬
‫ﻧﺴﺨﻪ ﺍﻳﺠﺎﺩ ﺷﺪﻩ ﺗﻮﺳﻂ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﻗﺪﻳﻤﻲ ﺗﺮ ﺍﺯ ‪ Ti‬ﺭﺍ ﻣﻲ ﺧﻮﺍﻧﺪ‪.‬‬

‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﺗﻀﻤﻴﻦ ﻛﻨﻨﺪﻩ ﺳﺮﻳﺎﻟﻲ‬

‫ﻗﻔﻞ ‪ 2‬ﻓﺎﺯ ﭼﻨﺪ ﻧﺴﺨﻪ ﺍﻱ‬

‫‪ ‬ﺗﻔﺎﻭﺕ ﺑﻴﻦ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﻓﻘﻂ ﺧﻮﺍﻧﺪﻧﻲ ﻭ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﺑﻪ ﺭﻭﺯ ﺷﺪﻩ‬

‫‪ ‬ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﺑﻪ ﺭﻭﺯ ﺷﺪﻩ ﻗﻔﻞ ﻫﺎﻱ ﺧﻮﺍﻧﺪﻥ ﻭ ﻧﻮﺷﺘﻦ ﺭﺍ ﺑﻪ ﺩﺳﺖ ﻣﻲ ﺁﻭﺭﻧﺪ‪ ،‬ﻭ ﺗﻤﺎﻣﻲ ﻗﻔﻞ ﻫﺎ ﺭﺍ ﺗﺎ ﭘﺎﻳﺎﻥ‬
‫ﺗﺮﺍﻛﻨﺶ ﻧﮕﻪ ﻣﻲ ﺩﺍﺭﻧﺪ‪ .‬ﺗﺮﺍﻛﻨﺶ ﻫﺎﻳﻲ ﻛﻪ ﺑﻪ ﺩﻧﺐﺍﻝ ﻗﻔﻞ ‪ 2‬ﻓﺎﺯ ﻫﺴﺘﻨﺪ‪.‬‬

‫• ﻫﺮﻳﻚ ﺍﺯ ﻧﺘﺎﻳﺞ ﻧﻮﺷﺘﻦ ﻣﻮﻓﻘﻴﺖ ﺁﻣﻴﺰ ﺩﺭ ﺍﻳﺠﺎﺩ ﻳﻚ ﻧﺴﺨﻪ ﺟﺪﻳﺪ ﺍﺯ ﺁﻳﺘﻢ ﺩﺍﺩﻩ ﻧﻮﺷﺘﻪ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬

‫• ﻫﺮ ﻧﺴﺨﻪ ﺍﺯ ﺁﻳﺘﻢ ﺩﺍﺩﻩ ﺩﺍﺭﺍﻱ ﻳﻚ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﻭﺍﺣﺪ ﺍﺳﺖ ﻛﻪ ﺍﺭﺯﺵ ﺁﻥ ﺍﺯ ﻃﺮﻳﻖ ﺷﻤﺎﺭﻧﺪﻩ‬

‫‪ Ts-counter‬ﻛﻪ ﺩﺭ ﻃﻮﻝ ﺗﺄﻳﻴﺪ ﭘﺮﺩﺍﺯﺵ ﺍﻓﺰﺍﻳﺶ ﻣﻲ ﻳﺎﺑﺪ ﺑﻪ ﺩﺳﺖ ﻣﻲ ﺁﻳﺪ‪.‬‬

‫‪91‬‬
 ‫‪ ‬ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﻓﻘﻂ ﺧﻮﺍﻧﺪﻧﻲ ﺩﺭ ﻳﻚ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﺑﺎ ﺧﻮﺍﻧﺪﻥ ﺍﺭﺯﺵ ﻓﻌﻠﻲ ‪ ts-counter‬ﻗﺒﻞ ﺍﺯ ﺷﺮﻭﻉ‬
‫ﺍﺟﺮﺍ ﺗﻌﻴﻴﻦ ﻣﻲ ﺷﻮﻧﺪ‪ ،‬ﺁﻧﻬﺎ ﺑﻪ ﺩﻧﺒﺎﻝ ﭘﺮﻭﺗﻜﻞ ﺩﺳﺘﻮﺭ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﭼﻨﺪ ﻧﺴﺨﻪ ﺍﻱ ﺑﺮﺍﻱ ﺍﺟﺮﺍﻱ ﺧﻮﺍﻧﺪﻥ‬
‫ﻣﻲ ﺑﺎﺷﻨﺪ‪.‬‬

‫‪ ‬ﭼﻪ ﻭﻗﺖ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﻭﻳﺮﺍﻳﺶ ﺷﺪﻩ ﺗﻤﺎﻳﻞ ﺩﺍﺭﺩ ﻳﻚ ﺑﺨﺶ ﺍﺯ ﺩﺍﺩﻩ ﺭﺍ ﺑﺨﻮﺍﻧﺪ‪:‬‬

‫ﻳﻚ ﻗﻔﻞ ﻣﺸﺘﺮﻙ ﺭﻭﻱ ﺁﻥ ﺑﻪ ﺩﺳﺖ ﻣﻲ ﺁﻭﺭﺩ‪ ،‬ﻭ ﻧﺴﺨﻪ ﻫﺎﻱ ﭘﺎﻳﻴﻦ ﺗﺮ ﺭﺍ ﻣﻲ ﺧﻮﺍﻧﺪ‪.‬‬ ‫•‬

‫‪ ‬ﻭﻗﺘﻲ ﻛﻪ ﺗﻤﺎﻳﻞ ﺩﺍﺭﺩ ﻳﻚ ﺑﺨﺶ ﺭﺍ ﺑﻨﻮﻳﺴﺪ‬

‫ﻗﻔﻞ ‪ x‬ﺭﺍ ﺑﻪ ﺩﺳﺖ ﻣﻲ ﺁﻭﺭﺩ‪ ،‬ﺳﭙﺲ ﻳﻚ ﻧﺴﺨﻪ ﺟﺪﻳﺪﺍﺯ ﺍﻳﻦ ﺑﺨﺶ ﻭ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﻧﺴﺨﻪ ﻫﺎﻱ‬ ‫•‬
‫ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﺑﻪ ∞ ﺭﺍﺍﻳﺠﺎﺩ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫‪ ‬ﭼﻪ ﻭﻗﺖ ﻭﻳﺮﺍﻳﺶ ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﻛﺎﻣﻞ ﻣﻲ ﺷﻮﺩ‪ ،‬ﭘﺬﻳﺮﺵ ﭘﺮﺩﺍﺯﺵ ﺭﺥ ﻣﻲ ﺩﻫﺪ ‪:‬‬

‫• ﻣﺠﻤﻮﻋﻪ ﻫﺎﻱ ‪ Ti‬ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ ﺭﻭﻱ ﻧﺴﺨﻪ ﻫﺎﻱ ﺁﻥ ‪ ts-counter+1‬ﺭﺍ ﺍﻳﺠﺎﺩ ﻣﻲ ﻛﻦﻥﺩ ‪.‬‬

‫• ‪ ts-counter ،Ti‬ﺭﺍ ﺑﺎ ‪ 1‬ﺍﻓﺰﺍﻳﺶ ﻣﻲ ﺩﻫﺪ‬

‫‪ ‬ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﻓﻘﻂ ﺧﻮﺍﻧﺪﻧﻲ ﻛﻪ ﺑﻌﺪ ﺍﺯ ‪ Ti‬ﺷﺮﻭﻉ ﻣﻲ ﺷﻮﺩ ‪ ts-counter‬ﺭﺍ ﺑﺮﺍﻱ ﺩﻳﺪﻥ ﻣﻘﺎﺩﻳﺮ ﻭﻳﺮﺍﻳﺶ‬
‫ﺷﺪﻩ ﺗﻮﺳﻂ ‪ Ti‬ﺍﻓﺰﺍﻳﺶ ﻣﻲ ﺩﻫﺪ‬

‫‪ ‬ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﻓﻘﻂ ﺧﻮﺍﻧﺪﻧﻲ ﻛﻪ ﺷﺮﻭﻉ ﻣﻲ ﺷﻮﻧﺪ ﻗﺒﻞ ﺍﺯ ‪ Ti‬ﺑﺮﺍﻱ ﺍﻓﺰﺍﻳﺶ ‪ ts-counter‬ﺑﺮﺍﻱ ﺩﻳﺪﻥ‬
‫ﻣﻘﺎﺩﻳﺮ ﻗﺒﻞ ﺍﺯ ﻭﻳﺮﺍﻳﺶ ﻫﺎﻱ ‪Ti‬‬

‫‪ ‬ﺑﺮﻧﺎﻣﻪ ﻫﺎﻳﻲ ﻛﻪ ﺗﻨﻬﺎ ﺑﻪ ﺻﻮﺭﺕ ﺳﺮﻳﺎﻟﻲ ﺗﻮﻟﻴﺪ ﺷﺪﻩ ﺍﻧﺪ‪.‬‬

‫‪ : MVCC‬ﻣﺴﺎﺋﻞ ﻣﺮﺑﻮﻁ ﺑﻪ ﭘﻴﺎﺩﻩ ﺳﺎﺯﻱ‬

‫‪ ‬ﺍﻳﺠﺎﺩ ﻧﺴﺨﻪ ﻫﺎﻱ ﻣﺘﻌﺪﺩ ﺍﻓﺰﺍﻳﺶ ﻣﻨﺒﻊ ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ ﺩﺭ ﻫﻮﺍ‬

‫• ‪ Tuple‬ﻫﺎﻱ ﺍﺿﺎﻓﻲ‬
‫• ﻓﻀﺎﻱ ﺍﺿﺎﻓﻲ ﺩﺭ ﻫﺮ ‪ tuple‬ﺑﺮﺍﻱ ﺫﺧﻴﺮﻩ ﻛﺮﺩﻥ ﻣﺘﻦ ﺍﻃﻼﻋﺎﺕ‬
‫‪ ‬ﺗﻮﺍﻧﺎﻳﻲ ﻧﺴﺨﻪ ﻫﺎ‪ ،‬ﻫﺮ ﭼﻨﺪ‪ ،‬ﮔﺮﺩﺁﻭﺭﻱ ﭘﺲ ﻣﺎﻧﺪﻩ‬

‫‪92‬‬
 ‫• ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ‪ .‬ﺍﮔﺮ ‪ 2 Q‬ﻧﺴﺨﻪ ﻱ ‪ Q5‬ﻭ ‪ Q9‬ﺭﺍ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ‪ ،‬ﻭ ﺗﺮﺍﻛﻨﺶ ﻓﻌﺎﻝ ﻗﺪﻳﻤﻲ‬
‫ﺗﺮ ﻛﻪ ﺑﺮﭼﺴﺐ ﺯﻣﺎﻥ <‪ 9‬ﺑﺎﺷﺪ ‪ ،‬ﻧﺴﺒﺖ ﺑﻪ ‪ Q5‬ﻛﻪ ﻫﺮﮔﺰ ﺩﻭﺑﺎﺭﻩ ﺑﻪ ﺩﺳﺖ ﻧﻤﻲ ﺁﻳﺪ‪.‬‬

‫ﻋﻤﻠﻴﺎﺕ ﺩﺭﺝ ﻭ ﺣﺬﻑ‬

‫‪ ‬ﺍﮔﺮ ﻗﻔﻞ ‪ 2‬ﻓﺎﺯ ﺍﺳﺘﻔﺎﺩﻩ ﺷﻮﺩ ‪:‬‬

‫• ﻳﻚ ﻋﻤﻠﻴﺎﺕ ﺣﺬﻑ ﻣﻤﻜﻦ ﺍﻧﺠﺎﻡ ﺷﻮﺩ ﺗﻨﻬﺎ ﺍﮔﺮ ﺗﺮﺍﻛﻨﺶ ﺣﺬﻑ ﻛﻨﺪ ﭼﻨﺪ ﺗﺎﻳﻲ ﺭﺍ ﻛﻪ ﻳﻚ ﻗﻔﻞ‬
‫ﺍﻧﺤﺼﺎﺭﻱ ﺩﺍﺭﺩ ﺭﻭﻱ ﭼﻨﺪﺗﺎﻳﻲ ﻛﻪ ﺣﺬﻑ ﺷﺪﻩ ﺍﻧﺪ‪.‬‬

‫• ﻳﻚ ﺗﺮﺍﻛﻨﺸﻲ ﻛﻪ ﻳﻚ ﭼﻨﺪﺗﺎﻳﻲ ﺟﺪﻳﺪ ﺭﺍ ﺩﺍﺧﻞ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺩﺭﺝ ﻣﻲ ﻛﻨﺪ ﻛﻪ ﻳﻚ ﻗﻔﻞ ‪x-mode‬‬
‫ﺭﺍ ﺭﻭﻱ ﭼﻨﺪﺗﺎﻳﻲ ﻛﻪ ﻫﺴﺖ ﻣﻲ ﺩﻫﺪ‪.‬‬

‫‪ ‬ﺩﺭﺝ ﻭ ﺣﺬﻑ ﻣﻲ ﺗﻮﺍﻧﺪ ﻣﻨﺠﺮ ﺑﻪ ﭘﺪﻳﺪﻩ ﻱ ﻓﺎﻧﺘﻮﻡ ﺷﻮﺩ‪.‬‬

‫• ﻳﻚ ﺗﺮﺍﻛﻨﺸﻲ ﻛﻪ ﻳﻚ ﺭﺍﺑﻄﻪ ﺭﺍ ﺑﻪ ﻃﻮﺭ ﺍﺟﻤﺎﻟﻲ ﺑﺮﺭﺳﻲ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫‪) ‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ‪ .‬ﻳﺎﻓﺘﻦ ﻣﻘﺪﺍﺭ ﻣﺎﻧﺪﻩ ﻫﺎﻱ ﻫﻤﻪ ﻱ ﻣﺤﺎﺳﺒﺎﺕ ﺩﺭ ‪ ( perrgridge‬ﻭ ﻳﻚ‬
‫ﺗﺮﺍﻛﻨﺸﻲ ﻛﻪ ﻳﻚ ﭼﻨﺪﺗﺎﻱﻱ ﺭﺍ ﺩﺭ ﺭﺍﺑﻄﻪ ﺩﺭﺝ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫‪) ‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ‪ .‬ﺩﺭﺝ ﻳﻚ ﻣﺤﺎﺳﺒﻪ ﺟﺪﻳﺪ ﺩﺭ ‪) (Perryridge‬ﺗﺼﻮﺭﻱ( ﻧﺎﺳﺎﺯﮔﺎﺭﻱ ﻋﻠﻲ‬
‫ﺭﻏﻢ ﻫﺮ ﭼﻨﺪﺗﺎﻳﻲ ﻣﺸﺘﺮﻙﻱ ﻛﻪ ﺩﺭ ﺩﺳﺘﺮﺱ ﻧﻴﺴﺖ‪.‬‬

‫• ﺍﮔﺮ ﺕﻧﻬﺎ ﭼﻨﺪ ﻗﻔﻞ ﺍﺳﺘﻔﺎﺩﻩ ﺷﻮﺩ‪ ،‬ﻧﺘﻴﺠﻪ ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﻏﻴﺮﺳﺮﻳﺎﻟﻲ ﺑﺎﺷﺪ ‪.‬‬

‫‪ ‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ‪ .‬ﺑﺮﺭﺳﻲ ﺍﺟﻤﺎﻟﻲ ﺗﺮﺍﻛﻨﺶ ﻣﺤﺎﺳﺒﺎﺕ ﺟﺪﻳﺪ ﺭﺍ ﻧﺸﺎﻥ ﻧﻤﻲ ﺩﻫﺪ‪ ،‬ﺍﻣﺎ ﭼﻨﺪ ﺗﺎ‬
‫ﺍﺯ ﻧﻮﺷﺘﻪ ﻫﺎﻳﻲ ﻛﻪ ﺗﻮﺳﻂ ﺗﺮﺍﻛﻨﺶ ﻭﻳﺮﺍﻳﺶ ﺷﺪﻩ ﺭﺍ ﻣﻲ ﺧﻮﺍﻧﺪ‪.‬‬

‫ﻙ ﻧﺸﺎﻥ ﻣﻲ ﺩﻫﻨﺪ ﭼﻨﺪ ﺗﺎ ﺭﺍﺑﻄﻪ ﺭﺍ‬

‫‪ ‬ﺗﺮﺍﻛﻨﺶ ﺑﻪ ﻃﻮﺭ ﺍﺟﻤﺎﻟﻲ ﺑﺮﺭﺳﻲ ﻣﻲ ﻛﻨﺪ ﺭﺍﺑﻄﻪ ﺧﻮﺍﻧﺪﻥ ﺍﻃﻼﻋﺎﺗﻲ ﺭﺍ ﻩ‬
‫ﺷﺎﻣﻞ ﻣﻲ ﺷﻮﻥﺩ‪ ،‬ﺩﺭ ﺣﺎﻟﻲ ﻛﻪ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﻳﻚ ﭼﻨﺪﺗﺎﻳﻲ ﺭﺍ ﺩﺭﺝ ﻣﻲ ﻛﻨﺪ ﻛﻪ ﻫﻤﺎﻥ ﺍﻃﻼﻋﺎﺕ ﺭﺍ ﻭﻳﺮﺍﻳﺶ‬
‫ﻣﻲ ﻛﻨﺪ‪.‬‬

‫ﺍﻃﻼﻋﺎﺕ ﺑﺎﻳﺪ ﻗﻔﻞ ﺷﻮﺩ‪.‬‬ ‫•‬

‫‪93‬‬
 ‫‪ ‬ﺭﺍﻩ ﺣﻞ ﺍﻭﻝ ‪:‬‬

‫• ﺍﺭﺗﺒﺎﻁ ﺩﺍﺩﻥ ﻳﻚ ﺑﺨﺶ ﺩﺍﺩﻩ ﺑﺎ ﺧﻮﻳﺶ‪ ،‬ﺑﺎﻧﻤﺎﻳﺶ ﺩﺍﺩﻥ ﺍﻃﻼﻋﺎﺗﻲ ﺩﺭ ﻣﻮﺭﺩ ﺍﻳﻦ ﻛﻪ ﭼﻪ ﭼﻨﺪﺗﺎﻳﻲ‬
‫ﻫﺎﻳﻲ ﺭﺍ ﺍﺯ ﺭﺍﺑﻄﻪ ﺷﺎﻣﻞ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫• ﺗﺮﺍﻛﻨﺶ ﻫﺎﻳﻲ ﻛﻪ ﺑﻪ ﻃﻮﺭ ﺍﺟﻤﺎﻟﻲ ﺭﺍﺑﻄﻪ ﺣﺎﺻﻞ ﺍﺯ ﻳﻚ ﻗﺴﻤﺖ ﺍﺯ ﻗﻔﻞ ﺭﺍ ﺩﺭ ﺑﺨﺸﻲ ﺍﺯ ﺩﺍﺩﻩ‬
‫ﺑﺮﺭﺳﻲ ﻣﻲ ﻛﻨﻨﺪ‪.‬‬

‫• ﺗﺮﺍﻛﻨﺶ ﻫﺎﻳﻲ ﻛﻪ ﺩﺭﺝ ﻳﺎ ﺣﺬﻑ ﻣﻲ ﻛﻨﻨﺪ ﻳﻚ ﭼﻨﺪﺗﺎﻳﻲ ﺭﺍ ﺑﺮﺍﻱ ﻳﻚ ﻗﻔﻞ ﺍﻧﺤﺼﺎﺭﻱ ﺑﺮﺭﻭﻱ‬
‫ﺑﺨﺸﻲ ﺍﺯ ﺩﺍﺩﻩ ﺑﻪ ﺩﺳﺖ ﻣﻲ ﺁﻭﺭﻧﺪ)ﺗﻮﺟﻪ‪ :‬ﻗﻔﻞ ﻫﺎﻱ ﺑﺮﺭﻭﻱ ﺑﺨﺸﻲ ﺍﺯ ﺩﺍﺩﻩ‪ ،‬ﻧﺎﺳﺎﺯﮔﺎﺭﻱ ﺑﺎ ﻗﻔﻞ ﻫﺎ‬
‫ﺑﺮﺭﻭﻱ ﭼﻨﺪﺗﺎﻳﻲ ﺍﺧﺘﺼﺎﺻﻲ ﺭﺍ ﺍﻧﺠﺎﻡ ﻧﻤﻲ ﺩﻫﻨﺪ‪(.‬‬

‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﺑﺎﻻ ﻫﻤﺎﻥ ﻫﻢ ﺯﻣﺎﻧﻲ ﻛﻢ ﺭﺍ ﺑﺮﺍﻱ ﺩﺭﺝ ﻫﺎ‪/‬ﺣﺬﻑ ﻫﺎ ﻓﺮﺍﻫﻢ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﺷﺎﺧﺺ ﻗﻔﻞ ﺑﻴﺸﺘﺮﻳﻦ ﻫﻤﺰﻣﺎﻧﻲ ﺭﺍ ﻓﺮﺍﻫﻢ ﻣﻲ ﻛﻨﺪ ﺩﺭﺣﺎﻟﻲ ﻛﻪ ﻣﺎﻧﻊ ﭘﺪﻳﺪﻩ ﻱ ﻓﺎﻧﺘﻮﻡ ﻣﻲ ﺷﻮﺩ‪،‬‬
‫ﺑﻮﺍﺳﻄﻪ ﻗﻔﻞ ﻫﺎﻳﻲ ﻛﻪ ﺑﻪ ﺳﻄﻞ ﻫﺎﻱ ﺷﺎﺧﺺ ﻣﻌﻴﻦ ﻧﻴﺎﺯ ﺩﺍﺭﻧﺪ‪.‬‬

‫ﭘﺮﻭﺗﻜﻞ ﻗﻔﻞ ﻓﻬﺮﺳﺖ‬

‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﻗﻔﻞ ﺷﺎﺧﺺ ‪:‬‬

‫• ﻫﻤﻪ ﺭﻭﺍﺑﻂ ﺑﺎﻳﺪ ﺩﺳﺖ ﻛﻢ ﻳﻚ ﺷﺎﺧﺺ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻨﺪ‪.‬‬

‫• ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﻣﻲ ﺗﻮﺍﻧﺪ ﻳﻚ ﭼﻨﺪ ﺗﺎﻳﻲ ﺭﺍ ﺑﻪ ﺩﺳﺖ ﺁﻭﺭﺩ ﺗﻨﻬﺎ ﺑﻌﺪ ﺍﺯ ﺍﻳﻨﻜﻪ ﺁﻧﻬﺎ ﺭﺍ ﺍﺯ ﻃﺮﻳﻖ ﻳﻚ‬
‫ﺷﺎﺧﺺ ﻳﺎ ﺑﻴﺸﺘﺮ ﺭﻭﻱ ﺭﻭﺍﺑﻂ ﭘﻴﺪﺍ ﻛﻨﺪ‪.‬‬

‫• ﻳﻚ ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﻳﻚ ﻣﺮﺍﺟﻌﻪ ﺍﻱ ﺭﺍ ﻛﻪ ﺑﺎﻳﺪ ﺗﻤﺎﻡ ﮔﺮﻩ ﻫﺎﻱ ﺑﺮگ ﺷﺎﺧﺺﻱ ﺭﺍ ﻛﻪ ﺑﻪ ﺁﻥ ﺩﺳﺘﺮﺳﻲ‬
‫ﺩﺍﺭﻧﺪ ﺭﺍ ﻗﻔﻞ ﻛﻨﺪ‪ ،‬ﺩﺭ ‪ s-mode‬ﻧﻤﺎﻳﺶ ﻣﻲ ﺩﻫﺪ‪.‬‬

‫‪ ‬ﻭﻟﻮﺁﻧﻜﻪ ﮔﺮﻩ ﺑﺮگ ﺷﺎﻣﻞ ﻫﺮﭼﻨﺪﺗﺎ ﻣﺮﺍﺟﻌﻪ ﺷﺎﺧﺺ ﻗﺎﻧﻊ ﻛﻨﻨﺪﻩ ﻧﺒﺎﺷﺪ‪).‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ‪.‬‬
‫ﺑﺮﺍﻱ ﻳﻚ ﻧﺎﺣﻴﻪ ﭘﺮﺱ ﻭ ﺟﻮ‪ ،‬ﺩﺭﺁﻥ ﻧﺎﺣﻴﻪ ﺑﻴﺸﺘﺮ ﺍﺯ ﭼﻨﺪ ﺗﺎ ﺑﺮگ ﻧﻴﺴﺖ‪(.‬‬

‫• ﻳﻚ ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﻛﻪ ﺩﺭﺝ ﻣﻲ ﻛﻨﺪ‪ ،‬ﻭﻳﺮﺍﻳﺶ ﻳﺎ ﺣﺬﻑ ﻣﻲ ﻛﻨﺪ ﭼﻨﺪ ﺗﺎ ‪ ti‬ﺭﺍﺩﺭ ﻳﻚ ﺭﺍﺑﻄﻪ ‪.r‬‬

‫‪94‬‬
 ‫‪ ‬ﺑﺎﻳﺪ ﺗﻤﺎﻡ ﺷﺎﺧﺺ ﻫﺎ ﺭﺍ ﺗﺎ ‪ r‬ﻭﻳﺮﺍﻳﺶ ﻛﻨﺪ‪.‬‬

‫‪ ‬ﺑﺎﻳﺪ ﻗﻔﻞ ﻫﺎﻱ ﺍﻧﺤﺼﺎﺭﻱ ﺭﺍ ﺭﻭﻱ ﺗﻤﺎﻡ ﮔﺮﻩ ﻫﺎﻱ ﺑﺮگ ﺷﺎﺧﺺ ﺳﺎﺧﺘﮕﻲ ﺗﻮﺳﻂ‬
‫ﺩﺭﺝ‪/‬ﻭﻳﺮﺍﻳﺶ‪/‬ﺣﺬﻑ ﺑﻪ ﺩﺳﺖ ﺁﻭﺭﺩ‪.‬‬

‫• ﻗﻮﺍﻋﺪﻱ ﺍﺯ ﭘﺮﻭﺗﻜﻞ ﻗﻔﻞ ‪ 2‬ﻓﺎﺯ ﺑﺎﻳﺪ ﺭﻋﺎﻳﺖ ﺷﻮﺩ‪.‬‬

‫‪ ‬ﺗﻀﻤﻴﻦ ﻣﻲ ﻛﻨﺪ ﻛﻪ ﭘﺪﻳﺪﻩ ﻱ ﻓﺎﻧﺘﻮﻡ ﺩﺭ ﺁﻥ ﺍﺗﻔﺎﻕ ﻧﻤﻲ ﺍﻓﺘﺪ‪.‬‬

‫ﺳﻄﻮﺡ ﺿﻌﻴﻒ ﺳﺎﺯگﺍﺭﻱ‬

‫‪ ‬ﺳﺎﺯﮔﺎﺭﻱ ﺩﺭﺟﻪ ‪ : 2‬ﺍﺧﺘﻼﻑ ﺍﺯ ﻗﻔﻞ ‪ 2‬ﻓﺎﺯ ﺩﺭ ﺍﻳﻦ ﻛﻪ ‪ s-lock‬ﻣﻤﻜﻦ ﺩﺭ ﻫﺮ ﺯﻣﺎﻥ ﺁﺯﺍﺩ ﺷﻮﺩ‪ ،‬ﻭ ﻗﻔﻞ‬
‫ﻫﺎﻳﻲ ﻛﻪ ﻣﻤﻜﻦ ﺩﺭ ﻫﺮ ﺯﻣﺎﻥ ﺑﻪ ﺩﺳﺖ ﺑﻴﺎﻳﻨﺪ‪.‬‬

‫• ‪ X-lock‬ﻫﺎ ﺑﺎﻳﺪ ﭘﺎﻳﺎﻥ ﺗﺮﺍﻛﻨﺶ ﺭﺍ ﻧﮕﻪ ﺩﺍﺭﻧﺪ‪.‬‬

‫• ‪ Serializability‬ﺗﻀﻤﻴﻦ ﺷﺪﻩ ﻧﻴﺴﺖ‪ ،‬ﺑﺮﻧﺎﻣﻪ ﻧﻮﻳﺲ ﺑﺎﻳﺪ ﻣﻄﻤﺌﻦ ﺑﺎﺷﺪ ﻛﻪ ﻫﻴﭻ ﻭﻗﺖ ﺍﺷﺘﺒﺎﻫﻲ‬
‫ﺍﺯ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺭﺥ ﻧﻤﻲ ﺩﻫﺪ‪.‬‬

‫‪ ‬ﻣﻜﺎﻥ ﻧﻤﺎﻱ ﭘﺎﻳﺪﺍﺭ ‪:‬‬

‫• ﺑﺮﺍﻱ ﻣﺸﺎﻫﺪﻩ‪ ،‬ﻫﺮ ﺗﺎﭘﻠﻲ ﻛﻪ ﻗﻔﻞ ﺷﺪﻩ ﺍﺳﺖ‪ ،‬ﻣﻲ ﺧﻮﺍﻧﺪ ﻭ ﻓﻮﺭﺍً ﺁﻥ ﻗﻔﻞ ﺭﺍ ﺁﺯﺍﺩ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫• ‪ X-lock‬ﻫﺎ ﺗﺎ ﭘﺎﻳﺎﻥ ﺗﺮﺍﻛﻨﺶ ﻧﮕﻪ ﺩﺍﺷﺘﻪ ﻣﻲ ﺷﻮﻧﺪ‪.‬‬

‫• ﺣﺎﻟﺖ ﺧﺎﺻﻲ ﺍﺯ ﺳﺎﺯﮔﺎﺭﻱ ﺩﺭﺟﻪ ‪2‬‬

‫ﺳﻄﻮﺡ ﺿﻌﻴﻒ ﺳﺎﺯﮔﺎﺭﻱ ﺩﺭ ‪SQL‬‬

‫‪ SQL ‬ﺍﺟﺎﺯﻩ ﻣﻲ ﺩﻫﺪ ﺗﺎ ‪ non-serializable‬ﺍﺟﺮﺍ ﺷﻮﺩ ‪.‬‬

‫• ‪ : Serializable‬ﭘﻴﺶ ﻓﺮﺽ ﺍﺳﺖ‪.‬‬

‫• ﺧﻮﺍﻧﺪﻥ ﺗﻜﺮﺍﺭﭘﺬﻳﺮ ‪ :‬ﺍﺟﺎﺯﻩ ﻣﻲ ﺩﻫﺪ ﺗﺎ ﺳﻮﺍﺑﻖ ﺗﻨﻬﺎ ﺧﻮﺍﻧﺪﻥ ﺭﺍ ﻗﺒﻮﻝ ﻛﻨﻨﺪ‪ ،‬ﻭ ﺗﻜﺮﺍﺭ ﻛﻨﻨﺪ ﻳﻚ‬
‫ﺧﻮﺍﻧﺪﻥﻱ ﻛﻪ ﺑﺎﻳﺪ ﺑﻪ ﻫﻤﺎﻥ ﻣﻘﺪﺍﺭ ﺑﺮﮔﺮﺩﺩ)ﺑﻨﺎﺑﺮﺍﻳﻦ ﺧﻮﺍﻧﺪﻥ ﻗﻔﻞ ﻫﺎ ﺑﺎﻳﺪ ﺣﻒﻅ ﺷﻮﺩ (‬

‫‪95‬‬
 ‫‪ ‬ﺑﺎ ﺍﻳﻦ ﺣﺎﻝ‪ ،‬ﭘﺪﻳﺪﻩ ﻱ ﻓﺎﻧﺘﻮﻡ ﻧﻴﺎﺯ ﺩﺍﺭﺩ ﻛﻪ ﻣﺎﻧﻊ ﺁﻥ ﻧﺸﻮﺩ‪.‬‬

‫– ‪ T1‬ﻣﻤﻜﻦ ﺑﺮﺧﻲ ﺍﺯ ﺳﻮﺍﺑﻖ ﻭﺍﺭﺩ ﺷﺪﻩ ﺗﻮﺳﻂ ‪ T2‬ﺭﺍ ﺑﺒﻴﻨﺪ‪ ،‬ﺍﻣﺎ ﺍﻧﻮﺍﻉ ﺩﻳﮕﺮﻱ ﻛﻪ‬
‫ﺗﻮﺳﻂ ‪ T2‬ﻭﺍﺭﺩ ﺷﺪﻧﺪ ﺭﺍ ﻧﺒﻴﻨﺪ‪.‬‬

‫• ﺧﻮﺍﻧﺪﻩ ﺷﺪﻩ ‪ :‬ﻫﻤﺎﻥ ﺩﺭﺟﻪ‪ 2‬ﺳﺎﺯﮔﺎﺭﻱ‪ ،‬ﺍﻣﺎ ﺳﻴﺴﺘﻢ ﻫﺎ ﺑﺎﻳﺪ ﺁﻥ ﺭﺍ ﺑﻪ ﻋﻨﻮﺍﻥ ﻳﻚ ﻣﻜﺎﻥ ﻧﻤﺎﻱ‬
‫ﭘﺎﻳﺪﺍﺭ ﭘﻴﺎﺩﻩ ﺳﺎﺯﻱ ﻛﻨﻨﺪ‪.‬‬

‫• ﺧﻮﺍﻧﺪﻩ ﻧﺸﺪﻩ ‪ :‬ﺍﺟﺎﺯﻩ ﻣﻲ ﺩﻫﺪ ﺣﺘﻲ ﺩﺍﺩﻩ ﻏﻴﺮ ﻗﺎﺑﻞ ﻗﺒﻮﻝ ﺧﻮﺍﻧﺪﻩ ﺷﻮﺩ‪.‬‬

‫‪ ‬ﺩﺭ ﺑﺴﻴﺎﺭﻱ ﺍﺯ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ‪ :‬ﺳﻄﺢ ﺛﺒﺎﺕ ﭘﻴﺶ ﻓﺮﺽ ﺭﺍ ﺑﻪ ﻃﻮﺭ ﻗﺎﺑﻞ ﻗﺒﻮﻝ ﻣﻲ ﺧﻮﺍﻧﺪ‪.‬‬

‫ﺗﻐﻴﻲ ﻣﻲ ﺩﻫﺪ ﺯﻣﺎﻧﻲ ﻛﻪ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﺍﺳﺖ‪.‬‬

‫• ﺑﻪ ﺻﺮﺍﺣﺖ ‪ serializable‬ﺭﺍ ﺭ‬

‫‪ ‬ﻣﺠﻤﻮﻋﻪ ﺟﺪﺍ ﺩﺭ ﺳﻄﺢ ‪serializable‬‬

‫ﻫﻤﺰﻣﺎﻧﻲ ﺩﺭ ﺳﺎﺧﺘﺎﺭﻫﺎﻱ ﻓﻬﺮﺳﺖ‬

‫ﺩﻳﮓ ﻣﻮﺍﺭﺩ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻛﻪ ﺗﻨﻬﺎ ﻛﺎﺭﺵ ﺑﺮﺍﻱ ﻛﻤﻚ ﺑﻪ ﺩﺳﺘﺮﺳﻲ ﺑﻪ ﺩﺍﺩﻩ ﻫﺎ ﻣﻲ ﺑﺎﺷﺪ‪.‬‬
‫‪ ‬ﺷﺎﺧﺺ ﺑﺮ ﺧﻼﻑ ﺭ‬

‫‪ ‬ﺷﺎﺧﺺ ﺳﺎﺧﺘﺎﺭﻫﺎ ﻣﻌﻤﻮﻻ ﺍﻏﻠﺐ ﺩﺭ ﺩﺳﺘﺮﺱ‪ ،‬ﺧﻴﻠﻲ ﺑﻴﺸﺘﺮ ﺍﺯ ﺩﻳﮕﺮ ﻣﻮﺍﺭﺩ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺍﺳﺖ‪.‬‬

‫• ﺩﺭﻣﺎﻥ ﺷﺎﺧﺺ ﺳﺎﺧﺘﺎﺭﻫﺎﻳﻲ ﻣﺎﻧﻨﺪ ﺩﻳﮕﺮ ﻡﻭﺍﺭﺩ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ‪ ،‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ ﻓﺎﺯ‪ 2‬ﻗﻔﻞ ﮔﺮﻩ ﻫﺎﻱ‬
‫ﺷﺎﺧﺺ ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﻪ ﻫﻤﺰﻣﺎﻧﻲ ﻛﻢ ﻣﻨﺠﺮ ﺷﻮﺩ‪.‬‬

‫‪ ‬ﭼﻨﺪﻳﻦ ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﻫﻤﺰﻣﺎﻧﻲ ﺍﻭﻝ ﻛﻪ ﺩﺭ ﺁﻥ ﻕﻓﻞ ﺑﺮ ﺭﻭﻱ ﮔﺮﻩ ﻫﺎﻱ ﺩﺍﺧﻠﻲ ﺩﺭ ﺍﻭﻟﻲ ﺑﻪ ﺻﻮﺭﺕ ﺁﺯﺍﺩ ﻭﺟﻮﺩ‬
‫ﺩﺍﺭﺩ‪ ،‬ﻭ ﻧﻪ ﺩ ﺭﻣﺪ ﺩﻭ ﻓﺎﺯ ﺍﺳﺖ‪.‬‬

‫• ﺁﻥ ﻗﺎﺑﻞ ﻗﺒﻮﻝ ﺍﺳﺖ ﺑﺮﺍﻱ ﺩﺳﺘﺮﺳﻲ ﻫﻤﺰﻣﺎﻥ ‪ nonserializable‬ﺑﻪ ﻳﻚ ﺷﺎﺧﺺ ﻃﻮﻻﻧﻲ ﻛﻪ‬
‫ﺩﻗﺖ ﺷﺎﺧﺺ ﺣﻔﻆ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬

‫‪ ‬ﺑﻪ ﻃﻮﺭ ﺧﺎﺹ ‪ ،‬ﺧﻮﺍﻧﺪﻥ ﻣﻘﺎﺩﻳﺮ ﺩﻗﻴﻖ ﺁﻥ ﺩﺭ ﻳﻚ ﮔﺮﻩ ﺩﺍﺧﻠﻲ ﺍﺯ ﻳﻚ ﺩﺭﺧﺖ ‪ B+‬ﺑﻲ ﺭﺑﻂ‬
‫ﺯﻣﻴﻦ ﺭﺍ ﺩﺭ ﮔﺮﻩ ﺑﺮگ ﺩﺭﺳﺖ ﻛﻨﻴﻢ‪.‬‬
‫ﻫﺴﺘﻨﺪ ﺗﺎ ﺯﻣﺎﻧﻲ ﻛﻪ ﻣﺎ ﻩ‬

‫‪ ‬ﻣﺜﺎﻟﻲ ﺍﺯ ﭘﺮﻭﺗﻜﻞ ﻫﻤﺰﻣﺎﻧﻲ ﻓﻬﺮﺳﺖ ‪:‬‬

‫‪96‬‬
 ‫‪ ‬ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ‪ crabbing‬ﺑﻪ ﺟﺎﻱ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻗﻔﻞ ‪ 2‬ﻓﺎﺯ ﺑﺮ ﺭﻭﻱ ﮔﺮﻩ ﻫﺎﻱ ‪ B+‬ﺩﺭﺧﺖ ‪،‬ﺩﺭ ﻃﻮﻝ‬
‫ﺟﺴﺘﺠﻮ‪/‬ﺩﺭﺝ‪/‬ﺣﺬﻑ ﺑﻪ ﺷﺮﺡ ﺯﻳﺮ ﺍﺳﺖ ‪:‬‬

‫• ﺑﺮﺍﻱ ﺍﻭﻟﻴﻦ ﺑﺎﺭ ﮔﺮﻩ ﺭﻳﺸﻪ ﺩﺭ ﺣﺎﻟﺖ ﻣﺸﺘﺮﻙ ﺭﺍ ﻗﻔﻞ ﻛﻨﻴﺪ‪.‬‬

‫• ﭘﺲ ﺍﺯ ﻗﻔﻞ ﻛﺮﺩﻥ ﺗﻤﺎﻡ ﻓﺮﺯﻧﺪﺍﻥ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﻳﻚ ﮔﺮﻩ ﺩﺭ ﺣﺎﻟﺖ ﻣﺸﺘﺮﻙ ‪ ،‬ﺍﻧﺘﺸﺎﺭ ﻗﻔﻞ ﺑﺮ ﺭﻭﻱ ﮔﺮﻩ‬

‫• ﺩﺭ ﻃﻮﻝ ﺩﺭﺝ ‪ /‬ﺣﺬﻑ‪ ،‬ﺍﺭﺗﻘﺎء ﻗﻒﻝ ﻫﺎﻱ ﮔﺮﻩ ﺑﺮگ ﺑﻪ ﺣﺎﻟﺖ ﻫﺎﻱ ﻣﻨﺤﺼﺮ ﺑﻪ ﻓﺮﺩ‬

‫• ﻩ ﻧﮕﺎﻣﻲ ﻛﻪ ﺷﻜﺎﻑ ﻳﺎ ‪ coalescing‬ﺑﺮﺍﻱ ﺗﻐﻴﻴﺮﺍﺕ ﻳﻚ ﻭﺍﻟﺪ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﺍﺳﺖ‪ ،‬ﻗﻔﻞ ﻭﺍﻟﺪ ﺩﺭ ﺣﺎﻟﺖ‬
‫ﻣﻨﺤﺼﺮﺑﻪ ﻓﺮﺩ ﺍﺳﺖ‪.‬‬

‫‪ ‬ﺑﺎﻻﺗﺮ ﺍﺯ ﺣﺪ ﭘﺮﻭﺗﻜﻞ ﺑﻦ ﺑﺴﺖ ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﺎﺷﺪ‪.‬‬

‫• ﺟﺴﺘﺠﻮﻱ ﭘﺎﻳﻴﻦ ﺁﻣﺪﻥ ﺩﺭﺧﺖ ﺑﻦ ﺑﺴﺖ ﺑﺎ ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ ﺑﺎﻻﺭﻓﺘﻦ ﺍﺯ ﺩﺭﺧﺖ‬

‫• ﺁﻳﺎ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﺑﻲ ﻧﺘﻴﺠﻪ ﺑﻤﺎﻧﻨﺪ ﻭ ﺟﺴﺘﺠﻮ ﻛﻨﻨﺪ ﺭﺍﻩ ﺍﻧﺪﺍﺯﻱ ﻣﺠﺪﺩ‪ ،‬ﺑﺪﻭﻥ ﺗﺎﺛﻴﺮ ﺑﺮ ﺗﺮﺍﻛﻨﺶ‬

‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﺑﻬﺘﺮ ﺩﺭ ﺩﺳﺘﺮﺱ ﻫﺴﺘﻨﺪ‪ ،‬ﻧﮕﺎﻩ ﻛﻨﻴﺪ ﺑﻪ ﺑﺨﺶ ‪ 16.9‬ﺑﺮﺍﻱ ﻳﻚ ﭘﺮﻭﺗﻜﻞ‪ ،‬ﭘﺮﻭﺗﻜﻞ ﺩﺭﺧﺖ ‪B-‬‬
‫‪LINK‬‬

‫• ﺷﻬﻮﺩ ‪ :‬ﻗﻔﻞ ﺁﺯﺍﺩﻱ ﺩﺭ ﻭﺍﻟﺪ ﻗﺒﻞ ﺍﺯ ﻛﺴﺐ ﻗﻔﻞ ﻛﻮﺩﻙ‬

‫‪ ‬ﻭ ﻣﻘﺎﺑﻠﻪ ﺑﺎ ﺗﻐﻴﻴﺮﺍﺗﻲ ﻛﻪ ﻣﻤﻜﻦ ﺍﺳﺖ ﺑﻴﻦ ﺁﺯﺍﺩﻱ ﻗﻔﻞ ﻭ ﺑﻪ ﺩﺳﺖ ﺁﻭﺭﺩﻥ ﺍﺗﻔﺎﻕ ﺍﻓﺘﺎﺩﻩ‬
‫ﺑﺎﺷﺪ‪.‬‬

‫ﻗﻔﻞ ﻛﻠﻴﺪ ﺑﻌﺪﻱ‬

‫‪ ‬ﭘﺮﻭﺗﻜﻞ ﺷﺎﺧﺺ ﻗﻔﻞ ﺑﺮﺍﻱ ﺟﻠﻮﮔﻴﺮﻱ ﺍﺯ ﻓﺎﻧﺘﻮﻡ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﻗﻔﻞ ﻛﻞ ﺑﺮگ‬

‫• ﻣﻲ ﺗﻮﺍﻧﺪ ﺩﺭ ﻫﻤﺰﻣﺎﻧﻲ ﺿﻌﻴﻒ ﺷﻮﺩ ﺩﺭ ﺻﻮﺭﺗﻲ ﻛﻪ ﺩﺭﺝ ﻫﺎﻱ ﺑﺴﻴﺎﺭﻱ ﻭﺟﻮﺩ ﺩﺍﺭﺩ‬

‫‪ ‬ﺟﺎﻳﮕﺰﻳﻦ ‪ :‬ﺑﺮﺍﻱ ﻣﺮﺍﺟﻌﻪ ﺑﻪ ﺷﺎﺧﺺ‬

‫‪97‬‬
‫• ﻗﻔﻞ ﻛﺮﺩﻥ ﺗﻤﺎﻡ ﺍﺭﺯﺵ ﻫﺎﻳﻲ ﻛﻪ ﻣﺮﺍﺟﻌﻪ ﺑﻪ ﺷﺎﺧﺺ ﺭﺍ ﺑﺮﺁﻭﺭﺩﻩ ﻣﻲ ﻛﻨﺪ)ﻣﻄﺎﺑﻘﺖ ﺑﺎ ﺍﺭﺯﺵ ﻣﺮﺍﺟﻌﻪ‬
‫ﻭ ﻳﺎ ﺳﻘﻮﻁ ﺩﺭ ﻣﺤﺪﻭﺩﻩ ﻣﺮﺍﺟﻌﻪ(‬

‫• ﻫﻤﭽﻨﻴﻦ ﺍﺭﺯﺵ ﻛﻠﻴﺪ ﺑﻌﺪﻱ ﺩﺭ ﺷﺎﺧﺺ ﻗﻔﻞ‬

‫• ﺣﺎﻟﺖ ﻗﻔﻞ ‪ S :‬ﺑﺮﺍﻱ ‪ X ، lookups‬ﺑﺮﺍﻱ ﺩﺭﺝ‪/‬ﺣﺬﻑ‪/‬ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ‬

‫‪ ‬ﻣﺤﺪﻭﺩﻩ ﻧﻤﺎﻳﺶ ﺩﺍﺩﻩ ﺷﺪﻩ ﺩﺭ ﺗﻀﺎﺩ ﺑﺎ ﺩﺭﺝ ‪ /‬ﺣﺬﻑ ‪ /‬ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ ﺭﺍ ﺗﻀﻤﻴﻦ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫• ﺻﺮﻑ ﻧﻈﺮ ﺍﺯ ﺍﻳﻦ ﻛﻪ ﺍﺗﻔﺎﻕ ﻣﻲ ﺍﻓﺘﺪ ‪ ،‬ﺗﺎ ﺯﻣﺎﻧﻲ ﻛﻪ ﻫﺮﺩﻭ ﻫﻤﺰﻣﺎﻥ ﻫﺴﺘﻨﺪ‬

‫ﺍﺳﻼﻳﺪﻫﺎﻱ ﺍﺿﺎﻓﻲ‬

‫ﺟﺪﺍﺳﺎﺯﻱ ﻓﻮﺭﻱ‬

‫‪ ‬ﺍﻧﮕﻴﺰﻩ ‪ :‬ﭘﺸﺘﻴﺒﺎﻧﻲ ﻧﻤﺎﻳﺶ ﺩﺍﺩﻩ ﺷﺪﻩ ﺗﺼﻤﻴﻢ ﻣﻲ ﮔﻴﺮﺩ ﻛﻪ ﺧﻮﺍﻧﺪﻥ ﻣﻘﺪﺍﺭﻫﺎﻱ ﺯﻳﺎﺩﻱ ﺍﺯ ﺩﺍﺩﻩ ﺑﻪ ﻃﻮﺭ‬
‫ﻫﻤﺰﻣﺎﻥ ﺩﺭ ﺗﻀﺎﺩﻧﺪ ﺑﺎ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ‪ OLTP‬ﻛﻪ ﭼﻨﺪ ﺭﺩﻳﻒ ﺭﺍ ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ ﻣﻲ ﻛﻨﻨﺪ‪.‬‬

‫• ﻧﺘﺎﻳﺞ ﻋﻤﻠﻜﺮﺩ ﺿﻌﻴﻒ‬

‫‪ ‬ﺭﺍﻩ ﺣﻞ ‪ : 1‬ﺩﺍﺩﻥ "ﻋﻜﺲ ﻓﻮﺭﻱ” ﻣﻨﻄﻘﻲ ﺍﺯ ﺣﺎﻟﺖ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺗﻨﻬﺎ ﺑﺮﺍﻱ ﺧﻮﺍﻧﺪﻥ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ‪ ،‬ﺧﻮﺍﻧﺪﻥ ﻭ‬
‫ﻧﻮﺷﺖﻥ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻗﻔﻞ ﻃﺒﻴﻌﻲ‬

‫• ‪ Multiversion‬ﻗﻔﻞ ‪ 2‬ﻓﺎﺯ‬

‫• ﻛﺎﺭ ﻣﻲ ﻛﻨﻨﺪ ﺑﻪ ﺧﻮﺑﻲ ‪ ،‬ﺍﻣﺎ ﭼﮕﻮﻧﻪ ﺑﺎﻳﺪ ﺑﺪﺍﻥﻱﻡ ﻛﻪ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﻓﻘﻂ ﺧﻮﺍﻧﺪﻧﻲ ﺍﺳﺖ؟‬

‫‪ ‬ﺭﺍﻩ ﺣﻞ ‪ :2‬ﺗﺼﻮﻳﺮ ﻓﻮﺭﻱ ﺍﺯ ﺣﺎﻟﺖ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺑﺮﺍﻱ ﻫﺮ ﺕﺭﺍﻛﻨﺶ ﺭﺍ ﻣﻲ ﺩﻫﺪ‪ ،‬ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ ﻫﺎ ﺗﻨﻬﺎ ﻣﻮﺭﺩ‬
‫ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻗﻔﻞ ‪ 2‬ﻓﺎﺯ ﺑﻪ ﻣﺤﺎﻓﻈﺖ ﺩﺭ ﺑﺮﺍﺑﺮ ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ ﻫﻤﺰﻣﺎﻥ ﻫﺴﺘﻨﺪ‪.‬‬

‫• ﻣﺸﻜﻞ ‪ :‬ﺍﻧﻮﺍﻉ ﻧﺎﻫﻨﺠﺎﺭﻱ ﻫﺎ ﻣﺎﻧﻨﺪ ﺍﺯﺩﺳﺖ ﺭﻓﺘﻦ ﻧﺘﺎﻳﺞ ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ‬

‫• ﺭﺍﻩ ﺣﻞ ﺟﺰﺋﻲ ‪ :‬ﺳﻄﺢ ﺟﺪﺍﺳﺎﺯﻱ ﻋﻜﺲ ﻓﻮﺭﻱ )ﺍﺳﻼﻳﺪ ﺑﻌﺪﻱ(‬

‫‪ ‬ﭘﻴﺸﻨﻬﺎﺩ ﺷﺪﻩ ﺗﻮﺳﻂ ‪ Berenson‬ﻭ ﻫﻤﻜﺎﺭﺍﻥ ‪SIGMOD 1995 ،‬‬

‫‪98‬‬
 ‫‪ ‬ﺍﻧﻮﺍﻉ ﺍﺟﺮﺍﻫﺎ ﺩﺭ ﺑﺴﻴﺎﺭﻱ ﺍﺯ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ‬

‫– ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ ‪ .‬ﺍﻭﺭﺍﻛﻞ‪ ،‬ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ‪ PostgreSQL ،SQL‬ﺳﺮﻭﺭ ‪2005‬‬

‫‪ ‬ﺍﺟﺮﺍﻱ ﺗﺮﺍﻛﻨﺶ ‪ T1‬ﺑﺎ ﺟﺪﺍﺳﺎﺯﻱ ﻓﻮﺭﻱ‬

‫• ﭘﺬﻳﺮﺵ ﻓﻮﺭﻱ ﺩﺍﺩﻩ ﻫﺎ ﺩﺭ ﺷﺮﻭﻉ ﻃﻮﻝ ﻣﻲ ﻛﺸﺪ‪.‬‬

‫• ﻫﻤﻴﺸﻪ ﺗﻐﻴﻴﺮ ﺩﺍﺩﻩ ﻫﺎ ﺭﺍ ﺩﺭ ﻋﻜﺲ ﻓﻮﺭﻱ ﺧﻮﺩ ﻣﻲ ﺧﻮﺍﻧﺪ‬

‫• ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﻫﻤﺰﻣﺎﻥ ﺑﺮﺍﻱ ‪ T1‬ﻗﺎﺑﻞ ﺭﺅﻳﺖ ﻧﻴﺴﺖ‪.‬‬

‫• ‪ T1‬ﺭﺍ ﺑﻪ ﻃﻮﺭ ﻛﺎﻣﻞ ﻣﻲ ﻧﻮﻳﺴﺪ ﺯﻣﺎﻧﻲ ﻛﻪ ﺁﻥ ﭘﺬﻳﺮﻓﺘﻪ ﺷﺪ‬

‫• ﺍﻭﻟﻴﻦ ﻗﺎﻧﻮﻥ ﭘﺬﻳﺮﺵ ﺑﺮﺩﻥ‪:‬‬

‫‪ ‬ﻣﻲ ﭘﺬﻳﺮﺩ ﺗﻨﻬﺎ ﺍﮔﺮ ﻫﻴﭻ ﺗﺮﺍﻛﻨﺶ ﻫﻤﺰﻣﺎﻥ ﺩﻳﮕﺮﻱ ﺩﺭ ﺣﺎﻝ ﺣﺎﺿﺮ ﻧﻮﺷﺘﻪ ﻧﺸﺪﻩ ﺍﺳﺖ ﺩﺍﺩﻩ‬
‫ﺍﻱ ﻛﻪ ‪ T1‬ﺑﺮﺍﻱ ﻧﻮﺷﺘﻦ ﺩﺭ ﻧﻈﺮ ﺩﺍﺭﺩ‪.‬‬

‫‪T1‬‬ ‫‪T2‬‬ ‫‪T3‬‬

‫‪99‬‬
 ‫)‪W(Y := 1‬‬

‫‪Commit‬‬

‫‪Start‬‬

‫‪R(X)  0‬‬

‫‪R(Y) 1‬‬

‫)‪W(X:=2‬‬

‫)‪W(Z:=3‬‬

‫‪Commit‬‬

‫‪R(Z)  0‬‬

‫‪R(Y)  1‬‬

‫)‪W(X:=3‬‬

‫‪Commit-Req‬‬

‫ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﯽ ﻫﻣﺯﻣﺎﻥ ﻗﺎﺑﻝ ﺭﺅﻳﺕ ﻧﻳﺳﺕ‬ ‫‪Abort‬‬

‫ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﯽ ﻫﺎ ﻗﺎﺑﻝ ﻣﺷﺎﻫﺩﻩ ﻣﯽ ﺑﺎﺷﻧﺩ‬
‫ﻧﻪ ﺍﻭﻝ ﭘﺫﻳﺭﺵ ‪x‬‬
‫ﺧﻁﺎی ‪ T2 ، serialization‬ﻫﺳﺕ ‪Rollback‬‬

‫ﻣﺰﺍﻳﺎﻱ ‪SI‬‬

‫‪ ‬ﺧﻮﺍﻧﺪﻧﻲ ﻛﻪ ﻫﺮﮔﺰ ﻣﺴﺪﻭﺩ ﻧﺸﺪﻩ‬

‫• ﻭﻫﻤﭽﻴﻨﻴﻦ ﻓﻌﺎﻟﻴﺖ ﻫﺎﻱ ﺩﻳﮕﺮ ‪ txns‬ﻫﺎ ﺭﺍ ﻣﺴﺪﻭﺩ ﻧﻤﻲ ﻛﻨﺪ‪.‬‬

‫‪ ‬ﻋﻤﻠﻜﺮﺩ ﻣﺸﺎﺑﻪ ﺧﻮﺍﻧﺪﻥ ﭘﺬﻳﺮﻓﺘﻪ ﺷﺪﻩ‬

‫‪ ‬ﺟﻠﻮﮔﻴﺮﻱ ﺍﺯ ﻧﺎﻫﻨﺠﺎﺭﻱ ﻫﺎﻱ ﻣﻌﻤﻮﻝ‬

‫• ﺑﺪﻭﻥ ﻋﻨﻮﺍﻥ ﺯﺷﺖ ﺧﻮﺍﻧﺪﻩ ﺷﺪﻩ‬

‫‪100‬‬
 ‫• ﺑﺪﻭﻥ ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ ﺍﺯ ﺩﺳﺖ ﺭﻓﺘﻪ‬

‫• ﺑﺪﻭﻥ ﻋﻨﻮﺍﻥ ﺧﻮﺍﻧﺪﻩ ﺷﺪﻥ ﻏﻴﺮ ﻗﺎﺑﻞ ﺗﻜﺮﺍﺭ‬

‫• ﮔﺰﺍﺭﻩ ﻫﺎ ﺑﺮ ﺍﺳﺎﺱ ﺍﻧﺘﺨﺎﺏ ﺗﻜﺮﺍﺭ ﻫﺴﺘﻨﺪ )ﺑﺪﻭﻥ ﻓﺎﻧﺘﻮﻡ(‬

‫‪ ‬ﻣﺸﻜﻼﺕ ﺑﺎ ‪SI‬‬

‫• ‪ SI‬ﻫﻤﻴﺸﻪ ﻗﺎﺑﻞ ﺍﺟﺮﺍ ﺑﻪ ﺻﻮﺭﺕ ‪ serializable‬ﻧﻴﺴﺖ‪.‬‬

‫‪ : Serializable ‬ﻣﻴﺎﻥ ﺩﻭ ‪ txns‬ﻫﻤﺰﻣﺎﻥ‪ ،‬ﻳﻜﻲ ﺍﺯ ﺍﺛﺮﺍﺕ ﺩﻳﮕﺮ ﺭﺍ ﻣﻲ ﺑﻴﻨﺪ‪.‬‬

‫‪ ‬ﺩﺭ ‪ SI‬ﺍﺛﺮﺍﺕ ﺩﻳﮕﺮ ﺭﺍ ﻧﻤﻲ ﺑﻴﻨﺪ‬

‫ﻭﺍﻧﺪ ﻧﻘﺾ ﻛﻨﺪ ‪.‬‬

‫• ﻧﺘﻴﺠﻪ ‪ :‬ﻣﺤﺪﻭﺩﻳﺖ ﻫﺎﻱ ﺗﻤﺎﻣﻴﺖ ﺭﺍ ﻣﻲ ﺕ‬

‫ﺟﺪﺍﺳﺎﺯﻱ ﻓﻮﺭﻱ‬

‫‪ ‬ﻣﺜﺎﻟﻲ ﺍﺯ ﻣﺸﻜﻼﺕ ﺑﺎ ‪SI‬‬

‫• ‪T1: x:=y‬‬

‫• ‪T2: y:= x‬‬

‫• ﺩﺭ ﺍﺑﺘﺪﺍ ‪x = 3 and y = 17‬‬

‫‪ ‬ﺍﺟﺮﺍﻱ ﺳﺮﻳﺎﻝ ‪x = ??, y = ?? :‬‬

‫‪ ‬ﺍﮔﺮ ﻫﺮﺩﻭ ﺗﺮﺍﻛﻨﺶ ﻫﻤﺰﻣﺎﻥ ﺷﺮﻭﻉ ﺑﺸﻮﻧﺪ ﺑﺎ ﺟﺪﺍﺳﺎﺯﻱ ﻓﻮﺭﻱ ‪x = ?? , y = ?? :‬‬

‫‪ ‬ﺑﻪ ﻧﺎﻡ ﻧﻮﺷﺘﻦ ﻣﻮﺭﺏ‬

‫‪ ‬ﻣﻮﺭﺏ ﺑﻮﺩﻥ ﻧﻴﺰ ﺑﺎ ﺩﺭﺝ ﺭﺥ ﻣﻲ ﺩﻫﺪ‪.‬‬

‫• ﻣﺜﺎﻝ ‪:‬‬

‫‪101‬‬
 ‫‪ ‬ﻳﺎﻓﺘﻦ ﺣﺪﺍﻛﺜﺮ ﺗﻌﺪﺍﺩ ﺩﺳﺘﻮﺭﻫﺎ ﺩﺭ ﻣﻴﺎﻥ ﺗﻤﺎﻡ ﺩﺳﺘﻮﺭﺍﺕ‬

‫‪ ‬ﺍﻳﺠﺎﺩ ﻳﻚ ﻧﻈﻢ ﺟﺪﻳﺪ ﺑﺎ ﺩﺳﺘﻮﺭ ﺗﻌﺪﺍﺩ = ﺣﺪﺍﻛﺜﺮ ﻗﺒﻠﻲ ‪1+‬‬

‫ﻧﺎﻫﻨﺠﺎﺭﻱ ﻫﺎﻱ ﺟﺪﺍﺳﺎﺯﻱ ﻓﻮﺭﻱ‬

‫‪ serializability SI ‬ﺭﺍ ﻣﻲ ﺷﻜﻨﺪ ﺯﻣﺎﻧﻲ ﻛﻪ ‪ txns‬ﻫﺎ ﺁﻳﺘﻢ ﻫﺎﻱ ﻣﺨﺘﻠﻒ ﺭﺍ ﺗﻐﻴﻴﺮ ﻣﻲ ﺩﻫﻨﺪ‪ ،‬ﻫﺮ ﻛﺪﺍﻡ‬
‫ﺑﺮﺍﺳﺎﺱ ﻳﻚ ﺣﺎﻟﺖ ﻗﺒﻠﻲ ﺍﺯ ﺗﻐﻴﻴﺮﺍﺕ ﺁﻳﺘﻢ ﺩﻳﮕﺮ‬

‫• ﻩﻥ ﺩﺭ ﻋﻤﻞ ﺑﺴﻴﺎﺭ ﻗﺎﺑﻞ ﻗﺒﻮﻝ ﺍﺳﺖ‪.‬‬

‫‪ ‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ ‪ .‬ﻣﺤﻚ ‪ TPC-C‬ﺑﻪ ﺩﺭﺳﺘﻲ ﺗﺤﺖ ‪ SI‬ﺍﺟﺮﺍ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪ ‬ﺯﻣﺎﻧﻲ ﻛﻪ ‪ txns‬ﺩﺭﮔﻴﺮ ﺍﺻﻼﺡ ﺩﺍﺩﻩ ﻫﺎﻱ ﻣﺨﺘﻠﻒ ﺍﺳﺖ‪ ،‬ﻛﻪ ﻣﻌﻤﻮﻻ ﻫﺮ ﺩﻭ ﺁﻧﻬﺎ ﺭﻭﻱ ﻳﻚ‬
‫ﺁﻳﺘﻢ ﻣﺸﺘﺮﻙ ﺗﻐﻴﻴﺮ ﺩﺍﺭﻧﺪ ﻫﻤﭽﻨﻴﻦ )ﻣﺎﻧﻨﺪ ﻣﻘﺪﺍﺭ ﻛﻞ( ‪ SI‬ﻳﻜﻲ ﺍﺯ ﺁﻧﻬﺎ ﺑﻲ ﻧﺘﻴﺠﻪ ﺧﻮﺍﻫﺪ‬
‫ﻣﺎﻧﺪ‪.‬‬

‫• ﺍﻣﺎ ﺁﻳﺎ ﺭﺥ ﻡﻱ ﺩﻫﺪ‪.‬‬

‫‪ ‬ﺗﻮﺳﻌﻪ ﺩﻫﻨﺪﮔﺎﻥ ﻧﺮﻡ ﺍﻓﺰﺍﺭ ﺑﺎﻳﺪ ﺩﺭ ﻣﻮﺭﺩ ﻧﻮﺷﺘﻦ ﻣﻮﺭﺏ ﻣﺮﺍﻗﺐ ﺑﺎﺷﻨﺪ‪.‬‬

‫‪ SI‬ﻧﻴﺰ ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﺎﻋﺚ ﻧﺎﻫﻨﺠﺎﺭﻱ ﻓﻘﻂ ﺧﻮﺍﻧﺪﻧﻲ ﺗﺮﺍﻛﻨﺶ ﺷﻮﺩ ‪ ،‬ﻛﻪ ﺩﺭ ﺁﻥ ﺗﺮﺍﻛﻨﺶ ﻣﻤﻜﻦ ﺍﺳﺖ ﻳﻚ‬ ‫‪‬‬
‫ﺣﺎﻟﺖ ﻧﺎﻣﻨﺎﺳﺐ ﺣﺘﻲ ﺍﮔﺮ ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ ‪ serializable‬ﺑﺎﺷﺪ ﺭﺍ ﺑﺒﻴﻨﺪ‪.‬‬

‫• ﻣﺎ ﺟﺰﺋﻴﺎﺕ ﺭﺍ ﺣﺬﻑ ﻣﻲ ﻛﻨﻴﻢ‬

‫‪ SI‬ﺩﺭ ﺍﻭﺭﺍﻛﻞ ﻭ ‪PostgreSQL‬‬

‫‪ ‬ﻫﺸﺪﺍﺭ ‪ SI :‬ﺯﻣﺎﻧﻲ ﻛﻪ ﺳﻄﺢ ﺍﻧﺰﻭﺍ ﻳﻚ ﻣﺠﻤﻮﻋﻪ ﺍﺯ ‪ Serializable‬ﺍﺳﺖ ‪ ،‬ﺗﻮﺳﻂ ﺍﻭﺭﺍﻛﻞ ﻭ‬

‫‪ PostgreSQL‬ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫• ﭘﻴﺎﺩﻩ ﺳﺎﺯﻱ ‪ PostgreSQL‬ﺍﺯ ‪ SI‬ﺩﺭ ﺑﺨﺶ ‪ 26.4.1.3‬ﺷﺮﺡ ﺩﺍﺩﻩ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬

‫‪102‬‬
 ‫• ﻗﺎﻧﻮﻥ ﭘﻴﺎﺩﻩ ﺳﺎﺯﻱ ﺍﻭﺭﺍﻛﻞ “ ﺍﻭﻟﻴﻦ ﺑﺮﻧﺪﮔﺎﻥ ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ “)ﺍﻧﻮﺍﻉ ﻣﺨﺘﻠﻔﻲ ﺍﺯ “ ﺍﻭﻟﻴﻦ ﺑﺮﻧﺪﮔﺎﻥ‬
‫ﭘﺬﻳﺮﻓﺘﻪ ﺷﺪﻩ“ (‬

‫‪ ‬ﻧﻮﻳﺴﻨﺪﻩ ﻫﻤﺰﻣﺎﻥ ﺩﺭ ﺯﻣﺎﻥ ﻧﻮﺷﺘﻦ ﻋﻤﻞ ﭼﻚ ﻛﺮﺩﻥ ﺭﺍ ﺍﻧﺠﺎﻡ ﻣﻲ ﺩﻫﺪ‪ ،‬ﻧﻪ ﺩﺭ ﺯﻣﺎﻥ ﺗﺄﻳﻴﺪ‬

‫‪ ‬ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺑﻪ ‪ Rollback‬ﺍﻳﻦ ﺍﺟﺎﺯﻩ ﺭﺍ ﺯﻭﺩﺗﺮ ﻣﻲ ﺩﻫﻨﺪ‪.‬‬

‫• ﻧﻪ ﭘﺸﺘﻴﺒﺎﻧﻲ ﺍﺟﺮﺍﻱ ‪ Serializable‬ﺑﻪ ﺩﺭﺳﺘﻲ ﺍﻧﺠﺎﻡ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪ ‬ﺁﻳﺎ ﻣﻲ ﺗﻮﺍﻥ ﺭﺍﻩ ﻓﺮﺍﺭﻱ ﺑﺮﺍﻱ ﭘﺮﺱ ﻭ ﺟﻮﻱ ﺧﺎﺹ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺁﻥ ﺭﺍ ‪ ...‬ﺑﺮﺍﻱ ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ ﺩﺭ ‪ Oracle‬ﻭ‬
‫‪ PostgreSQL‬ﺍﻧﺘﺨﺎﺏ ﻛﺮﺩ‪.‬‬

‫• ‪ Lock‬ﻫﺎﻱ ﺩﺍﺩﻩ ﺍﻱ ﻛﻪ ‪ ،‬ﺑﺮﺍﻱ ﺟﻠﻮﮔﻴﺮﻱ ﺍﺯ ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ ﻫﻤﺰﻣﺎﻥ ﺧﻮﺍﻧﺪﻩ ﻣﻲ ﺷﻮﻧﺪ‪.‬‬

‫• ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ ‪،‬‬

‫‪ ‬ﺣﺪﺍﻛﺜﺮ )‪ (orderno‬ﺍﺯ ﺩﺳﺘﻮﺍﺕ ﺭﺍ ﺑﺮﺍﻱ ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ ﺍﻧﺘﺨﺎﺏ ﻛﻨﻴﺪ‪.‬‬

‫‪ ‬ﺍﺭﺯﺵ ﺧﻮﺍﻧﺪﻩ ﺷﺪﻩ ﺑﻪ ﻣﺘﻐﻴﺮ ﻣﺤﻠﻲ ‪maxorder‬‬

‫‪ ‬ﺩﺭﺝ ﺩﺳﺘﻮﺭﺍﺕ )…‪(maxorder+1,‬‬

‫ﭘﺎﻳﺎﻥ ﻓﺼﻞ‬

‫ﺧﻮﺍﻧﺪﻥ ﻓﻮﺭﻱ‬

‫ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ ﻫﺎﻱ ﻫﻤﺰﻣﺎﻥ ﺍﺯ ﺧﻮﺍﻥﺩﻥ ﻫﺎﻱ ﻓﻮﺭﻱ ﻣﺨﻔﻲ ﻫﺴﺘﻨﺪ‪.‬‬

‫‪103‬‬
 ‫ﻧﻮﺷﺘﻦ ﻓﻮﺭﻱ ‪ :‬ﺍﻋﻤﺎﻝ ﻛﻨﻨﺪﻩ ﺍﻭﻟﻴﻦ ﭘﻴﺮﻭﺯﻱ‬

‫‪104‬‬
 ‫• ﮔﻮﻧﺎﮔﻮﻧﻲ ‪ :‬ﺍﻭﻟﻴﻦ ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ ﻛﻪ ﺑﺮﻧﺪﻩ ﺷﺪ‬

‫‪ ‬ﭼﻚ ﻛﺮﺩﻥ ﺑﺮﺍﻱ ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ ﻫﻤﺰﻣﺎﻥ ﺩﺭ ﻫﻨﮕﺎﻡ ﻧﻮﺷﺘﻦ ﺭﺥ ﻣﻲ ﺩﻫﺪ‪.‬‬

‫‪) ‬ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺍﻭﺭﺍﻛﻞ ﺑﻪ ﻋﻼﻭﻩ ﺑﺮﺧﻲ ﺍﺯ ﻭﻳﮋﮔﻲ ﻫﺎﻱ ﺍﺿﺎﻓﻲ(‬

‫‪ ‬ﺗﻔﺎﻭﺕ ﺗﻨﻬﺎ ﺩﺭ ﺯﻣﺎﻧﻲ ﻛﻪ ﺑﻲ ﻧﺘﻴﺠﻪ ﻣﻲ ﻣﺎﻧﺪ ‪ ،‬ﺩﺭ ﻏﻴﺮ ﺍﻳﻦ ﺻﻮﺭﺕ ﻣﻌﺎﺩﻝ ﺍﺳﺖ‪.‬‬

‫‪ SI Non-Serializability‬ﺣﺘﻲ ﺑﺮﺍﻱ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﻓﻘﻂ ﺧﻮﺍﻧﺪﻧﻲ‬

‫‪105‬‬
 ‫ﺟﺰﺋﻴﺎﺕ ﺯﻣﺎﻧﺒﻨﺪ ﺗﺤﺖ ﻗﻔﻞ ‪ 2‬ﻓﺎﺯ‬

‫‪106‬‬
 ‫ﺯﻣﺎﻧﺒﻨﺪ ﻧﺎﻗﺺ ﺑﺎ ﻗﻔﻞ ﺗﺒﺪﻳﻞ‬

‫‪107‬‬
 ‫ﻧﻤﻮﺩﺍﺭ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺑﺎ ﺳﺎﺧﺘﺎﺭ ﺩﺭﺧﺖ‬

‫ﺯﻣﺎﻧﺒﻨﺪ ‪ Serializable‬ﺗﺤﺖ ﭘﺮﻭﺗﻜﻞ ﺩﺭﺧﺖ‬

‫‪108‬‬
 ‫ﺯﻣﺎﻧﺒﻨﺪ ‪3‬‬

‫‪109‬‬
 ‫ﺯﻣﺎﻧﺒﻨﺪ ‪4‬‬

‫ﺑﺮﻧﺎﻣﻪ ﺯﻣﺎﻧﺒﻨﺪ ‪ ،5‬ﺯﻣﺎﻧﺒﻨﺪ ﺗﻮﻟﻴﺪ ﺷﺪﻩ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺍﻋﺘﺒﺎﺭ ﺳﻨﺠﻲ‬

‫ﺳﺎﺯﮔﺎﺭﻱ ﺑﺎ ﻣﺎﺗﺮﻳﺲ‬

‫‪110‬‬
 ‫ﺯﻣﺎﻧﺒﻨﺪ ‪ nonserializable‬ﺑﺎ ﺩﺭﺟﻪ ﺳﺎﺯﮔﺎﺭﻱ ‪2‬‬

‫ﺩﺭﺧﺖ ‪ +B‬ﺑﺮﺍﻱ ﻣﺤﺎﺳﺒﻪ ﻱ ﻓﺎﻳﻞ ﻫﺎﻱ ‪n=3‬‬

‫ﺩﺭﺝ‪ Clearview‬ﺑﻪ ﺩﺍﺧﻞ ﺩﺭﺧﺖ ‪ +B‬ﺍﺯ ﺷﻜﻞ ‪16-21‬‬

‫‪111‬‬
 ‫ﻗﻔﻞ ﻣﺎﺗﺮﻳﺲ ﺳﺎﺯﮔﺎﺭﻱ‬

‫‪112‬‬
 ‫ﻓﺼﻞ ‪:6‬‬
‫ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺗﻮﺯﻳﻊ ﺷﺪﻩ‬

‫‪113‬‬
 ‫ﺳﻴﺴﺘﻢ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺗﻮﺯﻳﻊ ﺷﺪﻩ‪:‬‬
‫ﺳﻴﺴﺘﻢ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺗﻮﺯﻳﻊ ﺷﺪﻩ ﻣﺘﺸﻜﻞ ﺍﺯ ﺳﺎﻳﺖ ﻫﺎﻱ ﺁﺯﺍﺩﺍﻧﻪ ﺟﻔﺖ ﺷﺪﻩ ﺍﻱ ﺍﺳﺖ ﻛﻪ ﺩﺭ ﻫﻴﭻ ﻣﻮﻟﻔﻪ‬
‫ﻓﻴﺰﻳﻜﻲ ﺍﺷﺘﺮﺍﻙ ﻧﺪﺍﺭﻧﺪ ‪ .‬ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻛﻪ ﺩﺭ ﻫﺮ ﺳﺎﻳﺖ ﺍﺟﺮﺍ ﻣﻲ ﺷﻮﺩ‪ ،‬ﻣﺴﺘﻘﻞ ﺍﺯ‬
‫ﻳﻜﺪﻳﮕﺮ ﻣﻲ ﺑﺎﺷﺪ‪ .‬ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﻣﻤﻜﻦ ﺍﺳﺖ ﺩﺭ ﻳﻚ ﻳﺎ ﭼﻨﺪ ﺳﺎﻳﺖ ﺑﻪ ﺩﺍﺩﻩ ﻫﺎ ﺩﺳﺘﺮﺳﻲ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ ‪.‬‬

‫ﭘﺎﻳﮕﺎﻩ ﻫﺎﻱ ﺩﺍﺩﻩ ﺗﻮﺯﻳﻊ ﺷﺪﻩ ﻫﻤﮕﻦ‪:‬‬

‫‪ ‬ﺩﺭ ﻳﻚ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺗﻮﺯﻳﻊ ﺷﺪﻩ ﻫﻤﮕﻦ‬

‫‪ ‬ﺗﻤﺎﻣﻲ ﺳﺎﻳﺖ ﻫﺎ ﻧﺮﻡ ﺍﻓﺰﺍﺭ ﻳﻜﺴﺎﻥ ﺩﺍﺭﻧﺪ‪.‬‬

‫‪ ‬ﺑﺎ ﻳﻜﺪﻳﮕﺮ ﺁﺷﻨﺎﻳﻲ ﺩﺍﺭﻧﺪ ﻭ ﻣﻮﺍﻓﻖ ﻫﻤﻜﺎﺭﻱ ﺩﺭ ﺩﺭﺧﻮﺍﺳﺖ ﻫﺎﻱ ﻛﺎﺭﺑﺮ ﭘﺮﺩﺍﺯﺵ ﻫﺴﺘﻨﺪ ‪.‬‬

‫‪ ‬ﻫﺮ ﺳﺎﻳﺖ ﺑﺨﺸﻲ ﺍﺯ ﺍﺳﺘﻘﻼﻝ ﺍﺵ ﺭﺍ ﺑﺮ ﺣﺴﺐ ﺣﻖ ﺗﻐﻴﻴﺮ ﻃﺮﺡ ﻫﺎ ﻳﺎ ﻧﺮﻡ ﺍﻓﺰﺍﺭ ﻭﺍﮔﺬﺍﺭ‬
‫ﻣﻲ ﻛﻨﺪ‪.‬‬

‫‪ ‬ﺑﺮﺍﻱ ﻛﺎﺭﺑﺮ ﺑﻪ ﻋﻨﻮﺍﻥ ﻳﻚ ﺳﻴﺴﺘﻢ ﺧﺎﺹ ﺑﻪ ﻧﻈﺮ ﻣﻲ ﺭﺳﺪ ‪.‬‬

‫‪ ‬ﺩﺭﻳﻚ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺗﻮﺯﻳﻊ ﺷﺪﻩ ﻧﺎﻫﻤﮕﻦ‬

‫‪ ‬ﺳﺎﻳﺖ ﻫﺎﻱ ﻣﺘﻔﺎﻭﺕ ﻣﻤﻜﻦ ﺍﺳﺖ ﺍﺯ ﻧﺮﻡ ﺍﻓﺰﺍﺭ ﻭ ﻃﺮﺡ ﻫﺎﻱ ﻣﺨﺘﻠﻒ ﺍﺳﺘﻔﺎﺩﻩ ﻛﻨﻨﺪ ‪.‬‬

‫‪ ‬ﺗﻔﺎﻭﺕ ﺩﺭ ﻃﺮﺡ ﻣﺸﻜﻞ ﺍﺻﻠﻲ ﺑﺮﺍﻱ ﭘﺮﺩﺍﺯﺵ ﭘﺮﺱ ﻭ ﺟﻮ ﺍﺳﺖ ‪.‬‬

‫‪ ‬ﺗﻔﺎﻭﺕ ﺩﺭ ﻧﺮﻡ ﺍﻓﺰﺍﺭﻣﺸﻜﻞ ﺍﺻﻠﻲ ﺑﺮﺍﻱ ﭘﺮﺩﺍﺯﺵ ﺗﺮﺍﻛﻨﺶ ﺍﺳﺖ ‪.‬‬

‫ﺳﺎﻳﺖ ﻫﺎ ﻣﻤﻜﻦ ﻧﻴﺴﺖ ﺍﺯ ﻳﻜﺪﻳﮕﺮ ﺁﮔﺎﻫﻲ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻨﺪ ﻭ ﻣﻤﻜﻦ ﺍﺳﺖ ﻓﻘﻂ ﺗﺴﻬﻴﻼﺕ ﻣﺤﺪﻭﺩ ﺭﺍ ﺑﺮﺍﻱ‬
‫ﻫﻤﻜﺎﺭﻱ ﺩﺭ ﭘﺮﺩﺍﺯﺵ ﺗﺮﺍﻛﻨﺶ ﻓﺮﺍﻫﻢ ﻛﻨﻨﺪ ‪.‬‬

‫‪114‬‬
 ‫ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ ﺩﺍﺩﻩ ﺗﻮﺯﻳﻊ ﺷﺪﻩ‪:‬‬

‫ﻓﺮﺽ ﻣﻲ ﻛﻨﻴﻢ ﻛﻪ ﻳﻚ ﻣﺪﻝ ﺩﺍﺩﻩ ﺍﺭﺗﺒﺎﻃﻲ ﺩﺍﺭﻳﻢ‪،‬‬

‫‪ ‬ﺗﻜﺮﺍﺭ‬

‫• ﺳﻴﺴﺘﻢ ﻧﺴﺨﻪ ﻫﺎﻱ ﻣﺘﻌﺪﺩ ﺍﺯ ﺩﺍﺩﻩ ﺭﺍ ﺣﻔﻆ ﻣﻲ ﻛﻨﺪ‪،‬ﺩﺭ ﻣﺤﻞ ﻫﺎﻱ ﻣﺨﺘﻠﻒ ﺑﺮﺍﻱ‬
‫ﺑﺎﺯﻳﺎﺑﻲ ﺳﺮﻳﻌﺘﺮ ﻭ ﺗﻮﻟﺮﺍﻧﺲ ﺷﻜﺴﺖ ﻭ ﻋﻴﺐ ﺫﺧﻴﺮﻩ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫‪ ‬ﺍﻓﺮﺍﺯ ﻛﺮﺩﻥ‬

‫• ﺍﺭﺗﺒﺎﻁ ﺑﻪ ﭼﻨﺪﻳﻦ ﺑﺨﺶ ﺫﺧﻴﺮﻩ ﺷﺪﻩ ﺩﺭ ﺳﺎﻳﺖ ﻫﺎﻱ ﻣﺸﺨﺺ ﻭ ﻣﺠﺰﺍ ﺍﻓﺮﺍﺯﻛﺮﺩﻥ ﻣﻲ‬
‫ﺷﻮﺩ‪.‬‬

‫‪ ‬ﻛﭙﻲ ﻭ ﺍﻓﺮﺍﺯﻛﺮﺩﻥ ﻣﻲ ﺗﻮﺍﻧﺪ ﺗﺮﻛﻴﺐ ﺷﻮﺩ‪.‬‬

‫• ﺍﺭﺗﺒﺎﻁ ﺑﻪ ﭼﻨﺪﻳﻦ ﺑﺨﺶ ﺍﻓﺮﺍﺯﻛﺮﺩﻥ ﻣﻲ ﺷﻮﺩ ‪ :‬ﺳﻴﺴﺘﻢ ﭼﻨﺪﻳﻦ ﻛﭙﻲ ﻳﻜﺴﺎﻥ ﺍﺯ ﭼﻨﻴﻦ‬
‫ﺑﺨﺸﻲ ﺭﺍ ﺣﻔﻆ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫ﻛﭙﻲ ﺩﺍﺩﻩ‪:‬‬
‫ﻳﻚ ﺭﺍﺑﻄﻪ ﻳﺎ ﺑﺨﺸﻲ ﺍﺯ ﻳﻚ ﺭﺍﺑﻄﻪ ﻛﭙﻲ ﻣﻲ ﺷﻮﺩ ﺍﮔﺮ ﺑﻪ ﻃﻮﺭ ﺍ ﺿﺎﻓﻲ ﺩﺭ ﺩﻭ ﻳﺎ ﭼﻨﺪ ﺳﺎﻳﺖ ﺫﺧﻴﺮﻩ‬
‫ﺷﻮﺩ‪ .‬ﻛﭙﻲ ﻛﺎﻣﻞ ﻳﻚ ﺭﺍﺑﻄﻪ ﻣﻮﺭﺩﻱ ﺍﺳﺖ ﻛﻪ ﺭﺍﺑﻄﻪ ﺩﺭ ﺗﻤﺎﻣﻲ ﺳﺎﻳﺖ ﻫﺎ ﺫﺧﻴﺮﻩ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫ﭘﺎﻳﮕﺎﻩ ﻫﺎﻱ ﺩﺍﺩﻩ ﻛﺎﻣﻼ ﺩﺍﺭﺍﻱ ﺍﻃﻨﺎﺏ ﻭ ﺣﺸﻮ ‪ ،‬ﭘﺎﻳﮕﺎﻩ ﻫﺎﻳﻲ ﻫﺴﺘﻨﺪ ﻛﻪ ﻛﻪ ﺩﺭ ﺁﻥ ﻫﺮ ﺳﺎﻳﺖ ﺷﺎﻣﻞ‬
‫ﻧﺴﺨﻪ ﺍﻱ ﺍﺯ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻛﻞ ﺍﺳﺖ‪.‬‬

‫ﻣﺰﺍﻳﺎﻱ ﻛﭙﻲ‬
‫‪ ‬ﻗﺎﺑﻠﻴﺖ ﺩﺳﺘﺮﺳﻲ‪ :‬ﺷﻜﺴﺖ ﺳﺎﻳﺖ ﺷﺎﻣﻞ ﺭﺍﺑﻄﻪ ‪ r‬ﺩﺭ ﺻﻮﺭﺕ ﻭﺟﻮﺩ ﻛﭙﻲ ﻣﻨﺠﺮﺑﻪ ﻋﺪﻡ‬
‫ﺩﺳﺘﺮﺳﻲ ‪ r‬ﻧﻤﻲ ﺷﻮﺩ‪.‬‬

‫‪ ‬ﺗﺸﺎﺑﻪ ﻭ ﺗﻮﺍﺯﻱ‪ :‬ﺳﻮﺍﻻﺕ ﺩﺭ ‪ r‬ﻣﻤﻜﻦ ﺍﺳﺖ ﺗﻮﺳﻂ ﭼﻨﺪﻳﻦ ﮔﺮﻩ ﻣﻮﺍﺯﻱ ﭘﺮﺩﺍﺯﺵ ﺷﻮﺩ ‪.‬‬

‫‪115‬‬
‫‪ ‬ﺍﻧﺘﻘﺎﻝ ﻛﺎﻫﺶ ﻳﺎﻓﺘﻪ ﺩﺍﺩﻩ ‪ :‬ﺭﺍﺑﻄﻪ ‪ r‬ﺑﻪ ﻃﻮﺭ ﻣﺤﻠﻲ ﺩﺭ ﻫﺮ ﺳﺎﻳﺖ ﻣﺤﺘﻮﻱ ﻳﻚ ﻛﭙﻲ ‪r‬‬
‫ﺩﺭﺩﺳﺘﺮﺱ ﻣﻲ ﺑﺎﺷﺪ‪.‬‬

‫ﻣﻌﺎﻳﺐ ﻛﭙﻲ‬

‫‪ ‬ﻫﺰﻳﻦﻩ ﺍﻓﺰﺍﻳﺶ ﻳﺎﻓﺘﻪ ﺑﻪ ﺭﻭﺯ ﺭﺳﺎ ﻧﻲ ﻫﺎ ‪ :‬ﻫﺮ ﻛﭙﻲ ﺭﺍﺑﻄﻪ ‪ r‬ﺑﺎﻳﺴﺘﻲ ﺑﻪ ﺭﻭﺯ ﺷﻮﺩ‪.‬‬
‫‪ ‬ﭘﻲﭼﻴﺪﮔﻲ ﺍﻓﺰﺍﻳﺶ ﻳﺎﻓﺘﻪ ﻛﻨﺘﺮﻝ ﻫﻤﺰﻣﺎﻥ ‪ :‬ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ ﻫﺎﻱ ﻫﻤﺰﻣﺎﻥ ﺗﻜﺮﺍﺭﻫﺎﻱ‬
‫ﻣﺸﺨﺺ ﻭ ﻣﺠﺰﺍ ﻣﻤﻜﻦ ﺍﺳﺖ ﻣﻨﺠﺮ ﺑﻪ ﺩﺍﺩﻩ ﺷﻮﺩ‪.‬‬
‫‪ ‬ﭘﻴﭽﻴﺪﮔﻲ ﺍﻓﺰﺍﻳﺶ ﻳﺎﻓﺘﻪ ﻛﻨﺘﺮﻝ ﻫﻤﺰﻣﺎﻥ‪ :‬ﻛﭙﻲ ﻫﺎﻱ ﻣﺠﺰﺍ ﻭ ﻣﺸﺨﺺ ﺏﺭﻭﺯﺭﺳﺎﻧﻲ ﻫﺎﻱ‬
‫ﻫﻤﺰﻣﺎﻥ ﻣﻤﻜﻦ ﺍﺳﺖ ﻣﻨﺠﺮ ﺑﻪ ﺩﺍﺩﻩ ﻏﻴﺮ ﻣﺘﻨﺎﺳﺐ ﺷﻮﺩ ﻣﮕﺮﺁﻧﻜﻪ ﻣﻜﺎﻧﻴﺴﻢ ﻫﺎﻱ ﻛﻨﺘﺮﻝ‬
‫ﻫﻤﺰﻣﺎﻥ ﺧﺎﺹ ﺍﺟﺮﺍ ﺷﻮﺩ ‪.‬‬

‫ﻳﻚ ﺭﺍﻩ ﺣﻞ ‪ :‬ﻳﻚ ﻧﺴﺨﻪ ﺭﺍ ﺑﻪ ﻋﻨﻮﺍﻥ ﻧﺴﺨﻪ ﺍﻭﻟﻴﻪ ﻭ ﺍﺻﻠﻲ ﺍﻧﺘﺨﺎﺏ ﻛﻨﻴﺪ ﻭ ﻋﻤﻠﻴﺎﺕ ﻛﻨﺘﺮﻝ ﻫﻤﺰﻣﺎﻥ ﺑﺮ‬
‫ﻧﺴﺨﻪ ﺍﻭﻟﻴﻪ ﺭﺍ ﺑﻜﺎﺭ ﺑﺮﻳﺪ‪.‬‬

‫ﺑﺨﺶ ﺑﻨﺪﻱ ﺩﺍﺩﻩ‪:‬‬

‫ﺑﺨﺶ ﺭﺍﺑﻄﻪ ‪ r‬ﺩﺭ ﺑﺨﺶ ﻫﺎﻱ ‪ r1, r2, …, rn‬ﻛﻪ ﺷﺎﻣﻞ ﺍﻃﻼﻋﺎﺕ ﺍﻃﻼﻋﺎﺕ ﻛﺎﻓﻲ ﻭ ﻣﻮﺛﺮ ﺟﻬﺖ‬
‫ﺳﺎﺧﺘﺎﺭ ﺑﻨﺪﻱ ﻣﺠﺪﺩ ﺭﺍﺑﻄﻪ ‪ r‬ﺍﺳﺖ‪.‬‬

‫ﺑﺨﺶ ﺑﻨﺪﻱ ﺍﻓﻘﻲ ‪ :‬ﻫﺮﺗﺎﭘﻞ ‪ r‬ﺑﻪ ﻳﻚ ﻳﺎ ﭼﻨﺪ ﺑﺨﺶ ﺗﻌﻴﻴﻦ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﺑﺨﺶ ﻋﻤﻮﺩﻱ ‪ :‬ﻃﺮﺡ ﺑﺮﺍﻱ ﺍﺭﺗﺒﺎﻁ ‪ r‬ﺑﻪ ﻳﻚ ﻳﺎ ﭼﻨﺪﻳﻦ ﻃﺮﺡ ﻛﻮﭼﻜﺘﺮﺗﻌﻴﻴﻦ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫)ﻳﺎ ﻣﻮﺭﺩ ﺑﺮﺗﺮ ( ﺟﻬﺖ ﺍﻃﻤﻴﻨﺎﻥ ﺍﺯ‬ ‫ﺗﻤﺎﻣﻲ ﻃﺮﺡ ﻫﺎ ﺑﺎﻳﺴﺘﻲ ﺷﺎﻣﻞ ﻳﻚ ﻣﻮﺭﺩ ﻣﻨﺘﺨﺐ ﻣﺸﺘﺮﻙ‬
‫ﻭﻳﮋﮔﻲ ﻣﺸﺘﺮﻙ ﺑﻲ ﺯﻳﺎﻥ ﺑﺎﺷﺪ ‪ .‬ﻭﻳﮋﮔﻲ ﺧﺎﺹ ‪ ،‬ﻭﻳﮋﮔﻲ ﺗﺎﭘﻞ ﻣﻤﻜﻦ ﺍﺳﺖ ﺑﻪ ﻫﺮ ﻃﺮﺡ ﺍﺿﺎﻓﻪ ﺷﻮﺩ‬
‫ﺗﺎ ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﻮﺭﺩ ﻣﻨﺘﺨﺐ ﺑﻪ ﺧﺪﻣﺖ ﮔﺮﻓﺘﻪ ﺷﻮﺩ ‪.‬‬

‫‪116‬‬
 ‫ﻣﺜﺎﻝ ‪ :‬ﻣﺤﺎﺳﺒﻪ ﺭﺍﺑﻄﻪ ﺑﺎ ﻃﺮﺡ ﺯﻳﺮ‪ ،‬ﺣﺴﺎﺏ ‪) :‬ﻧﺎﻡ ﺷﻌﺒﻪ ‪ ،‬ﺷﻤﺎﺭﻩ ﺣﺴﺎﺏ ‪،‬ﻣﺎﻧﺪﻩ (‬

‫ﺍﻓﺮﺍﺯﻛﺮﺩﻥ ﺍﻓﻘﻲ ﺭﺍﺑﻄﻪ ﺣﺴﺎﺏ‬

‫ﺍﻓﺮﺍﺯﻛﺮﺩﻥ ﺍﻓﻘﻲ ﻛﺎﺭﻣﻨﺪ ﺑﻪ ﺣﺴﺎﺏ‬

‫‪117‬‬
 ‫ﻣﺰﺍﻳﺎﻱ ﺍﻓﺮﺍﺯﻛﺮﺩﻥ‬

‫‪ ‬ﺍﻓﻘﻲ‪ :‬ﺩﺭ ﺍﻳﻦ ﺣﺎﻟﺖ ﭘﺮﺩﺍﺯﺵ ﻣﻮﺍﺯﻱ ﺩﺭ ﺑﺨﺶ ﻫﺎﻱ ﻳﻚ ﺭﺍﺑﻄﻪ ﺭﺍ ﻣﺠﺎﺯ ﻣﻲ ﺳﺎﺯﺩ ‪ .‬ﺭﺍﺑﻄﻪ ﺭﺍ‬
‫ﻣﺠﺎﺯ ﺑﻪ ﺟﺪﺍﻳﻲ ﻭ ﺗﻘﺴﻴﻢ ﻣﻲ ﺳﺎﺯﺩ ﺗﺎ ﺟﺎﻳﻲ ﻛﻪ ﺗﺎﭘﻞ ﻫﺎ ﺩﺭ ﺟﺎﻳﻲ ﻭﺍﻗﻊ ﻣﻲ ﺷﻮﺩ ﻛﻪ ﻏﺎﻟﺒﺎ‬
‫ﻗﺎﺑﻞ ﺩﺳﺘﺮﺳﻲ ﻫﺴﺘﻨﺪ‪.‬‬

‫‪ ‬ﻋﻤﻮﺩﻱ ‪ :‬ﺩﺭ ﺍﻳﻨﺠﺎ ﺗﺎﭘﻞ ﻫﺎ ﺭﺍ ﻣﺠﺎﺯ ﺑﻪ ﺟﺪﺍﻳﻲ ﻣﻲ ﺳﺎﺯﺩ ﺗﺎ ﺟﺎﻳﻲ ﻛﻪ ﻫﺮ ﺑﺨﺶ ﺗﺎﭘﻞ ﺩﺭ ﺟﺎﻳﻲ‬
‫ﺫﺧﻴﺮﻩ ﻣﻲ ﺷﻮﺩ ﻛﻪ ﻏﺎﻟﺒﺎ ﻗﺎﺑﻞ ﺩﺳﺘﺮﺳﻲ ﺍﺳﺖ‪.‬‬

‫• ﻭﻳﮋﮔﻲ ﺗﺎﭘﻠﻲ ﺍﺗﺼﺎﻝ ﻣﻮﺛﺮ ﺑﺨﺶ ﻫﺎﻱ ﻋﻤﻮﺩﻱ ﺭﺍ ﻣﺠﺎﺯ ﻣﻲ ﺳﺎﺯﺩ ‪.‬‬

‫ﭘﺮﺩﺍﺯﺵ ﻣﻮﺍﺯﻱ ﺩﺭ ﻳﻚ ﺭﺍﺑﻄﻪ ﺭﺍ ﻣﺠﺎﺯ ﻣﻲ ﺳﺎﺯﺩ ‪.‬‬ ‫•‬

‫• ﺍﻓﺮﺍﺯﻛﺮﺩﻥ ﺍﻓﻘﻲ ﻭ ﻋﻤﻮﺩﻱ ﻣﻲ ﺗﻮﺍﻧﺪ ﺗﺮﻛﻴﺐ ﺷﻮﺩ‪.‬‬

‫• ﺑﺨﺶ ﻫﺎ ﻣﻤﻜﻦ ﺍﺳﺖ ﺑﻪ ﻃﻮﺭ ﻣﻮﻓﻘﻴﺖ ﺁﻣﻴﺰﻱ ﺑﻪ ﻋﻤﻖ ﺍﺧﺘﻴﺎﺭﻱ ﺍﻓﺮﺍﺯﻛﺮﺩﻥ ﺷﻮﺩ ‪.‬‬

‫ﺷﻔﺎﻓﻴﺖ ﺩﺍﺩﻩ‬
‫ﺷﻔﺎﻓﻴﺖ ﺩﺍﺩﻩ ﺩﺭﺟﻪ ﺍﻱ ﻛﻪ ﻛﺎﺭﺑﺮ ﺳﻴﺴﺘﻢ ﻣﻤﻜﻦ ﺍﺳﺖ ﺍﺯ ﺟﺰﺋﻴﺎﺕ ﭼﮕﻮﻧﮕﻲ ﻭ ﺟﺎﻳﮕﺎﻩ ﺫﺧﻴﺮﻩ ﺍﻗﻼﻡ‬
‫ﺩﺍﺩﻩ ﺩﺭ ﺳﻴﺴﺘﻢ ﺗﻮﺯﻳﻊ ﺷﺪﻩ ﺑﻲ ﺍﻃﻼﻉ ﺑﺎﺷﺪ‬

‫ﻣﻮﺿﻮﻋﺎﺕ ﺷﻔﺎﻓﻴﺖ ﺩﺭ ﺍﺭﺗﺒﺎﻁ ﺑﺎ ﻣﻮﺍﺭﺩ ﺯﻳﺮ ﺭﺍ ﺩﺭ ﻧﻈﺮ ﺑﮕﻴﺮﻳﺪ ‪:‬‬

‫ﺷﻔﺎﻓﻴﺖ ﺍﻓﺮﺍﺯﻛﺮﺩﻥ‬

‫ﺷﻔﺎﻓﻴﺖ ﺗﻜﺮﺍﺭ‬

‫ﺷﻔﺎﻓﻴﺖ ﻣﻮﻗﻌﻴﺖ‬

‫‪118‬‬
 ‫ﻣﻌﻴﺎﺭﻫﺎﻱ ﻧﺎﻣﮕﺬﺍﺭﻱ ﺍﻗﻼﻡ ﺩﺍﺩﻩ‬
‫ﺑﺎﻱ ﻧﺎﻡ ﻣﻨﺤﺼﺮ ﺑﻪ ﻓﺮﺩ ﮔﺴﺘﺮﺩﻩ ﺳﻴﺴﺘﻢ ﺭﺍ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ ‪.‬‬
‫ﻫﺮ ﻓﻘﺮﻩ ﺩﺍﺩﻩ ﺩ‬ ‫‪(1‬‬
‫ﻫﻤﭽﻨﻴﻦ ﺑﺎﻳﺴﺘﻲ ﺟﻬﺖ ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﻣﻮﻗﻌﻴﺖ ﺍﻗﻼﻡ ﺩﺍﺩﻩ ﺑﻪ ﻃﻮﺭ ﻣﻮﺛﺮ ﻭ ﺳﻮﺩﻣﻨﺪ ﻋﻤﻠﻲ ﻭ‬ ‫‪(2‬‬
‫ﻣﻤﻜﻦ ﺑﺎﺷﺪ‪.‬‬
‫ﺟﻬﺖ ﺗﻐﻴﻴﺮ ﻣﻮﻗﻌﻴﺖ ﺍﻗﻼﻡ ﺩﺍﺩﻩ ﺑﻪ ﻃﻮﺭ ﺷﻔﺎﻑ ﻋﻤﻠﻲ ﻭ ﻣﻤﻜﻦ ﺑﺎﺷﺪ ‪.‬‬ ‫‪(3‬‬
‫ﻫﻤﻴﻨﻄﻮﺭ ﻫﺮ ﺳﺎﻳﺖ ﺑﺎﻳﺴﺘﻲ ﻗﺎﺩﺭ ﺑﻪ ﺍﻳﺠﺎﺩ ﺍﻗﻼﻡ ﺩﺍﺩﻩ ﺟﺪﻳﺪ ﺑﻪ ﻃﻮﺭ ﻣﺴﺘﻘﻞ ﺑﺎﺷﺪ ‪.‬‬ ‫‪(4‬‬

‫ﻃﺮﺡ ﻣﺘﻤﺮﻛﺰ ﺷﺪﻩ‪ -‬ﺳﺮﻭﺭ ﻧﺎﻡ‬

‫‪ ‬ﺳﺎﺧﺘﺎﺭ‪ :‬ﺳﺮﻭﺭ ﻧﺎﻡ ﺑﺮﺍﻱ ﺗﻤﺎﻣﻲ ﺍﺳﺎﻣﻲ ﺗﻌﻴﻴﻦ ﻣﻲ ﺷﻮﺩ ﻭ ﻫﺮ ﺳﺎﻳﺖ ﺛﺒﺘﻲ ﺍﺯ ﺍﻗﻼﻡ ﺩﺍﺩﻩ ﻣﺤﻠﻲ‬
‫ﺭﺍ ﺍﻳﺠﺎﺩ ﻣﻲ ﻛﻨﺪ ‪ .‬ﺳﺎﻳﺖ ﻫﺎ ﺩﺭﺑﺎﺭﻩ ﺳﺮﻭﺭ ﻧﺎﻡ ﺑﺮﺍﻱ ﺗﻌﻴﻴﻦ ﻣﻮﻗﻌﻴﺖ ﺍﻗﻼﻡ ﺩﺍﺩﻩ ﻏﻴﺮﻣﺤﻠﻲ ﭘﺮﺱ‬
‫ﻭ ﺟﻮ ﻣﻲ ﻛﻨﻨﺪ‪.‬‬

‫‪ ‬ﻣﺰﺍﻳﺎ ‪ :‬ﻣﻌﻴﺎﺭ ‪3‬ﻭ‪ 1‬ﻧﺎﻣﮕﺬﺍﺭﻱ ﺭﺍ ﺑﺮﺁﻭﺭﺩﻩ ﻣﻲ ﺳﺎﺯﺩ‪.‬‬

‫‪ ‬ﻣﻌﺎﻳﺐ ‪ :‬ﻣﻌﻴﺎﺭ ‪ 4‬ﻧﺎﻣﮕﺬﺍﺭﻱ ﺭﺍ ﺑﺮﺁﻭﺭﺩﻩ ﻧﻤﻲ ﺳﺎﺯﺩ ‪ .‬ﺳﺮﻭﺭ ﻧﺎﻡ ﺑﻦ ﺑﺴﺖ ﻋﻤﻠﻜﺮﺩ ﺑﺎﻟﻘﻮﻩ ﺍﺳﺖ ‪.‬‬
‫ﺳﺮﻭﺭ ﻧﺎﻡ ﻭﻳﮋﮔﻲ ﺧﺎﺹ ﺷﻜﺴﺖ ﺍﺳﺖ‪.‬‬

‫ﻛﺎﺭﺑﺮﺩ ﻧﺎﻡ ﻫﺎﻱ ﻣﺴﺘﻌﺎﺭ‬

‫‪ ‬ﺟﺎﻳﮕﺰﻳﻨﻲ ﺑﻪ ﻃﺮﺡ ﻣﺘﻤﺮﻛﺰ ﺷﺪﻩ ‪ :‬ﻫﺮ ﺳﺎﻳﺖ ﺩﺍﺭﺍﻱ ﭘﻴﺸﻮﻧﺪ ﺷﻨﺎﺳﺎﻳﻲ ﻛﻨﻨﺪﻩ ﺳﺎﻳﺖ ﺧﻮﺩﺵ‬
‫ﺑﻪ ﻫﺮ ﻧﺎﻣﻲ ﺍﺳﺖ ﻛﻪ ﺍﻳﺠﺎﺩ ﻣﻲ ﻛﻦ ﺩ ‪ .‬ﺩﺍﺷﺘﻦ ﺷﻨﺎﺳﺎﻳﻲ ﻛﻨﻨﺪﻩ ﻣﻨﺤﺼﺮﺑﻪ ﻓﺮﺩ ﺭﺍ ﺗﻜﻤﻴﻞ ﻣﻲ‬
‫ﻛﻨﺪ ﻭ ﻣﺎﻧﻊ ﻣﺸﻜﻼﺕ ﻣﺮﺗﺒﻂ ﺑﺎ ﻛﻨﺘﺮﻝ ﻣﺮﻛﺰﻱ ﻣﻲ ﺷﻮﺩ ‪ .‬ﺑﺎ ﻭﺟﻮﺩ ﺍﻳﻦ ‪ ،‬ﻣﻮﻓﻖ ﺑﻪ ﺩﺳﺘﻴﺎﺑﻲ ﺑﻪ‬
‫ﺷﻔﺎﻓﻴﺖ ﺷﺒﻜﻪ ﻧﻤﻲ ﺷﻮﺩ‪.‬‬

‫ﺭﺍﻩ ﺣﻞ ‪ :‬ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺍﺳﺎﻡ ﻱ ﻣﺴﺘﻌﺎﺭ ﺭﺍ ﺑﺮﺍﻱ ﺍﻗﻼﻡ ﺩﺍﺩﻩ ﺍﻳﺠﺎﺩ ﻣﻲ ﻛﻨﺪ؛ ﻧﮕﺎﺷﺖ ﻧﺎﻡ ﻫﺎﻱ ﻣﺴﺘﻌﺎﺭ‬
‫ﺑﻪ ﺍﺳﺎﻣﻲ ﻭﺍﻗﻌﻲ ﺩﺭ ﻫﺮ ﺳﺎﻳﺖ ﺭﺍ ﺫﺧﻴﺮﻩ ﻣﻲ ﻛﻨﺪ ‪ .‬ﻛﺎﺭﺑﺮ ﻣﻲ ﺗﻮﺍﻧﺪ ﺍﺯ ﻣﻮﻗﻌﻴﺖ ﻓﻴﺰﻳﻜﻲ ﺁﻳﺘﻢ ﺩﺍﺩﻩ ﻧﺎﺁﮔﺎﻩ‬
‫ﺑﺎﺷﺪ ﻭﺍﮔﺮ ﺁﻳﺘﻢ ﺩﺍﺩﻩ ﺍﺯ ﻳﻚ ﺳﺎﻳﺖ ﺑﻪ ﺳﺎﻳﺖ ﺩﻳﮕﺮ ﺣﺮﻛﺖ ﻛﻨﺪ ‪ ،‬ﺗﺤﺖ ﺗﺎﺛﻴﺮ ﻧﻤﻲ ﺑﺎﺷﺪ ‪.‬‬

‫‪119‬‬
‫ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﺗﻮﺯﻳﻊ ﺷﺪﻩ ‪ :‬ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﻣﻤﻜﻦ ﺍﺳﺖ ﺩﺭ ﭼﻨﺪ ﺳﺎﻳﺖ ﺑﻪ ﺩﺍﺩﻩ ﺩﺳﺘﺮﺳﻲ ﺩﺍﺷﺘﻪ‬
‫ﺑﺎﺵﻥﺩ ‪.‬‬

‫‪ ‬ﻫﺮ ﺳﺎﻳﺖ ﻣﺪﻳﺮ ﺗﺮﺍﻛﻨﺶ ﻣﺤﻠﻲ ﻣﺴﺌﻮﻝ ﺭﺍ ﺩﺍﺭﺩ ‪ :‬ﺛﺒﺘﻲ ﺑﺮﺍﻱ ﺍﻫﺪﺍﻑ ﺑﻬﺒﻮﺩ ﺭﺍ ﺣﻔﻆ ﻣﻲ ﻛﻨﺪ ‪.‬‬
‫ﺩﺭ ﻫﻤﺎﻫﻨﮓ ﻛﺮﺩﻥ ﺍﺟﺮﺍﻱ ﻫﻤﺰﻣﺎﻥ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺍﺟﺮﺍﺷﺪﻩ ﺩﺭﺁﻥ ﺳﺎﻳﺖ ﻣﺸﺎﺭﻛﺖ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫‪ ‬ﻫﺮ ﺳﺎﻳﺖ ﺩﺍﺭﺍﻱ ﻳﻚ ﻫﻤﺎﻫﻨﮓ ﻛﻨﻨﺪﻩ ﺗﺮﺍﻛﻨﺶ ﺍﺳﺖ ﻛﻪ ﻣﺴﺊﻭﻝ ﻣﻮﺍﺭﺩ ﺯﻳﺮ ﻣﻲ ﺑﺎﺷﺪ ‪:‬‬

‫‪ .1‬ﺍﺟﺮﺍﻱ ﻣﻌﺎﻣﻼﺗﻲ ﺭﺍ ﺷﺮﻭﻉ ﻣﻲ ﻛﻨﺪ ﻛﻪ ﺩﺯﺭ ﺳﺎﻳﺖ ﺷﺮﻭﻉ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫‪ .2‬ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﻓﺮﻋﻲ ﺩﺭ ﺳﺎﻳﺖ ﻫﺎﻱ ﻣﻨﺎﺳﺐ ﺑﺮﺍﻱ ﺍﺟﺮﺍ ﺭﺍ ﺗﻮﺯﻳﻊ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫‪ .3‬ﭘﺎﻳﺎﻥ ﻫﺮ ﺗﺮﺍﻛﻨﺶ ﺩﺭ ﺳﺎﻳﺘﻲ ﺭﺍ ﻛﻪ ﺷﺮﻭﻉ ﻣﻲ ﺷﻮﺩ ‪،‬ﻫﻤﺎﻫﻨﮓ ﻣﻲ ﻛﻨﺪ ؛ ﭘﺎﻳﺎﻥ ﻫﺮ ﻳﻚ‬

‫ﻣﻤﻜﻦ ﺍﺳﺖ ﻣﻨﺠﺮ ﺑﻪ ﺗﺮﺍﻛﻨﺶ ﺍﻱ ﺷﻮﺩ ﻛﻪ ﺗﺮﺍﻛﻨﺶ ﺩﺭ ﺗﻤﺎﻣﻲ ﺳﺎﻳﺖ ﻫﺎ ﺍﺭﺍﺋﻪ ﻣﻲ‬
‫ﺷﻮﺩ ﻳﺎ ﺩﺭ ﺗﻤﺎﻣﻲ ﺳﺎﻳﺖ ﻫﺎ ﺑﻲ ﻧﺘﻴﺠﻪ ﻣﻲ ﻣﺎﻧﺪ ‪.‬‬

‫ﻣﻌﻤﺎﺭﻱ ﺳﻴﺴﺘﻢ ﺗﺮﺍﻛﻨﺶ‪:‬‬

‫ﺣﺎﻻﺕ ﺷﻜﺴﺖ ﺳﻴﺴﺘﻢ‪:‬‬

‫ﺷﻜﺴﺖ ﻫﺎﻱ ﻣﻨﺤﺼﺮ ﺑﻪ ﻓﺮﺩ ﺑﻪ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﺗﻮﺯﻳﻊ ﺷﺪﻩ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ‪ :‬ﺷﻜﺴﺖ ﻳﻚ ﺳﺎﻳﺖ ﻭﻳﺎ ﻛﻤﺒﻮﺩ‬
‫ﭘﻴﺎﻡ ﻫﺎ ﺑﺎﺷﻨﺪ ﻛﻪ ﺑﺎ ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﻛﻨﺘﺮﻝ ﺍﻧﺘﻘﺎﻝ ﺷﺒﻜﻪ ﺍﺯ ﻗﺒﻴﻞ ‪ TCP-IP‬ﻧﮕﻪ ﺩﺍﺷﺘﻪ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪120‬‬
 ‫ﺷﻜﺴﺖ ﻳﻚ ﻟﻴﻨﻚ ﺍﺭﺗﺒﺎﻁ‬

‫• ﺑﺎ ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﺷﺒﻜﻪ ﺗﻮﺳﻂ ﭘﻴﺎﻡ ﻫﺎﻱ ﻣﺴﻴﺮﻳﺎﺑﻲ ﺍﺯﻃﺮﻳﻖ ﻟﻴﻨﻚ ﻫﺎﻱ‬
‫ﺟﺎﻳﮕﺰﻳﻦ ﻧﮕﻪ ﺩﺍﺷﺘﻪ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫• ﺷﺒﻜﻪ ﺯﻣﺎﻧﻲ ﺍﻓﺮﺍﺯﻛﺮﺩﻥ ﻣﻲ ﺷﻮﺩ ﻛﻪ ﺑﻪ ﺩﻭ ﻳﺎ ﭼﻨﺪ ﺳﻴﺴﺘﻢ ﻓﺮﻋﻲ ﻣﺠﺰﺍ ﺷﻮﺩ‬

‫ﻛﻪ ﺩﺍﺭﺍﻱ ﻛﻤﺒﻮﺩ ﻫﺮ ﺍﺭﺗﺒﺎﻃﻲ ﺑﻴﻦ ﺁﻧﻬﺎ ﺍﺳﺖ ‪.‬‬

‫‪ .‬ﺑﻪ ﻃﻮﺭﻛﻠﻲ ﺍﻓﺮﺍﺯﻛﺮﺩﻥ ﻭ‬ ‫*ﺗﺬﻛﺮ‪ :‬ﺳﻴﺴﺘﻢ ﻓﺮﻋﻲ ﻣﻤﻜﻦ ﺍﺳﺖ ﻣﺘﺸﻜﻞ ﺍﺯ ﻳﻚ ﮔﺮﻩ ﺧﺎﺹ ﺑﺎﺷﺪ‬
‫ﺷﻜﺴﺖ ﻫﺎﻱ ﺳﺎﻳﺖ ﻏﻴﺮﻗﺎﺑﻞ ﺗﺸﺨﻴﺺ ﻣﻲ ﺑﺎﺷﺪ ‪.‬‬

‫ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﺗﻌﻬﺪ ‪ :‬ﺍﻳﻦ ﭘﺮﻭﺗﻜﻞ ﻫﺎ ﺟﻬﺖ ﺍﻃﻤﻴﻨﺎﻥ ﺍﺯ ﻇﺮﻓﻴﺖ ﺍﺗﻤﻲ ﺩﺭ ﺳﺎﻳﺖ ﻫﺎ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ‬
‫ﺷﻮﺩ‪ .‬ﺗﺮﺍﻛﻨﺶ ﺍﻱ ﻛﻪ ﺩﺭ ﺳﺎﻳﺖ ﻫﺎﻱ ﻣﺘﻌﺪﺩ ﺍﺟﺮﺍ ﻣﻲ ﺷﻮﺩ ﺑﺎﻳﺴﺘﻲ ﺩﺭ ﺗﻤﺎﻣﻲ ﺳﺎﻳﺖ ﻫﺎ ﻭﺍﮔﺬﺍﺭ ﺷﻮﺩ ﻳﺎ‬
‫ﺩﺭ ﺗﻤﺎﻣﻲ ﺳﺎﻳﺖ ﻫﺎ ﻱ ﻧﺘﻴﺠﻪ ﺑﺎﺷﺪ ‪ .‬ﺩﺍﺷﺘﻦ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﻭﺍﮔﺬﺍﺭ ﺷﺪﻩ ﺩﺭ ﻳﻚ ﺳﺎﻳﺖ ﻭ ﺑﻲ ﻧﺘﻴﺠﻪ‬
‫ﺑﻮﺩﻥ ﺩﺭ ﺳﺎﻳﺖ ﺩﻳﮕﺮ ﻏﻴﺮ ﻗﺎﺑﻞ ﻗﺒﻮﻝ ﺍﺳﺖ ‪ .‬ﭘﺮﻭﺗﻜﻞ ﺗﻌﻬﺪ ﺩﻭﻓﺎﺯﻱ )‪ (2PC‬ﺑﻪ ﻃﻮﺭﮔﺴﺘﺮﺩﻩ ﺍﻱ ﺍﺳﺘﻔﺎﺩﻩ‬
‫ﻣﻲ ﺷﻮﺩ‪ .‬ﭘﺮﻭﺗﻜﻞ ﺗﻌﻬﺪ ﺳﻪ ﻓﺎﺯﻱ )‪ (3pc‬ﭘﻴﭽﻴﺪﻩ ﺗﺮ ﻭ ﭘﺮﻫﺰﻳﻨﻪ ﺗﺮ ﺍﺳﺖ ﺍﻣﺎ ﻣﺎﻧﻊ ﭼﻨﺪ ﻋﻴﺐ ﭘﺮﻭﺗﻜﻞ‬
‫ﺗﻌﻬﺪ ﺩﻭﻓﺎﺯﻱ ﻣﻲ ﺷﻮﺩ‪ .‬ﺍﻳﻦ ﭘﺮﻭﺗﻜﻞ ﺩﺭ ﻋﻤﻞ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﭘﺮﻭﺗﻜﻞ ﺗﻌﻬﺪ ﺩﻭﻓﺎﺯﻱ )‪(2PC‬‬

‫ﻓﺮﺽ ﻛﻨﻴﺪ ﺳﺎﻳﺖ ﻫﺎﻱ ﻧﺎﻛﺎﻡ ﺷﺪﻩ ﻣﺪﻝ ﻋﺪﻡ ﻣﻮﻓﻘﻴﺖ ‪ -‬ﺗﻮﻗﻒ ﻓﻘﻂ ﻛﺎﺭ ﺭﺍ ﻣﺘﻮﻗﻒ ﻣﻲ ﻛﻨﺪ ﻭ ﺑﺎﻋﺚ‬
‫ﻫﻴﭻ ﺁﺳﻴﺐ ﺩﻳﮕﺮ ﺍﺯﻗﺒﻴﻞ ﻓﺮﺳﺘﺎﺩﻥ ﭘﻴﺎﻡ ﻫﺎﻱ ﻧﺎﺩﺭﺳﺖ ﻳﻪ ﺩﻳﮕﺮ ﺳﺎﻳﺖ ﻫﺎ ﻧﻤﻲ ﺷﻮﺩ ‪ .‬ﺍﺟﺮﺍﻱ ﭘﺮﻭﺗﻚﻝﻱ‬
‫ﺗﻮﺳﻂ ﻫﻤﺎﻫﻨﮓ ﻛﻨﻨﺪﻩ ﺑﻌﺪ ﺍﺯ ﺁﺧﺮﻳﻦ ﻣﺮﺣﻠﻪ ﺩﺳﺘﺮﺳﻲ ﺑﻪ ﺗﺮﺍﻛﻨﺶ ﺷﺮﻭﻉ ﻣﻲ ﺷﻮﺩ ‪ .‬ﭘﺮﻭﺗﻜﻞ ﺷﺎﻣﻞ‬
‫ﺗﻤﺎﻣﻲ ﺳﺎﻳﺖ ﻫﺎﻱ ﻣﺤﻠﻲ ﺍﺳﺖ ﻛﻪ ﺩﺭ ﺁﻥ ﺗﺮﺍﻛﻨﺶ ﺍﺟﺮﺍ ﻣﻲ ﺷﻮﺩ ‪ .‬ﺍﺟﺎﺯﻩ ﺩﻫﻴﺪ ‪ T‬ﺗﺮﺍﻛﻨﺶ ﺷﺮﻭﻉ‬
‫ﺷﺪﻩ ﺩﺭ ﺳﺎﻳﺖ ‪ Si‬ﻭ ﻫﻤﺎﻫﻨﮓ ﻛﻨﻨﺪﻩ ﺗﺮﺍﻛﻨﺶ ﺩﺭ ‪ Ci ،Si‬ﺑﺎﺷﺪ‪.‬‬

‫ﻣﺮﺣﻠﻪ ﺍﻭﻝ ‪ :‬ﺗﺼﻤﻴﻢ ﮔﻴﺮﻱ‬

‫‪ ‬ﻫﻤﺎﻫﻨﮓ ﻛﻨﻨﺪﻩ ﺩﺭﺑﺎﺭﻩ ﺗﻤﺎﻣﻲ ﺷﺮﻛﺖ ﻛﻨﻨﺪﮔﺎ ﻥ ﺑﺮﺍﻱ ﺁﻣﺎﺩﻩ ﺷﺪﻥ ﺩﺭ ﻭﺍﮔﺬﺍﺭﻱ ﺗﺮﺍﻛﻨﺶ ‪Ti‬‬
‫ﭘﺮﺱ ﻭ ﺟﻮ ﻣﻲ ﻛﻨﺪ‪ Ci .‬ﺛﺒﺖ ﻫﺎﻱ>‪ <prepare T‬ﺭﺍ ﺑﻪ ﺛﺒﺖ ﺍﺿﺎﻓﻪ ﻡﻱ ﻛﻨﺪ ﻭ ﺛﺒﺖ ﺭﺍ ﻭﺍﺩﺍﺭ‬

‫‪121‬‬
 ‫ﺑﻪ ﺗﺜﺒﻴﺖ ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ ﻣﻲ ﻛﻨﺪ ‪ .‬ﭘﻴﺎﻡ ﻫﺎﻱ ‪ prepare T‬ﺭﺍ ﺑﻪ ﺗﻤﺎﻣﻲ ﺳﺎﻳﺖ ﻫﺎﻳﻲ ﻣﻲ‬
‫ﻓﺮﺳﺘﺪ ﻛﻪ ‪ T‬ﺍﺟﺮﺍ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪ ‬ﺑﺎ ﺩﺭﻳﺎﻓﺖ ﭘﻴﺎﻡ ‪ ،‬ﻣﺪﻳﺮ ﺗﺮﺍﻛﻨﺶ ﺩﺭ ﺳﺎﻳﺖ ﺗﻌﻴﻴﻦ ﻣﻲ ﻛﻨﺪ ﻛﻪ ﺁﻳﺎ ﻣﻲ ﺗﻮﺍﻧﺪ ﺗﺮﺍﻛﻨﺶ ﺭﺍ ﻭﺍﮔﺬﺍﺭ‬
‫ﻛﻨﺪ‪ .‬ﺍﮔﺮ ﻧﺘﻮﺍﻧﺪ ﻭﺍﮔﺬﺍﺭ ﻛﻨﺪ‪ ،‬ﺛﺒﺖ >‪ <no T‬ﺭﺍ ﺑﻪ ﺛﺒﺖ ﺍﺿﺎﻓﻪ ﻣﻲ ﻛﻨﺪ ﻭ ﻭﭘﻴﺎﻡ ‪ abort T‬ﺭﺍ‬
‫ﺑﻪ ‪ Ci‬ﻣﻲ ﻓﺮﺳﺘﺪ ‪.‬ﺍﮔﺮ ﺗﺮﺍﻛﻨﺶ ﺑﺘﻮﺍﻧﺪ ﻭﺍﮔﺬﺍﺭ ﺷﻮﺩ ‪ ،‬ﺳﭙﺲ‪:‬‬

‫• ﺛﺒﺖ >‪ <ready T‬ﺭﺍ ﺑﻪ ﺛﺒﺖ ﺍﺿﺎﻓﻪ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫• ﺗﻤﺎﻣﻲ ﺛﺒﺖ ﻫﺎ ﺑﺮﺍﻱ ‪T‬ﺭﺍ ﻭﺍﺩﺍﺭ ﺑﻪ ﺫﺥﻳﺮﻩ ﺳﺎﺯﻱ ﺛﺎﺑﺖ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫• ﭘﻴﺎﻡ ‪ ready T‬ﺭﺍ ﺑﻪ ‪ Ci‬ﻣﻲ ﻓﺮﺳﺘﺪ‪.‬‬

‫ﻣﺮﺣﻠﻪ ‪ :2‬ﺛﺒﺖ ﺗﺼﻤﻴﻢ‬

‫‪ T ‬ﻣﻲ ﺗﻮﺍﻧﺪ ﺍﺯ ‪ Ci‬ﻣﺘﻌﻬﺪ ﺑﻪ ﭘﻴﺎﻡ ‪ ready T‬ﺩﺭﻳﺎﻓﺖ ﺷﺪﻩ ﺍﺯ ﺗﻤﺎﻣﻲ ﺳﺎﻳﺖ ﻫﺎﻱ ﻣﺸﺎﺭﻛﺘﻲ‬
‫ﺷﻮﺩ‪ :‬ﺩﺭﻏﻴﺮﺍﻳﻨﺼﻮﺭﺕ ‪ T‬ﺑﺎﻳﺴﺘﻲ ﺑﻲ ﻧﺘﻴﺠﻪ ﺑﺎﺷﺪ‪.‬‬

‫‪ ‬ﻫﻤﺎﻫﻨﮓ ﻛﻨﻨﺪﻩ ﺛﺒﺖ ﺗﺼﻤﻴﻢ ‪ , <commit T> or <abort T>,‬ﺭﺍ ﺑﻪ ﺛﺒﺖ ﺍﺿﺎﻓﻪ ﻣﻲ‬
‫ﻛﻨﺪ ﻭ ﺛﺒﺖ ﺭﺍ ﻭﺍﺩﺍﺭ ﺑﻪ ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ ﻣﻘﺎﻭﻡ ﻣﻲ ﻛﻨﺪ ‪ .‬ﺑﻪ ﻫﻨﮕﺎﻡ ﺫﺧﻴﺮﻩ ﺱﺍﺯ ﻣﻘﺎﻭﻡ ﺛﺒﺖ ‪ ،‬ﻏﻴﺮ‬
‫ﻗﺎﺑﻞ ﻓﺴﺦ ﺍﺳﺖ)ﺣﺘﻲ ﺍﮔﺮ ﺷﻜﺴﺖ ﺭﻭﻱ ﺩﻫﺪ(‪.‬‬

‫)ﺗﻌﻬﺪ ﻳﺎ ﺑﻲ ﻧﺘﻴﺠﻪ ( ﻣﻲ‬ ‫‪ ‬ﻫﻤﺎﻫﻨﮓ ﻛﻨﻨﺪﻩ ﭘﻴﺎﻣﻲ ﺭﺍ ﺑﻪ ﻫﺮ ﺷﺮﻛﺖ ﻛﻨﻨﺪﻩ ﻣﻄﻠﻊ ﺍﺯ ﺗﺼﻤﻴﻢ‬
‫ﻓﺮﺳﺘﺪ ‪.‬‬

‫‪ ‬ﺷﺮﻛﺖ ﻛﻨﻨﺪﮔﺎﻥ ﻋﻤﻞ ﻣﻨﺎﺳﺐ ﺭﺍ ﺑﻪ ﻃﻮﺭ ﻣﺤﻠﻲ ﻣﻲ ﭘﺬﻳﺮﻧﺪ‪.‬‬

‫ﻣﺪﻝ ﻫﺎﻱ ﺟﺎﻳﮕﺰﻳﻦ ﭘﺮﺩﺍﺯﺵ ﺗﺮﺍﻛﻨﺶ‬

‫‪ ‬ﻣﻔﻬﻮﻡ ﺳﺎﻳﺖ ﻫﺎﻱ ﻣﺘﻌﺪﺩ ﺩﺭﮔﺴﺘﺮﻩ ﻱ ﺗﺮﺍﻛﻨﺶ ﺧﺎﺹ ﺑﺮﺍﻱ ﭼﻨﺪﻳﻦ ﻛﺎﺭﺑﺮﺩ ﻧﺎﻣﻨﺎﺳﺐ ﺍﺳﺖ ‪.‬‬
‫ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ ﺗﺮﺍﻛﻨﺶ ﻋﺒﻮﺭﻱ ﺍﺯ ﻣﺮﺯ ﺳﺎﺯﻣﺎﻧﻲ ‪ ،‬ﻫﻴﭻ ﺳﺎﺯﻣﺎﻧﻲ ﺗﻤﺎﻳﻠﻲ ﺑﻪ ﻣﺠﺎﺯ ﻛﺮﺩﻥ‬
‫ﻱ ﺩﻭﺭﻩ‬ ‫ﺗﺮﺍﻛﻨﺶ ﺑﻪ ﻃﻮﺭﺧﺎﺭﺟﻲ ﺷﺮﻭﻉ ﺷﺪﻩ ﺟﻬﺖ ﻣﺴﺪﻭﺩ ﻛﺮﺩﻥ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﻣﺤﻠﻲ ﺑﺮﺍ‬
‫ﻧﺎﻣﻌﻴﻦ ﻧﺨﻮﺍﻫﺪ ﺩﺍﺷﺖ ‪.‬‬

‫‪122‬‬
‫‪ ‬ﻣﺪﻝ ﻫﺎﻱ ﺟﺎﻳﮕﺰﻳﻦ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺭﺍ ﺑﺎ ﻓﺮﺳﺘﺎﺩﻥ ﭘﻴﺎﻡ ﻫﺎ ﺍﻧﺠﺎﻡ ﻣﻲ ﺩﻫﺪ ‪ .‬ﻛﺪ ﺟﻬﺖ ﻧﮕﻪ ﺩﺍﺷﺘﻦ‬
‫ﭘﻴﺎﻡ ﻫﺎ ﺑﺎﻳﺴﺘﻲ ﺑﻪ ﺩﻗﺖ ﺟﻬﺖ ﺍﻃﻤﻴﻨﺎﻥ ﺩﻫﻲ ﻗﺎﺑﻠﻴﺖ ﺗﺪﺍﻭﻡ ﻭﻳﮋﮔﻲ ﻫﺎ ﺑﺮﺍﻱ ﺑﻪ ﺭﻭﺯﺭﺳﺎﻧﻲ ﻫﺎ‬
‫ﻃﺮﺍﺣﻲ ﺷﻮﺩ ‪ .‬ﺟﺪﺍﻳﻲ ﻧﻤﻲ ﺗﻮﺍﻧﺪ ﺿﻤﺎﻧﺖ ﺷﻮﺩ ‪،‬ﺩﺭ ﻣﺮﺍﺣﻞ ﻣﻴﺎﻧﻲ ﻗﺎﺑﻞ ﻣﺸﺎﻫﺪﻩ ﺍﺳﺖ‪ ،‬ﺍﻣﺎ ﻛﺪ‬
‫ﺑﺎﻳﺴﺘﻲ ﻫﻴﭻ ﻧﺘﻴﺠﻪ ﻏﻴﺮﻣﺘﻨﺎﺳﺐ ﺑﻌﻠﺖ ﻫﻤﮕﺎﻫﻲ ﺭﺍ ﺍﻃﻤﻴﻨﺎﻥ ﺩﻫﺪ ‪ .‬ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﭘﻴﺎﻡ ﺩﻫﻲ‬
‫ﺛﺎﺑﺖ‪ ،‬ﺳﻴﺴﺘﻢ ﻫﺎﻳﻲ ﻫﺴﺘﻨﺪ ﻛﻪ ﻭﻳﮋﮔﻲ ﻫﺎﻱ ﻣﻌﺎﻣﻼﺗﻲ ﺭﺍ ﺑﻪ ﭘﻴﺎﻡ ﻫﺎ ﺭﺍ ﺍﺭﺍﺋﻪ ﻣﻲ ﺩﻫﺪ ‪ .‬ﭘﻴﺎﻡ ﻫﺎ‬
‫ﺿﻤﺎﻧﺖ ﻣﻲ ﺷﻮﺩ ﺗﺎ ﺩﻗﻴﻘﺎ ﻳﻜﺒﺎﺭ ﺍﺭﺍﺋﻪ ﺷﻮﻧﺪ ‪.‬‬

‫‪ ‬ﻣﺜﺎﻝ‪ :‬ﺗﺤﺮﻳﻚ ﻛﻨﻨﺪﻩ ‪ :‬ﺍﻧﺘﻘﺎﻝ ﻭﺟﻮﻩ ﺑﻴﻦ ﻭﺍﮔﺬﺍﺭﻱ ﺩﻭ ﻣﺮﺣﻠﻪ ﺍﻱ ﺩﻭ ﺑﺎﻧﻚ ﭘﺘﺎﻧﺴﻴﻠﻲ ﺟﻬﺖ‬
‫ﺭﻭﺯ ﺭﺳﺎﻧﻲ ﻫﺎ ﺩﺭ ﺣﺴﺎﺏ ﻫﺎﻱ ﺷﺎﻣﻞ ﺷﺪﻩ ﺩﺭ ﺍﻧﺘﻘﺎﻝ ﻭﺟﻮﻩ ﺧﻮﺍﻫﺪ ﺩﺍﺷﺖ ‪.‬‬
‫ﻣﺴﺪﻭﺩﻛﺮﺩﻥ ﺏ‬

‫• ﺭﺍﻩ ﺣﻞ ﺟﺎﻳﮕﺰﻳﻦ ‪:‬‬

‫– ﭘﻮﻝ ﺑﺪﻫﻲ ﺍﺯ ﻣﻨﺒﻊ ﻣﺤﺎﺳﺒﻪ ﻣﻲ ﺷﻮﺩ ﻭ ﭘﻴﺎﻣﻲ ﺭﺍ ﺑﻪ ﺳﺎﻳﺖ ﺩﻳﮕﺮ ﻣﻲ ﻓﺮﺳﺘﺪ ‪.‬‬

‫– ﺳﺎﻳﺖ ﭘﻴﺎﻡ ﺭﺍ ﺩﺭﻳﺎﻓﺖ ﻣﻲ ﻛﻨﺪ ﻭ ﺣﺴﺎﺏ ﻣﻘﺼﺪ ﺭﺍ ﺑﺴﺘﺎﻧﻜﺎﺭ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫• ﭘﻴﺎﻡ ﺩﻫﻲ ﺍﺯ ﻣﺪﺗﻬﺎ ﻗﺒﻞ )ﺣﺘﻲ ﻗﺒﻞ ﺍﺯ ﺍﺧﺘﺮﺍﻉ ﺭﺍﻳﺎﻧﻪ ﻫﺎ ( ﺑﺮﺍﻱ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺗﻮﺯﻳﻊ ﺷﺪﻩ‬
‫ﺍﺳﺘﻔﺎﺩﻩ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬

‫‪ ‬ﻣﻮﺿﻮﻉ ﺍﺗﻤﻴﺴﻴﺘﻲ‬

‫• ﺯﻣﺎﻧﻲ ﻛﻪ ﺗﺮﺍﻛﻨﺶ ﭘﻴﺎﻣﻲ ﺭﺍ ﻣﻲ ﻓﺮﺳﺘﺪ ‪ ،‬ﻭﺍﮔﺬﺍﺭ ﻣﻲ ﺷﻮﺩ ‪ .‬ﭘﻴﺎﻡ ﺑﺎﻳﺴﺘﻲ ﺑﺮﺍﻱ ﺍﺭﺍﺋﻪ‬
‫ﺿﻤﺎﻧﺖ ﺷﻮﺩ‪.‬‬

‫ﺭﻱ ﭘﻴﺎﻡ ﻫﺎﻱ ﻳﺮ‬ ‫ﺿﻤﺎﻧﺖ ﻣﺎﺩﺍﻣﻲ ﻛﻪ ﺳﺎﻳﺖ ﻣﻘﺼﺪ ﺑﻪ ﺭﻭﺯ ﻭ ﻗﺎﺑﻞ ﺩﺳﺘﺮﺳﻲ ﺍﺳﺖ‪ ،‬ﻛﺪ ﺟﻬﺖ ﻧﮕﻬﺪﺍ‬
‫ﻗﺎﺑﻞ ﺗﺤﻮﻳﻞ ﺑﺎﻳﺴﺖﻱ ﻫﻤﭽﻨﻴﻦ ﺩﺭﺩﺳﺘﺮﺱ ﺑﺎﺷﺪ‪.‬‬

‫ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ ‪ :‬ﭘﻮﻝ ﺍﻋﺘﺒﺎﺭ ﺑﺮﮔﺸﺘﻲ ﺑﻪ ﺣﺴﺎﺏ ﻣﻨﺒﻊ ‪ :‬ﺍﮔﺮ ﻓﺮﺳﺘﺎﺩﻥ ﺗﺮﺍﻛﻨﺶ ﺑﻲ ﻧﺘﻴﺠﻪ ﺷﻮﺩ ‪،‬ﭘﻴﺎﻡ‬
‫ﻧﺒﺎﻳﺪ ﻓﺮﺳﺘﺎﺩﻩ ﺷﻮﺩ‪.‬‬

‫‪123‬‬
 ‫ﻣﺪﻳﺮ ﻣﺎﻧﻊ ﺗﻮﺯﻳﻊ ﺷﺪﻩ‬
‫∆ ﺩﺭ ﺍﻳﻦ ﺭﻭﻳﻜﺮﺩ‪ ،‬ﻋﻤﻠﻜﺮﺩ ﻗﻔﻞ ﺗﻮﺳﻂ ﻣﺪﻳﺮﺍﻥ ﻣﺎﻧﻊ ﺩﺭ ﻫﺮ ﺳﺎﻳﺖ ﺍﺟﺮﺍ ﻣﻲ ﺷﻮﺩ ‪ .‬ﻣﺪﻳﺮﺍﻥ ﻣﺎﻧﻊ‬
‫ﺩﺳﺘﺮﺳﻲ ﺑﻪ ﺍﻗﻼﻡ ﺩﺍﺩﻩ ﻣﺤﻠﻲ ﺭﺍ ﻛﻨﺘﺮﻝ ﻣﻲ ﻛﻨﻨﺪ ‪ .‬ﺍﻣﺎ ﻛﻨﺘﺮﻝ ﻫﺎﻱ ﺧﺎﺹ ﻣﻤﻜﻦ ﺍﺳﺖ ﺑﺮﺍﻱ‬
‫ﺗﻜﺮﺍﺭﻫﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺷﻮﺩ‪.‬‬

‫∆ ﻣﺰﻳﺖ ‪ :‬ﻛﺎﺭ ﺗﻮﺯﻳﻊ ﻣﻲ ﺷﻮﺩ ﻭ ﻣﻲ ﺗﻮﺍﻧﺪ ﺩﺭ ﺷﻜﺴﺖ ﻫﺎ ﺟﺪﻱ ﺍﻧﺠﺎﻡ ﺷﻮﺩ ‪.‬‬

‫∆ ﻋﻴﺐ ‪ :‬ﺷﻨﺎﺳﺎﻳﻲ ﻭﻗﻔﻪ ﺩﺷﻮﺍﺭﺗﺮ ﺍﺳﺖ ‪.‬‬

‫∆ ﭼﻨﺪﻳﻦ ﺗﻐﻴﻴﺮ ﺑﻪ ﺍﻳﻦ ﺭﻭﻳﻜﺮﺩ ‪ :‬ﺭﻭﻧﻮﺷﺖ ﺍﻭﻟﻴﻪ‪ ،‬ﭘﺮﻭﺗﻜﻞ ﺍﻛﺜﺮﻳﺖ‪ ،‬ﭘﺮﻭﺗﻜﻞ ﺑﺎﻳﺎﺱ ﺷﺪﻩ ﻭ ﻣﻮﺍﻓﻘﺖ‬
‫ﻋﻤﻮﻣﻲ ﺣﺪﻧﺼﺎﺏ‬

‫ﺭﻭﻧﻮﺷﺖ ﺍﻭﻟﻴﻪ‬
‫ﻳﻚ ﻛﭙﻲ ﺁﻳﺘﻢ ﺩﺍﺩﻩ ﺭﺍ ﺍﻧﺘﺨﺎﺏ ﻛﻨﻴﺪ ﺗﺎ ﺭﻭﻧﻮﺷﺖ ﺍﻭﻟﻴﻪ ﺑﺎﺷﺪ ‪ .‬ﺳﺎﻳﺖ ﺷﺎﻣﻞ ﻛﭙﻲ ﺳﺎﻳﺖ ﺍﻭﻟﻴﻪ ﺑﺮﺍﻱ ﺁﻥ‬
‫ﺁﻳﺘﻢ ﺩﺍﺩﻩ ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﺩ ‪ .‬ﺁﻳﺘﻢ ﻫﺎﻱ ﻣﺘﻔﺎﻭﺕ ﺩﺍﺩﻩ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﺳﺎﻳﺖ ﻫﺎﻱ ﺍﻭﻟﻴﻪ ﻣﺘﻔﺎﻭﺕ ﺩﺍﺷﺘﻪ‬
‫ﺑﺎﺷﻨﺪ‪ .‬ﺯﻣﺎﻧﻲ ﻛﻪ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺑﺎﻳﺴﺘﻲ ﺁﻳﺘﻢ ‪ Q‬ﺩﺍﺩﻩ ﺭﺍ ﻣﺴﺪﻭﺩ ﻛﻨﺪ‪،‬ﻗﻔﻞ ﻱ ﺩﺭ ﺳﺎﻳﺖ ﺍﻭﻟﻴﻪ ‪ Q‬ﺭﺍ‬
‫ﺩﺭﺧﻮﺍﺳﺖ ﻣﻲ ﻛﻨﺪ‪ .‬ﺑﻄﻮﺭ ﺿﻤﻨﻲ ﺩﺭ ﺗﻤﺎﻣﻲ ﺗﻜﺮﺍﺭﻫﺎﻱ ﺍﻳﺘﻢ ﺩﺍﺩﻩ ﻣﺴﺪﻭﺩ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫• ﻓﺎﻳﺪﻩ‪ :‬ﻛﻨﺘﺮﻝ ﻫﻤﺰﻣﺎﻥ ﺩﺍﺩﻩ ﺗﻜﺮﺍﺭﺷﺪﻩ ﺑﻪ ﺳﺎﺩﮔﻲ ﺑﻪ ﺍﺟﺮﺍﻱ ﺳﺎﺩﻩ‪ -‬ﺩﺍﺩﻩ ﺗﻜﺮﺍﺭﻧﺸﺪﻩ ﻧﮕﻪ ﺩﺍﺷﺘﻪ‬
‫ﻣﻲ ﺷﻮﺩ‪.‬‬

‫• ﻋﻴﺐ‪ :‬ﺍﮔﺮ ﺳﺎﻳﺖ ﺍﻭﻟﻴﻪ ‪ Q‬ﻣﻮﻓﻖ ﻧﺸﻮﺩ‪ Q ،‬ﻏﻴﺮﻗﺎﺑﻞ ﺩﺱﺗﺮﺳﻲ ﺍﺳﺖ ﺣﺘﻲ ﮔﺮﭼﻪ ﺩﻳﮕﺮ ﺳﺎﻳﺖ‬
‫ﻫﺎﻱ ﺷﺎﻣﻞ ﻳﻚ ﻛﭙﻲ ﻣﻤﻜﻦ ﺍﺳﺖ ﻗﺎﺑﻞ ﺩﺳﺘﺮﺳﻲ ﺑﺎﺷﻨﺪ ‪.‬‬

‫ﭘﺮﻭﺗﻜﻞ ﺍﻛﺜﺮﻳﺖ‬
‫ﻣﺪﻳﺮ ﻗﻔﻞ ﻣﺤﻠﻲ ﺩﺭ ﻫﺮ ﺳﺎﻳﺖ ﺩﺭﺧﻮﺍﺳﺖ ﻫﺎﻱ ﺑﺪﻭﻥ ﻗﻔﻞ ﻭ ﻗﻔﻞ ﺭﺍ ﺑﺮﺍﻱ ﺁﻳﺘﻢ ﻫﺎﻱ ﺫﺧﻴﺮﻩ ﺷﺪﻩ ﺩﺭ‬
‫ﺳﺎﻳﺖ ﺍﺟﺮﺍ ﻣﻲ ﻛﻨﺪ ‪ .‬ﺯﻣﺎﻧﻲ ﻛﻪ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺗﻤﺎﻳﻞ ﺑﻪ ﻗﻔﻞ ﺍﻳﺘﻢ ‪ Q‬ﺩﺍﺩﻩ ﻛﭙﻲ ﻧﺸﺪﻩ ﺑﺎﻗﻲ ﻣﺎﻧﺪﻩ ﺩﺭ‬
‫ﺳﺎﻳﺖ ‪ Si‬ﺩﺍﺭﺩ‪ ،‬ﭘﻴﺎﻣﻲ ﺑﻪ ﻣﺪﻳﺮ ﻗﻔﻞ ‪ Si‬ﻓﺮﺳﺘﺎﺩﻩ ﻣﻲ ﺷﻮﺩ ‪ .‬ﺍﮔﺮ ‪ Q‬ﺩﺭ ﺣﺎﻟﺖ ﻧﺎﺳﺎﺯﮔﺎﺭ ﻣﺴﺪﻭﺩ ﺷﻮﺩ ‪،‬‬
‫ﺳﭙﺲ ﺩﺭﺧﻮﺍﺳﺖ ﺗﺎ ﺯﻣﺎﻧﻲ ﻛﻪ ﺑﺘﻮﺍﻧﺪ ﺍﻋﻄﺎ ﺷﻮﺩ ‪ ،‬ﺑﻪ ﺗﺎﺧﻴﺮ ﻣﻲ ﺍﻓﺘﺪ ‪ .‬ﺯﻣﺎﻧﻴﻜﻪ ﺩﺭﺧﻮﺍﺳﺖ ﻗﻔﻞ ﻣﻲ ﺗﻮﺍﻧﺪ‬

‫‪124‬‬
‫ﺍﻋﻄﺎ ﺷﻮﺩ ﻣﺪﻳﺮ ﻗﻔﻞ ﭘﻴﺎﻣﻲ ﺭﺍ ﺑﻪ ﺁﻏﺎﺯﮔﺮ ﺑﺎﺯﭘﺲ ﻣﻲ ﻓﺮﺳﺘﺪ ؛ ﻧﺸﺎﻥ ﻣﻲ ﺩﻫﺪ ﻛﻪ ﺩﺭﺧﻮﺍﺳﺖ ﻗﻔﻞ ﺍﻋﻄﺎ‬
‫ﺷﺪﻩ ﺍﺳﺖ‪.‬‬

‫ﺩﺭﻣﻮﺭﺩ ﺩﺍﺩﻩ ﻛﭙﻲ ﺷﺪﻩ ‪ :‬ﺍﮔﺮ ‪ Q‬ﺩﺭ ‪ n‬ﺳﺎﻳﺖ ﻛﭙﻲ ﺷﻮﺩ ‪ ،‬ﺳﭙﺲ ﭘﻴﺎﻡ ﺩﺭﺧﻮﺍﺳﺖ ﻗﻔﻞ ﺑﺎﻳﺴﺘﻲ‬
‫ﺑﻪ ﺑﻴﺶ ﺍﺯﻧﻴﻤﻲ ﺍﺯ ‪ n‬ﺳﺎﻳﺖ ﻓﺮﺳﺘﺎﺩﻩ ﺷﻮﺩ ﻛﻪ ﺩﺭ ﺍﻥ ‪ Q‬ﺫﺧﻴﺮﻩ ﻣﻲ ﺷﻮﺩ‪ .‬ﺗﺮﺍﻛﻨﺶ ﺩﺭ ‪ Q‬ﻋﻤﻞ‬
‫ﻧﻤﻲ ﻛﻨﺪ ﺗﺎ ﺯﻣﺎﻧﻴﻜﻪ ﻗﻔﻞ ﺩﺭ ﺍﻛﺜﺮﻳﺖ ﺗﻜﺮﺍﺭﻫﺎﻱ ‪ Q‬ﺭﺍ ﺑﺪﺳﺖ ﺁﻭﺭﺩﻩ ﺑﺎﺷﺪ ‪ .‬ﺑﻪ ﻫﻨﮕﺎﻡ ﻧﻮﺷﺘﻦ‬
‫ﺁﻳﺘﻢ ﺩﺍﺩﻩ ‪،‬ﺗﺮﺍﻛﻨﺶ ﺛﺒﺖ ﻫﺎﻳﻲ ﺭﺍ ﺩﺭ ﺗﻤﺎﻣﻲ ﺗﻜﺮﺍﺭﻫﺎ ﺍﻧﺠﺎﻡ ﻣﻲ ﺩﻫﺪ ‪.‬‬

‫‪ .‬ﺟﺰﺋﻴﺎﺕ‬ ‫ﻓﺎﻳﺪﻩ ‪ :‬ﺣﺘﻲ ﺯﻣﺎﻧﻲ ﻛﻪ ﭼﻨﺪ ﺳﺎﻳﺖ ﺩﺭﺩﺳﺘﺮﺱ ﻧﻤﻲ ﺑﺎﺷﺪ‪ ،‬ﻣﻴﺘﻮﺍﻧﺪ ﺍﺳﺘﻔﺎﺩﻩ ﺷﻮﺩ‬
‫ﺩﺭﺑﺎﺭﻩ ﭼﮕﻮﻧﮕﻲ ﻧﮕﻬﺪﺍﺭﻱ ﺛﺒﺖ ﻫﺎ ﺩﺭ ﺣﻀﻮﺭ ﺷﻜﺴﺖ ﺳﺎﻳﺖ ﺑﻌﺪﺍ ﺍﺭﺍﺋﻪ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫ﻋﻴﺐ ‪ :‬ﺑﻪ )‪ 2(n/2 + 1‬ﭘﻴﺎﻡ ﺑﺮﺍﻱ ﻧﮕﻪ ﺩﺍﺷﺘﻦ ﺩﺭﺧﻮﺍﺳﺖ ﻫﺎﻱ ﻗﻔﻞ ﻭ )‪(n/2 + 1‬ﭘﻴﺎﻡ ﺑﺮﺍﻱ‬
‫ﻧﮕﻪ ﺩﺍﺷﺘﻦ ﺩﺭﺧﻮﺍﺳﺖ ﻫﺎﻱ ﻏﻴﺮﻣﺴﺪﻭﺩ ﻧﻴﺎﺯ ﺩﺍﺭﺩ ‪ .‬ﺍﻣﻜﺎﻥ ﺑﺎﻟﻘﻮﻩ ﺍﻱ ﺑﺮﺍﻱ ﻭﻗﻔﻪ ﺣﺘﻲ ﺑﺎ ﺁﻳﺘﻢ‬
‫ﺧﺎﺹ ﻭﺟﻮﺩ ﺩﺍﺭﺩ ‪ ،‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ ﻫﺮ ﻳﻚ ﺍﺯ ﺳﻪ ﺗﺮﺍﻛﻨﺶ ﻣﻤﻜﻦ ﺍﺳﺖ ﻗﻔﻞ ﻫﺎﻳﻲ ﺩﺭ ‪1/3rd‬‬
‫ﺗﻜﺮﺍﺭﻫﺎﻱ ﺩﺍﺩﻩ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻨﺪ‪.‬‬

‫ﭘﺮﻭﺗﻜﻞ ﺑﺎﻳﺎﺱ ﺷﺪﻩ‬

‫ﻣﺪﻳﺮ ﻗﻔﻞ ﻣﺤﻠﻲ ﺩﺭ ﻫﺮ ﺳﺎﻳﺖ ﻫﻤﺎﻧﻨﺪ ﭘﺮﺭﻭﺗﻜﻞ ﺍﻛﺜﺮﻳﺖ‪ ،‬ﺑﺎﻭﺟﻮﺩ ﺍﻳﻦ ﺩﺭﺧﻮﺍﺳﺖ ﻫﺎ ﺑﺮﺍﻱ ﻗﻔﻞ ﻫﺎﻱ‬
‫ﻣﺸﺘﺮﻙ ﻣﺘﻔﺎﻭﺕ ﺍﺯ ﺩﺭﺧﻮﺍﺳﺖ ﻫﺎ ﺑﺮﺍﻱ ﻗﻔﻞ ﻫﺎﻱ ﺍﻧﺤﺼﺎﺭﻱ ﻧﮕﻪ ﺩﺍﺷﺘﻪ ﻣﻲ ﺷﻮﺩ ‪ .‬ﻗﻔﻞ ﻫﺎﻱ ﻣﺸﺘﺮﻙ‬
‫ﺯﻣﺎﻧﻴﻜﻪ ﻳﻚ ﺗﺮﺍﻛﻨﺶ ﺑﺎﻳﺴﺘﻲ ﺁﻳﺘﻢ ‪ Q‬ﺩﺍﺩﻩ ﺭﺍ ﻣﺴﺪﻭﺩ ﻛﻨﺪ ‪ ،‬ﻓﻘﻂ ﻗﻔﻠﻲ ﺑﺮ ‪ Q‬ﺍﺯ ﻣﺪﻳﺮ ﻗﻔﻞ ﺩﺭ ﻳﻚ‬
‫ﺳﺎﻳﺖ ﻣﺤﺘﻮﻱ ﻛﭙﻲ ‪ Q‬ﺭﺍ ﺩﺭﺧﻮﺍﺳﺖ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫ﻗﻔﻞ ﻫﺎﻱ ﺍﻧﺤﺼﺎﺭﻱ ‪ :‬ﺯﻣﺎﻧﻴﻜﻪ ﺗﺮﺍﻛﻨﺶ ﺑﺎﻳﺴﺘﻲ ﺁﻳﺘﻢ ‪ Q‬ﺩﺍﺩﻩ ﺭﺍ ﻣﺴﺪﻭﺩ ﻛﻨﺪ ‪،‬ﻗﻔﻞﻱ ﺑﺮ ‪ Q‬ﺍﺯ‬
‫ﻣﺪﻳﺮ ﻗﻔﻞ ﺩﺭ ﺗﻤﺎﻣﻲ ﺳﺎﻳﺖ ﻫﺎﻱ ﻣﺤﺘﻮﻱ ﻛﭙﻲ ‪ Q‬ﺭﺍ ﺩﺭﺧﻮﺍﺳﺖ ﻣﻲ ﻛﻨﺪ‬

‫ﻣﺰﻳﺖ‪ :‬ﺳﺮﺑﺎﺭ ﻛﻤﺘﺮﻱ ﺑﺮ ﻋﻤﻠﻴﺎﺕ ﺧﻮﺍﻧﺪﻧﻲ ﺭﺍ ﺗﺤﻤﻴﻞ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫ﻋﻴﺐ‪ :‬ﺳﺮﺑﺎﺭ ﺍﺿﺎﻓﻲ ﺑﺮ ﺛﺒﺖ ﻫﺎ‬

‫‪125‬‬
 ‫ﭘﺮﻭﺗﻜﻞ ﻣﻮﺍﻓﻘﺖ ﻋﻤﻮﻣﻲ ﺣﺪﻧﺼﺎﺏ‬
‫ﻋﻤﻮﻡﺍ ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﺍﻛﺜﺮﻳﺖ ﻭ ﺑﺎﻳﺎﺱ ﺷﺪﻩ ﻫﺮ ﺳﺎﻳﺖ ﺑﺎ ﻳﻚ ﻭﺯﻥ ﺗﻌﻴﻴﻦ ﻣﻲ ﺷﻮﺩ ‪ .‬ﺍﺟﺎﺯﻩ ﺩﻫﻴﺪ ‪ S‬ﻭﺯﻥ‬
‫ﺗﻤﺎﻣﻲ ﺳﺎﻳﺖ ﻫﺎ ﺑﺎﺷﺪ ‪ .‬ﺩﻭﻣﻘﺪﺍﺭﺣﺪﻧﺼﺎﺏ ﺧﻮﺍﻧﺪﻧﻲ ‪ Qr‬ﻭ ﺣﺪﻧﺼﺎﺏ ﺛﺒﺖ ‪ Qw‬ﺭﺍ ﺍﻧﺘﺨﺎﺏ ﻛﻨﻴﺪ ‪.‬‬
‫ﭼﻨﺎﻧﭽﻪ ‪ Qr + Qw > S‬ﻭ ‪ ، 2 * Qw > S‬ﺣﺪﻧﺼﺎﺏ ﻫﺎ ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﻪ ﻃﻮﺭ ﻣﺠﺰﺍ ﺑﺮﺍﻱ ﻫﺮ ﺁﻳﺘﻢ‬
‫ﺍﻧﺘﺨﺎﺏ ﺷﻮﺩ )ﻭ ‪ S‬ﻣﺤﺎﺳﺒﻪ ﺷﻮﺩ (‪ ،‬ﻫﺮ ﺧﻮﺍﻧﺪﻥ ﺑﺎﻳﺪ ﺗﻜﺮﺍﺭﻫﺎﻱ ﻛﺎﻓﻲ ﺭﺍ ﻣﺴﺪﻭﺩﻛﻨﺪ ﻛﻪ ﻣﺠﻤﻮﻉ ﻭﺯﻥ‬
‫ﻫﺎﻱ ﺳﺎﻳﺖ ﺑﻴﺸﺘﺮ ﻳﺎ ﻣﺴﺎﻭﻱ ‪ Qr‬ﺍﺳﺖ‪ .‬ﻫﺮ ﺧﻮﺍﻧﺪﻥ ﺑﺎﻳﺪ ﺗﻜﺮﺍﺭﻫﺎﻱ ﻛﺎﻓﻲ ﺭﺍ ﻣﺴﺪﻭﺩﻛﻨﺪ ﻛﻪ ﻣﺠﻤﻮﻉ‬
‫ﻭﺯﻥ ﻫﺎﻱ ﺳﺎﻳﺖ ﺑﻴﺸﺘﺮ ﻳﺎ ﻣﺴﺎﻭﻱ ‪ Qw‬ﺍﺳﺖ‪ .‬ﺩﺭ ﺣﺎﻝ ﺣﺎ ﺿﺮ ﻓﺮﺽ ﻣﻲ ﻛﻨﻴﻢ ﻛﻪ ﺗﻤﺎﻣﻲ ﺗﻜﺮﺍﺭﻫﺎ‬
‫ﻣﻜﺘﻮﺏ ﻣﻲ ﺑﺎﺷﺪ ‪ .‬ﺑﺴﻂ ﻫﺎﻳﻲ ﺟﻬﺖ ﻣﺠﺎﺯ ﻛﺮﺩﻥ ﭼﻨﺪ ﺳﺎﻳﺖ ﺑﺎ ﺭﺍﻱ ﻏﻴﺮ ﻗﺎﺑﻞ ﺩﺳﺘﺮﺱ ﺑﻮﺩﻥ ﺑﻌﺪﺍ‬
‫ﺗﻮﺻﻴﻒ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﻧﺸﺎﻥ ﺯﻣﺎﻧﻲ‬
‫ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﻛﻨﺘﺮﻝ ﻫﻤﮕﺎﻫﻲ ﻣﺒﺘﻨﻲ ﺑﺮﻧﺸﺎﻥ ﺯﻣﺎﻧﻲ ﻣﻲ ﺗﻮﺍﻧﺪ ﺩﺭ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﺗﻮﺯﻳﻊ ﺷﺪﻩ ﺍﺳﺘﻔﺎﺩﻩ‬
‫ﺷﻮﺩ‪ .‬ﻫﺮ ﺗﺮﺍﻛﻨﺶ ﺑﺎﻳﺴﺘﻲ ﻧﺸﺎﻥ ﺯﻣﺎﻧﻲ ﻣﻨﺤﺼﺮ ﺑﻪ ﻓﺮﺩﻱ ﺭﺍ ﺍﺭﺍﺋﻪ ﺩﻫﺪ ‪.‬‬

‫ﻣﺸﻜﻞ ﺍﺻﻠﻲ ‪ :‬ﭼﮕﻮﻧﮕﻲ ﺍﻳﺠﺎﺩ ﻧﺸﺎﻥ ﺯﻣﺎﻧﻲ ﺩﺭ ﺣﺎﻟﺖ ﺗﻮﺯﻳﻊ ﺷﺪﻩ‬

‫– ﻫﺮ ﺳﺎﻳﺖ ﻧﺸﺎﻥ ﺯﻡ ﺍ ﻧﻲ ﻣﺤﻠﻲ ﻣﻨﺤﺼﺮ ﺑﻪ ﻓﺮﺩﻱ ﺭﺍ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺷﻤﺎﺭﺷﮕﺮ ﻣﻨﻄﻘﻲ ﻳﺎ‬

‫ﻗﻔﻞ ﻣﺤﻠﻲ ﺍﻳﺠﺎﺩ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫– ﻧﺸﺎﻥ ﺯﻣﺎﻧﻲ ﻣﻨﺤﺼﺮ ﺑﻪ ﻓﺮﺩ ﺟﻬﺎﻧﻲ ﺑﺎ ﺍﻟﺤﺎﻕ ﻛﺮﺩﻥ ﻧﺸﺎﻥ ﺯﻣﺎﻧﻲ ﻣﺤﻠﻲ ﻣﻨﺤﺼﺮ ﺑﻪ ﻓﺮﺩ‬
‫ﺑﺎ ﺷﻨﺎﺳﺎﮔﺮ ﻣﻨﺤﺼﺮ ﺑﻪ ﻓﺮﺩ ﺑﺪﺳﺖ ﻣﻲ ﺁﻳﺪ‪.‬‬

‫‪126‬‬
‫ﺳﺎﻳﺘﻲ ﺑﺎ ﺳﺎﻋﺖ ﺁﻫﺴﺘﻪ ﻧﺸﺎﻥ ﻫﺎﻱ ﺯ ﻣﺎﻧﻲ ﻛﻮﭼﺘﺮ ﺭﺍ ﺗﻌﻴﻴﻦ ﺧﻮﺍﻫﺪ ﻛﺮﺩ ‪ .‬ﺍﻣﺎ ﺍﺯ ﻝﺣﺎﻅ ﻣﻨﻄﻘﻲ ﺩﺭﺳﺖ ‪:‬‬
‫ﻫﻨﻮﺯﻗﺎﺑﻠﻴﺖ ﺳﺮﻳﺎﻟﻲ ﺷﺪﻥ ﺗﺤﺖ ﺗﺎﺛﻴﺮ ﻧﻴﺴﺖ ‪.‬‬

‫ﺟﻬﺖ ﺗﺜﺒﻴﺖ ﺍﻳﻦ ﻣﺸﻚﻝ ‪:‬‬

‫♦ ﺩﺭ ﻣﺤﺪﻭﺩﻩ ﻫﺮ ﺳﺎﻳﺖ ‪ Si‬ﺳﺎﻋﺖ ﻣﻨﻄﻘﻲ )‪ (LCi‬ﺭﺍ ﺗﻌﺮﻳﻒ ﻛﻨﻴﺪ ﻛﻪ ﻧﺸﺎﻥ ﺯﻣﺎﻧﻲ ﻣﺤﻠﻲ‬
‫ﻣﻨﺤﺼﺮ ﺑﻪ ﻓﺮﺩﻱ ﺭﺍ ﺍﻳﺠﺎﺩ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫♦ ﻧﻴﺎﺯ ﺩﺍﺭﺩ ﻛﻪ ‪ Si‬ﭘﻴﺸﺎﭘﻴﺶ ﺳﺎﻋﺖ ﻣﻨﻄﻘﻲ ﺍﺵ ﺩﺭ ﻫﺮ ﺟﺎﻳﻲ ﺑﺎﺷﺪ ﻛﻪ ﺩﺭﺧﻮﺍﺳﺘﻲ ﺍﺯ‬

‫ﺗﺮﺍﻛﻨﺶ ‪ Ti‬ﺑﺎ ﻧﺸﺎﻥ ﺯﻣﺎﻧﻲ >‪ < x,y‬ﺩﺭﻳﺎﻓﺖ ﻣﻲ ﺷﻮﺩ ﻭ ‪ x‬ﺑﻴﺸﺘﺮ ﺍﺯ ﻣﻘﺪﺍﺭ ﻛﻨﻮﻧﻲ‬
‫‪ LCi‬ﺍﺱﺕ ‪.‬‬

‫♦ ﺩﺭ ﺍﻳﻦ ﻣﻮﺭﺩ ‪ ،‬ﺳﺎﻳﺖ ﺍﺯ ﺳﺎﻋﺖ ﻣﻨﻄﻘﻲ ﺍﺵ ﺑﻪ ﻣﻘﺪﺍﺭ ‪ x + 1‬ﭘﻴﺸﺮﻭﻱ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫ﺗﻜﺮﺍﺭ ﺑﺎ ﺳﺎﺯﮔﺎﺭﻱ ﺿﻌﻴﻒ‬

‫ﭼﻨﺪﻳﻦ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺗﺠﺎﺭﻱ ﺗﻜﺮﺍﺭ ﺩﺍﺩﻩ ﺑﺎ ﺩﺭﺟﻪ ﻫﺎﻱ ﺿﻌﻴﻒ ﺳﺎﺯﮔﺎﺭﻱ )ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ ﺑﺪﻭﻥ ﺿﻤﺎﻧﺖ‬
‫ﺭﻭﺯﺭﺳﺎﻧﻲ ﻫﺎ ﺩﺭ ﻓﺎﻳﻞ‬
‫ﻗﺎﺑﻠﻴﺖ ﺳﺮﻳﺎﻟﻲ ( ﺭﺍ ﺣﻤﺎﻳﺖ ﻣﻲ ﻛﻨﺪ ‪ .‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ ﺗﻜﺮﺍﺭ ﺍﺻﻠﻲ ‪ /‬ﺩﻧﺒﺎﻟﻪ ﺭﻭ ‪ :‬ﺏ‬
‫ﺍﺻﻠﻲ ﺧﺎﺹ ﻡﻱ ﺷﻮﺩ ﻭ ﺑﻪ ﺳﺎﻳﺖ ﻫﺎﻱ ﺩﻧﺒﺎﻟﻪ ﺭﻭ ﭘﺨﺶ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﻛﭙﻲ ﻫﺎ ﺑﺎﻳﺴﺘﻲ ﺭﻭﻧﻮﺷﺖ ﻟﺤﻈﻪ ﺍﻱ ﺳﺎﺯﮔﺎﺭ ﺑﺎ ﺗﺮﺍﻛﻨﺶ ﺍﺯ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺭﺍ ﻣﺸﺎﻫﺪﻩ ﻛﻨﻨﺪ ‪ .‬ﻳﻌﻨﻲ ﺣﺎﻟﺘﻲ‬
‫ﺍﺯ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺑﺎﺯﺗﺎﺏ ﺩﻫﻨﺪﻩ ﺗﻤﺎﻣﻲ ﺍﺛﺮﺍﺕ ﺗﻤﺎﻣﻲ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﺗﺎ ﭼﻨﺪ ﻣﺮﺣﻠﻪ ﺩﺭ ﺗﺮﺗﻴﺐ ﺳﺮﻳﺎﻟﻲ ﻭ‬
‫ﻫﻴﭻ ﺗﺎﺙﻳﺮﻱ ﺍﺯ ﺩﻳﮕﺮ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻱ ﺑﻌﺪﻱ ‪ ،‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ ‪ :‬ﺍﻭﺭﺍﻛﻞ ﺟﻤﻠﻪ ﺍﻳﺠﺎﺩ ﺭﻭﻧﻮﺷﺖ ﻟﺤﻈﻪ ﺍﻱ ﺭﺍ‬
‫ﺑﺮﺍﻱ ﺍﻳﺠﺎﺩ ﺭﻭﻧﻮﺷﺖ ﻟﺤﻈﻪ ﺍﻱ ﺍﺯ ﻳﻚ ﺍﺭﺗﺒﺎﻁ ﻳﺎ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺍﺭﺗﺒﺎﻃﺎﺕ ﺩﺭ ﺳﺎﻳﺖ ﺩﻭﺭ ﻓﺮﺍﻫﻢ ﻣﻲ‬
‫ﺁﻭﺭﺩ‪ .‬ﺭﻭﻧﻮﺷﺖ ﻟﺤﻈﻪ ﺍﻱ ﺑﺎ ﻣﺤﺎﺳﺒﻪ ﻣﺠﺪﺩ ﻳﺎ ﺑﺎ ﺑﻪ ﺭﻭﺯ ﺭﺳﺎﻧﻲ ﺍﻓﺰﺍﻳﺸﻲ ﻣﺠﺪﺩﺍ ﺗﺎﺯﻩ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫‪127‬‬
 ‫ﮔﺮﺩﺍﻧﺶ ﺑﻦ ﺑﺴﺖ‬
‫ﺩﻭ ﺗﺮﺍﻛﻨﺶ ﺯﻳﺮ ﻭ ﭘﻴﺸﻴﻨﻪ ﺍﻱ ﺑﺎ ﺁﻳﺘﻢ ‪ X‬ﻭ ﺗﺮﺍﻛﻨﺶ‪ T1‬ﺩﺭﺳﺎﻳﺖ ‪ 1‬ﻭ ﺁﻳﺘﻢ ‪ y‬ﻭ ﺗﺮﺍﻛﻨﺶ ‪T2‬ﺩﺭ ﺳﺎﻳﺖ‬
‫‪2‬ﺭﺍ ﺩﺭ ﻧﻈﺮ ﺑﮕﻴﺮﻳﺪ‪.‬‬

‫ﻧﺘﻴﺠﻪ ‪ :‬ﺑﻦ ﺑﺴﺘﻲ ﻛﻪ ﻧﻤﻲ ﺗﻮﺍﻧﺪ ﺑﻪ ﻃﻮﺭ ﻣﺤﻠﻲ ﺩﺭ ﻫﺮ ﺳﺎﻳﺖ ﺷﻨﺎﺳﺎﻳﻲ ﺷﻮﺩ ‪.‬‬

‫ﭘﻴﻜﺮﺑﻨﺪﻱ ﻣﺠﺪﺩ‬
‫ﻣﺘﻮﻗﻒ ﻛﺮﺩﻥ ﺗﻤﺎﻣﻲ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻳﻲ ﻛﻪ ﺩﺭ ﺳﺎﻳﺖ ﺧﺮﺍﺏ ﺷﺪﻩ ﻓﻌﺎﻝ ﺑﻮﺩ ‪ .‬ﺍﻧﺘﻈﺎﺭ ﺁﻧﻬﺎ ﻣﻲ ﺗﻮﺍﻧﺴﺖ ﺑﺎ‬
‫ﺩﻳﮕﺮ ﺗﺮﺍﻛﻨﺶ ﻫﺎ ﻣﺪﺍﺧﻠﻪ ﻛﻨﺪ ﭼﺮﺍ ﻛﻪ ﺁﻧﻬﺎ ﻣﻤﻜﻦ ﺍﺳﺖ ﻗﻔﻞ ﻫﺎﻳﻲ ﺑﻪ ﺩﻳﮕﺮ ﺳﺎﻳﺖ ﻫﺎ ﺭﺍ ﻧﮕﻪ ﺩﺍﺭﻧﺪ ‪ .‬ﺑﺎ‬
‫ﻭﺟﻮﺩ ﺍﻳﻦ ﺩﺭ ﻣﻮﺭﺩﻱ ﻛﻪ ﻓﻘﻂ ﭼﻨﺪ ﻛﭙﻲ ﺍﺯ ﻳﻚ ﺁﻳﺘﻢ ﺩﺍﺩﻩ ﺷﻜﺴﺖ ﻣﻲ ﺧﻮﺭﺩ ‪ ،‬ﻣﻤﻜﻦ ﺍﺳﺖ ﺟﻬﺖ‬
‫ﺍﺩﺍﻣﻪ ﺩﺍﺩﻥ ﺗﺮﺍﻛﻨﺶ ﻫﺎﻳﻲ ﻋﻤﻠﻲ ﻭ ﻣﻤﻜﻦ ﺑﺎﺷﺪ ﻛﻪ ﺩﺭ ﺳﺎﻳﺖ ﺧﺮﺍﺏ ﺷﺪﻩ ﺑﻪ ﺩﺍﺩﻩ ﺩﺳﺖ ﻳﺎﻓﺘﻪ ﺍﺳﺖ ‪.‬‬
‫ﺍﮔﺮ ﺍﻳﺘﻢ ﻫﺎﻱ ﺩﺍﺩﻩ ﺗﻜﺮﺍﺭ ﺷﺪﻩ ﺩﺭ ﺳﺎﻳﺖ ﺧﺮﺍﺏ ﺷﺪﻩ ﺑﺎﺷﺪ ‪،‬ﻛﺎﺗﺎﻟﻮگ ﺳﻴﺴﺘﻢ ﺟﻬﺖ ﺭﻓﻊ ﺍﻧﻬﺎ ﺍﺯ ﻟﻴﺴﺖ‬
‫ﻛﭙﻲ ﻫﺎ ﺑﻪ ﺭﻭﺯ ﻣﻲ ﺷﻮﺩ‪ .‬ﺯﻣﺎﻧﻲ ﻛﻪ ﺳﺎﻳﺖ ﺧﺮﺍﺏ ﺷﺪﻩ ﺑﻬﺒﻮﺩ ﻣﻲ ﻳﺎﺑﺪ ‪ ،‬ﺑﺎﻳﺴﺘﻲ ﺑﺮﻋﻜﺲ ﺷﻮﺩ ﺍﻣﺎ ﺩﻗﺖ‬
‫ﺍﺿﺎﻓﻲ ﺑﺎﻳﺴﺘﻲ ﺑﺮﺍﻱ ﺑﻪ ﺭﻭﺯ ﻛﺮﺩﻥ ﻣﻘﺎﺩﻳﺮ ﺍﻧﺠﺎﻡ ﺷﻮﺩ ‪ .‬ﺍﮔﺮ ﺳﺎﻳﺖ ﺧﺮﺍﺏ ﺵﺩﻩ ﺳﺮﻭﺭ ﻣﺮﻛﺰﻱ ﺑﺮﺍﻱ ﭼﻨﺪ‬
‫ﺳﻴﺴﺘﻢ ﺑﺎﺷﺪ ﺟﻬﺖ ﺗﻌﻴﻴﻦ ﺳﺮﻭﺭ ﺟﺪﻳﺪ ﺑﺎﻳﺴﺘﻲ ﮔﺰﻳﻨﺶ ﺩﺍﺷﺖ ‪ .‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ ﺳﺮﻭﺭ ﻧﺎﻡ ‪ ،‬ﻫﻤﺎﻫﻨﮓ‬
‫ﻛﻨﻨﺪﻩ ﻫﻤﮕﺎﻫﻲ ‪،‬ﺷﻨﺎﺳﺎﻳﻲ ﻛﻨﻨﺪﻩ ﺑﻦ ﺑﺴﺖ ﺟﻬﺎﻧﻲ ‪.‬‬

‫‪128‬‬
‫ﺍﺯ ﺁﻧﺠﺎ ﻛﻪ ﺍﻓﺮﺍﺯ ﻛﺮﺩﻥ ﺷﺒﻜﻪ ﻣﻤﻜﻦ ﻧﻴﺴﺖ ﺍﺯ ﺷﻜﺴﺖ ﺳﺎﻳﺖ ﻗﺎﺑﻞ ﺗﺸﺨﻴﺺ ﺑﺎﺷﺪ ‪ ،‬ﻣﻮﻗﻌﻴﺖ ﻫﺎﻱ ﺯﻳﺮ‬
‫ﺑﺎﻳﺴﺘﻲ ﺍﺟﺘﻨﺎﺏ ﺷﻮﺩ‪:‬‬

‫♦ ﺩﻭ ﻳﺎ ﭼﻨﺪ ﺳﺮﻭﺭ ﻣﺮﻛﺰﻱ ﺍﻧﺘﺨﺎﺏ ﺷﺪﻩ ﺩﺭ ﺍﻓﺮﺍﺯ ﻫﺎﻱ ﻣﺠﺰﺍﻭ ﻣﺸﺨﺺ ‪.‬‬

‫♦ ﺑﻴﺶ ﺍﺯ ﻳﻚ ﺍﻓﺮﺍﺯ ‪،‬ﺁﻳﺘﻢ ﺩﺍﺩﻩ ﺗﻜﺮﺍﺭ ﺷﺪﻩ ﺭﺍ ﺑﻪ ﺭﻭﺯ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫ﺭﻭﺯﺭﺳﺎﻧﻲ ﻫﺎ ﺑﺎﻳﺴﺘﻲ ﻗﺎﺩﺭ ﺑﻪ ﺍﺩﺍﻣﻪ ﺩﺍﺩﻥ ﺑﺎﺷﺪ ﺣﺘﻲ ﺍﮔﺮ ﭼﻨﺪ ﺳﺎﻳﺖ ﺍﺯﻛﺎﺭﺍﻓﺘﺎﺩﻩ ﺑﺎﺷﺪ ‪.‬‬
‫ﺏ‬

‫ﺭﺍﻩ ﺣﻞ ‪ :‬ﺭﻭﻳﻜﺮﺩ ﻣﺒﺘﻨﻲ ﺑﺮ ﺍﻛﺜﺮﻳﺖ‪:‬‬

‫ﺟﺎﻳﮕﺰﻳﻦ " ﺧﻮﺍﻧﺪﻥ ﻳﻚ ﺛﺒﺖ ﻛﺎﻣﻼ ﺩﺭﺩﺳﺘﺮﺱ " ﺁﺯﺍﺭ ﻣﻲ ﺑﻴﻨﺪ ﺍﻣﺎ ﺑﺎﻋﺚ ﻣﺸﻜﻼﺗﻲ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫ﺭﻭﻳﻜﺮﺩ ﻣﺒﺘﻨﻲ ﺑﺮ ﺍﻛﺜﺮﻳﺖ‬

‫‪ n‬ﭘﺮﻭﺗﻜﻞ ﺍﻛﺜﺮﻳﺖ ﺑﺮﺍﻱ ﻛﻨﺘﺮﻝ ﻫﻤﮕﺎﻫﻲ ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﺮﺍﻱ ﻛﺎﺭ ﺍﺻﻼﺡ ﺷﻮﺩ ﺣﺘﻲ ﺍﮔﺮ ﭼﻨﺪ ﺳﺎﻳﺖ‬
‫ﺩﺭﺩﺳﺘﺮﺱ ﻧﺒﺎﺷﺪ‪ .‬ﻛﭙﻲ ﻫﺮ ﺁﻳﺘﻢ ﺩﺍﺭﺍﻱ ﻳﻚ ﺷﻤﺎﺭﻩ ﻧﺴﺨﻪ ﺍﺳﺖ ﺯﻣﺎﻧﻲ ﺑﺮﻭﺯ ﻣﻲ ﺷﻮﺩ ﻛﻪ ﻛﭙﻲ‬
‫ﺑﺮﻭﺯ ﻣﻲ ﺷﻮﺩ ﻛﻪ ﺩﺭ ﺯﻳﺮ ﻣﻄﺮﺡ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫– ﺩﺭﺧﻮﺍﺳﺖ ﻗﻔﻞ ﺑﻪ ﺣﺪﺍﻗﻞ ﻧﺼﻒ ﺳﺎﻳﺖ ﻫﺎﻳﻲ ﻓﺮﺳﺘﺎﺩﻩ ﻣﻲ ﺷﻮﺩ ﻛﻪ ﺩﺭ ﺍﻥ ﻛﭙﻲ ﻫﺎﻱ‬
‫ﺁﻳﺘﻢ ﺫﺧﻴﺮﻩ ﻣﻲ ﺷﻮﺩ ﻭ ﻋﻤﻠﻴﺎﺕ ﻓﻘﻂ ﺯﻣﺎﻧﻲ ﺍﺩﺍﻣﻪ ﻣﻲ ﻳﺎﺑﺪ ﻛﻪ ﻳﻚ ﻗﻔﻞ ﺩﺭ ﺍﻛﺜﺮﻳﺖ‬
‫ﺳﺎﻳﺖ ﻫﺎ ﺑﺪﺳﺖ ﻣﻲ ﺁﻳﺪ‪.‬‬

‫ﺑﺎ‬ ‫– ﻋﻤﻠﻴﺎﺕ ﺧﻮﺍﻧﺪﻥ ﺗﻤﺎﻣﻲ ﻛﭙﻲ ﻫﺎﻱ ﻗﻔﻞ ﺷﺪﻩ ﺭﺍ ﺑﺮﺭﺳﻲ ﻣﻲ ﻛﻨﺪ ﻭ ﻣﻘﺪﺍﺭ ﺍﺯ ﻛﭙﻲ‬
‫ﺑﺰﺭﮔﺘﺮﻳﻦ ﺷﻤﺎﺭﻩ ﻧﺴﺨﻪ ﺭﺍ ﻣﻲ ﺧﻮﺍﻧﺪ ‪ .‬ﻣﻤﻜﻦ ﺍﺳﺖ ﺍﻳﻦ ﻣﻘﺪﺍﺭ ﺭﺍ ﺑﻨﻮﻳﺴﺪ ﻭ ﺷﻤﺎﺭﻩ‬
‫ﻧﺴﺨﻪ ﺑﻪ ﻛﭙﻲ ﻫﺎﻳﻲ ﺑﺎ ﺷﻤﺎﺭﻩ ﻫﺎﻱ ﻧﺴﺨﻪ ﻛﻤﺘﺮ )ﺑﺪﻭﻥ ﻧﻴﺎﺯ ﺑﻪ ﻛﺴﺐ ﻗﻔﻞ ﻫﺎ ﺩﺭ‬
‫ﺗﻤﺎﻣﻲ ﻛﭙﻲ ﻫﺎ ﺑﺮﺍﻱ ﺍﻳﻦ ﻛﺎﺭ ( ﺑﺮﻣﻲ ﮔﺮﺩﺩ‪.‬‬

‫– ﺛﺒﺖ ﻋﻤﻠﻴﺎﺕ ‪ :‬ﺑﺎﻻﺗﺮﻳﻦ ﺷﻤﺎﺭﻩ ﻧﺴﺨﻪ ﻣﺎﻧﻨﺪ ﺧﻮﺍﻧﺪﻥ ﺭﺍ ﭘﻴﺪﺍ ﻛﻨﻴﺪ ﻭ ﺷﻤﺎﺭﻩ ﻧﺴﺨﻪ‬
‫ﺟﺪﻳﺪ ﺑﻪ ﺑﺎﻻﺗﺮﻳﻦ ﻧﺴﺨﻪ ﻗﺪﻳﻤﻲ ‪ 1+‬ﺭﺍﺗﻌﻴﻴﻦ ﻛﻨﻴﺪ ‪ .‬ﺳﭙﺲ ﺛﺒﺖ ﻫﺎ ﺩﺭ ﺗﻤﺎﻣﻲ ﻛﭙﻲ‬

‫‪129‬‬
‫ﻫﺎﻱ ﻗﻔﻞ ﺷﺪﻩ ﺍﻧﺠﺎﻡ ﻣﻲ ﺷﻮﺩ ﻭ ﺷﻤﺎﺭﻩ ﻧﺴﺨﻪ ﺩﺭ ﺍﻳﻦ ﻛﭙﻲ ﻫﺎ ﺑﻪ ﺷﻤﺎﺭﻩ ﻧﺴﺨﻪ ﺟﺪﻳﺪ‬
‫ﺗﻌﻴﻴﻦ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫– ﺷﻜﺴﺖ ﻫﺎ )ﺷﺒﻜﻪ ﻭ ﺳﺎﻳﺖ ( ﺑﺎﻋﺚ ﺑﺮﻭﺯ ﻣﺸﻜﻞ ﻧﻤﻲ ﺷﻮﺩ ﻣﺎﺩﺍﻣﻲ ﻛﻪ ‪ :‬ﺳﺎﻳﺖ ﻫﺎ ﺩﺭ‬
‫ﻭﺍﮔﺬﺍﺭﻱ ﺷﺎﻣﻞ ﺍﻛﺜﺮﻳﺖ ﻛﭙﻲ ﻫﺎ ﺍﺯ ﻫﺮ ﺁﻱ ﺗﻢ ﺩﺍﺩﻩ ﺑﻪ ﺭﻭﺯ ﺷﺪﻩ ﻣﻲ ﺑﺎﺷﺪ ‪ .‬ﺩﺭﻃﻲ‬
‫ﺧﻮﺍﻧﺪﻥ ‪،‬ﺍﻛﺜﺮﻳﺖ ﻛﭙﻲ ﻫﺎ ﺟﻬﺖ ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﺷﻤﺎﺭﻩ ﻫﺎﻱ ﻧﺴﺨﻪ ﺩﺭﺩﺳﺘﺮﺱ ﻣﻲ ﺑﺎﺷﺪ ‪.‬‬
‫ﻣﻮﺿﻮﻋﻲ ﺑﻪ ﺑﺎﻻ ‪،‬ﻭﺍﮔﺬﺍﺭﻱ ﺩﻭﻓﺎﺯﻱ ﻣﻲ ﺗﻮﺍﻧﺪ ﺟﻬﺖ ﺑﻪ ﺭﻭﺯ ﻛﺮﺩﻥ ﻛﭙﻲ ﻫﺎ ﺍﺳﺘﻔﺎﺩﻩ‬
‫ﺷﻮﺩ‪.‬‬

‫ﺗﺬﻛﺮ‪:‬ﺧﻮﺍﻧﺪﻥ ﻫﺎ ﺟﻬﺖ ﻣﺸﺎﻫﺪﻩ ﺁﺧﺮﻳﻦ ﻧﺴﺨﻪ ﺁﻳﺘﻢ ﺩﺍﺩﻩ ﺿﻤﺎﻧﺖ ﻣﻲ ﺷﻮﺩ ‪ .‬ﻳﻜﭙﺎﺭﭼﮕﻲ ﻣﺠﺪﺩ ﺑﺪﻱﻫﻲ‬
‫ﺍﺳﺖ ‪:‬ﻫﻴﭻ ﭼﻴﺰ ﻧﺒﺎﻳﺪ ﺍﻧﺠﺎﻡ ﺷﻮﺩ‪.‬‬

‫*ﺍﻟﮕﻮﺭﻳﺘﻢ ﻣﻮﺍﻓﻘﺖ ﻋﻤﻮﻣﻲ ﺣﺪﻧﺼﺎﺏ ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﻪ ﻃﻮﺭﻣﺸﺎﺑﻪ ﮔﺴﺘﺮﺩﻩ ﺷﻮﺩ ‪.‬‬

‫ﺍﻟﮕﻮﺭﻳﺘﻢ ‪Bully‬‬
‫‪ Θ‬ﺍﮔﺮ ﺳﺎﻳﺖ ‪ Si‬ﺩﺭﺧﻮﺍﺳﺘﻲ ﺭﺍ ﺑﻔﺮﺳﺘﺪ ﻛﻪ ﺍﻟﺒﺘﻪ ﺗﻮﺳﻂ ﻫﻤﺎﻫﻨﮓ ﻛﻨﻨﺪﻩ ﺩﺭﻣﺤﺪﻭﺩﻩ ﻓﺎﺻﻠﻪ ﺯﻣﺎﻧﻲ‬
‫‪ T‬ﭘﺎﺳﺦ ﺩﺍﺩﻩ ﻧﻤﻲ ﺷﻮﺩ ‪ ،‬ﻓﺮﺽ ﻛﻨﻴﺪ ﻛﻪ ﻫﻤﺎﻫﻨﮓ ﻛﻨﻨﺪﻩ ‪ Si‬ﺭﺍ ﺷﻜﺴﺖ ﺩﺍﺩﻩ ﺍﺳﺖ ‪،‬ﺳﻌﻲ‬
‫ﻣﻲ ﻛﻨﺪ ﺗﺎ ﺧﻮﺩﺵ ﺭﺍ ﺑﻪ ﻋﻨﻮﺍﻥ ﻫﻤﺎﻫﻨﮓ ﻛﻨﻨﺪﻩ ﺟﺪﻳﺪ ﺍﻧﺘﺨﺎﺏ ﻛﻨﺪ ‪.‬‬

‫‪ Si Θ‬ﭘﻴﺎﻡ ﺍﻧﺘﺨﺎﺏ ﺭﺍ ﺑﻪ ﻫﺮ ﺳﺎﻳﺖ ﺑﺎ ﻋﺪﺩ ﺷﻨﺎﺳﺎﻳﻲ ﺑﺎﻻﺗﺮ ﻣﻲ ﻓﺮﺳﺘﺪ‪ ،‬ﺳﭙﺲ ‪ Si‬ﻣﻨﺘﻈﺮ ﻫﺮ ﻳﻚ‬
‫ﺍﺯ ﺍﻳﻦ ﻓﺮﺍﻳﻨﺪ ﻫﺎ ﺑﺮﺍﻱ ﭘﺎﺳﺦ ﺩﺍﺩﻥ ﺩﺭ ﻣﺤﺪﻭﺩﻩ ‪ T‬ﻣﻲ ﻣﺎﻧﺪ‪.‬‬

‫‪ Θ‬ﺍﮔﺮ ﻫﻴﭻ ﭘﺎﺳﺨﻲ ﺩﺭﻣﺤﺪﻭﺩﻩ ‪ T‬ﻭﺟﻮﺩ ﻧﺪﺍﺷﺘﻪ ﺑﺎﺷﺪ ‪ ،‬ﺑﭙﺬﻳﺮﻳﺪ ﻛﻪ ﺗﻤﺎﻣﻲ ﺳﺎﻳﺖ ﻫﺎﻳﻲ ﺑﺎ ﻋﺪﺩ‬
‫ﺑﻴﺸﺘﺮ ﺍﺯ ‪ i‬ﺷﻜﺴﺖ ﺧﻮﺭﺩﻩ ﺍﻧﺪ ‪ Si ،‬ﺧﻮﺩﺵ ﺭﺍ ﻫﻤﺎﻫﻨﮓ ﻛﻨﻨﺪﻩ ﺟﺪﻳﺪ ﺍﻧﺘﺨﺎﺏ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫‪ Θ‬ﺍﮔﺮ ﭘﺎﺳﺦ ﺩﺭﻳﺎﻓﺖ ﺷﻮﺩ‪ Si ،‬ﻓﺎﺻﻠﻪ ﺯﻣﺎﻧﻲ ‪ T‬ﺭﺍ ﺷﺮﻭﻉ ﻣﻲ ﻛﻨﺪ‪ ،‬ﻣﻨﺘﻈﺮ ﺩﺭﻳﺎﻓﺖ ﭘﻴﺎﻣﻲ ﻣﻲ ﻣﺎﻧﺪ‬
‫ﻛﻪ ﺳﺎﻳﺖ ﺑﺎ ﻋﺪﺩ ﺷﻨﺎﺳﺎﻳﻲ ﺑﺎﻻﺗﺮ ﺍﻧﺘﺨﺎﺏ ﻛﺮﺩﻩ ﺍﺳﺖ‪.‬‬

‫‪130‬‬
 ‫‪ Θ‬ﺍﮔﺮ ﻫﻴﭻ ﭘﻴﺎﻣﻲ ﺩﺭﻡ ﺣﺪﻭﺩﻩ ‪ T‬ﻓﺮﺳﺘﺎﺩﻩ ﻧﺸﻮﺩ‪ ،‬ﺑﭙﺬﻳﺮﻳﺪ ﻛﻪ ﺳﺎﻳﺘﻲ ﺑﺎ ﻋﺪﺩ ﺑﺎﺍﻻﺗﺮ ﺷﻜﺴﺖ‬
‫ﺧﻮﺭﺩﻩ ﺍﺳﺖ ؛ ‪ Si‬ﺍﻟﮕﻮﺭﻳﺘﻢ ﺭﺍ ﻣﺠﺪﺩﺍ ﺷﺮﻭﻉ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫‪ Θ‬ﺑﻌﺪ ﺍﺯ ﺍﻧﻜﻪ ﺳﺎﻳﺖ ﺷﻜﺴﺖ ﺧﻮﺭﺩﻩ ﺑﻬﺒﻮﺩ ﻣﻲ ﻳﺎﺑﺪ‪ ،‬ﻓﻮﺭﺍ ﺍﺟﺮﺍﻱ ﺍﻟﮕﻮﺭﻳﺘﻢ ﻣﺸﺎﺑﻪ ﺭﺍ ﺷﺮﻭﻉ ﻣﻲ‬
‫ﻛﻨﺪ‪.‬‬

‫‪ Θ‬ﺍﮔﺮ ﻫﻴﭻ ﺳﺎﻳﺖ ﻓﻌﺎﻟﻲ ﺑﺎ ﺍﻋﺪﺍﺩ ﺑﺎﻻﺗﺮ ﻭﺟﻮﺩ ﻧﺪﺍﺷﺘﻪ ﺑﺎﺷﺪ ‪،‬ﺳﺎﻳﺖ ﺑﻬﺒﻮﺩﻳﺎﻓﺘﻪ ﺗﻤﺎﻣﻲ ﻓﺮﺍﻳﻨﺪﻫﺎ ﺍ‬
‫ﺍﻋﺪﺍﺩﻛﻤﺘﺮ ﺭﺍ ﻭﺍﺩﺍﺭ ﻣﻲ ﻛﻨﺪ ﺗﺎ ﺁﻧﺮﺍ ﻣﺠﺎﺯ ﻛﻨﺪ ﻳﻚ ﺳﺎﻳﺖ ﻫﻤﺎﻫﻨﮓ ﻛﻨﻨﺪﻩ ﺑﺎﺷﺪ ‪،‬ﺣﺘﻲ ﺍﮔﺮ‬
‫ﻫﻤﺎﻫﻨﮓ ﻛﻨﻨﺪﻩ ﺩﺭ ﺣﺎﻝ ﺣﺎﺿﺮ ﻓﻌﺎﻝ ﺑﺎ ﺍﻋﺪﺍﺩ ﻛﻤﺘﺮ ﻭﺟﻮﺩ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ ‪.‬‬

‫ﭘﺮﺩﺍﺯﺵ ﭘﺮﺱ ﻭ ﺟﻮ ﺗﻮﺯﻳﻊ ﺷﺪﻩ‬

‫ﺑﺮﺍﻱ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﻣﺘﻤﺮﻛﺰ ﺷﺪﻩ ﻣﻌﻴﺎﺭ ﺍﻭﻟﻴﻪ ﺑﺮﺍﻱ ﺑﺮﺁﻭﺭﺩ ﻫﺰﻳﻨﻪ ﺍﺳﺘﺮﺍﺗﮋﻱ ﺧﺎﺹ ‪،‬ﺗﻌﺪﺍﺩ ﺩﺳﺘﺮﺳﻲ‬
‫ﻫﺎﻱ ﺍﺣﺘﻤﺎﻝ ﺧﻄﺮ ﺍﺳﺖ‪ .‬ﺩﺭ ﻳﻚ ﺳﻴﺴﺘﻢ ﺗﻮﺯﻳﻊ ﺷﺪﻩ ‪ ،‬ﺩﻳﮕﺮ ﻣﻮﺿﻮﻋﺎﺕ ﺑﺎﻳﺴﺘﻲ ﺩﺭﻧﻈﺮ ﮔﺮﻓﺘﻪ ﺷﻮﺩ ‪:‬‬

‫ﻫﺰﻳﻨﻪ ﺍﻧﺘﻘﺎﻝ ﺩﺍﺩﻩ ﺩﺭ ﺷﺒﻜﻪ‬

‫ﺳﻮﺩ ﺑﺎﻟﻘﻮﻩ ﺩﺭ ﻋﻤﻠﻜﺮﺩ ﺍﺯ ﺩﺍﺷﺘﻦ ﭼﻨﺪﻳﻦ ﺳﺎﻳﺖ ‪،‬ﺑﺨﺶ ﻫﺎﻱ ﭘﺮﺱ ﻭ ﺟﻮ ﺩﺭ ﺗﻮﺍﺯﻱ ﺭﺍ‬
‫ﭘﺮﺩﺍﺯﺵ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫ﺗﻐﻴﻴﺮ ﭘﺮﺱ ﻭ ﺟﻮ‬

‫‪ ‬ﭘﺮﺱ ﻭ ﺟﻮ ﻫﺎﻱ ﺟﺒﺮﻱ ﺩﺭ ﺑﺨﺶ ﻫﺎ ﺭﺍ ﺗﺮﺟﻤﻪ ﻛﻨﻴﺪ ‪.‬‬

‫‪ l‬ﺑﺎﻳﺴﺘﻲ ﺟﻬﺖ ﺳﺎﺧﺘﺎﺭﺑﻨﺪﻱ ﺭﺍﺑﻄﻪ ‪r‬ﺍﺯ ﺑﺨﺶ ﻫﺎﻳﺶ ﻋﻤﻠﻲ ﻭ ﻣﻤﻜﻦ ﺑﺎﺷﺪ ‪.‬‬

‫‪ l‬ﺭﺍﺑﻄﻪ ‪ r‬ﺭﺍ ﺑﺎ ﻋﺒﺎﺭﺗﻲ ﺟﻬﺖ ﺳﺎﺧﺘﺎﺭﺑﻨﺪﻱ ﺭﺍﺑﻄﻪ ‪ r‬ﺍﺯ ﺑﺨﺶ ﻫﺎﻳﺶ ﺟﺎﻳﮕﺰﻳﻦ ﻛﻨﻴﺪ ‪.‬‬

‫‪ ‬ﺑﺨﺶ ﺑﻨﺪﻱ ﺍﻓﻘﻲ ﺭﺍﺑﻄﻪ ﺣﺴﺎﺏ ﺑﻪ ﺯﻳﺮ ﺭﺍ ﺩﺭﻧﻈﺮ ﺑﮕﻴﺮﻳﺪ‪:‬‬

‫‪(account) _name = “Hillside” account1 = σ branch‬‬

‫”‪view‬‬ ‫‪(account) _name = “Valley account2 = σ branch‬‬

‫‪131‬‬
 ‫‪ ‬ﭘﺮﺱ ﻭ ﺟﻮ ”‪ σ branch _name = “Hillside‬ﺑﺼﻮﺭﺕ ﺯﻳﺮ ﻣﻲ ﺷﻮﺩ ‪:‬‬

‫)‪σ branch _name = “Hillside” (account1 ∪ account2‬‬

‫ﻛﻪ ﺩﺭ ﺯﻳﺮ ﺑﻬﻴﻨﻪ ﻣﻲ ﺷﻮﺩ ‪:‬‬

‫)‪σ branch _name = “Hillside” (account1) ∪ σ branch _name = “Hillside” (account2‬‬

‫‪ ‬ﺍﺯ ﺁﻧﺠﺎ ﻛﻪ ‪ account1‬ﻓﻘﻂ ﺗﺎﭘﻞ ﻫﺎﻱ ﻧﺴﺒﺖ ﺩﺍﺩﻧﻲ ﺑﻪ ﺷﻌﺒﻪ ‪ Hillside‬ﺭﺍ ﺩﺍﺭﺩ ‪،‬ﻣﻲ ﺗﻮﺍﻧﻴﻢ‬
‫ﻋﻤﻠﻴﺎﺕ ﺍﻧﺘﺨﺎﺏ ﺭﺍ ﺣﺬﻑ ﻛﻨﻴﻢ ‪.‬‬

‫‪ ‬ﺗﻌﺮﻳﻒ ‪account2‬ﺭﺍ ﺟﻬﺖ ﻛﺴﺐ ﺭﺍﺑﻄﻪ ﺯﻳﺮ ﺑﻜﺎﺭ ﻣﻲ ﺑﺮﻳﻢ ‪.‬‬

‫”‪view‬‬ ‫‪(account ) _name = “Valley _name = “Hillside” (σ branch σ branch‬‬

‫‪ ‬ﺍﻳﻦ ﻋﺒﺎﺭﺕ ﻣﺠﻤﻮﻋﻪ ﺧﺎﻟﻲ ﺩﺭﺧﺼﻮﺹ ﻣﺤﺘﻮﻳﺎﺕ ﺭﺍﺑﻄﻪ ﺣﺴﺎﺏ ﺍﺳﺖ ‪.‬‬

‫‪ ‬ﺍﺳﺘﺮﺍﺗﮋﻱ ﻧﻬﺎﻳﻲ ﺑﺮﺍﻱ ﺳﺎﻳﺖ ‪ Hillside‬ﺟﻬﺖ ﺑﺎﺯﮔﺮﺩﺍﻧﺪﻥ ‪ account1‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻧﺘﻴﺠﻪ ﭘﺮﺱ‬
‫ﻭ ﺟﻮ ﺍﺳﺖ‪.‬‬

‫ﭘﺮﺩﺍﺯﺵ ﺍﺗﺼﺎﻝ ﺳﺎﺩﻩ‬

‫ﺑﻴﺎﻥ ﺟﺒﺮﻱ ﺍﺭﺗﺒﺎﻃﻲ ﺭﺍ ﺩﺭﻧﻈﺮ ﺑﮕﻴﺮﻳﺪ ﻛﻪ ﺩﺭ ﺁﻥ ﺳﻪ ﺍﺭﺗﺒﺎﻁ ﺗﻜﺮﺍﺭ ﻳﺎ ﺑﺨﺶ ﺑﻨﺪﻱ ﻧﻤﻲ ﺷﻮﺩ ‪:‬‬

‫ﺣﺴﺎﺏ ‪ -‬ﺳﭙﺮﺩﻩ ﮔﺬﺍﺭ ‪ -‬ﺷﻌﺒﻪ‬

‫• ﺣﺴﺎﺏ ﺩﺭ ﺳﺎﻳﺖ ‪S1‬ﺫﺧﻴﺮﻩ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫• ﺳﭙﺮﺩﻩ ﮔﺬﺍﺭ ﺩﺭ ‪S2‬‬

‫• ﺷﻌﺒﻪ ﺩﺭ ‪S 3‬‬

‫• ﺑﺮﺍﻱ ﭘﺮﺱ ﻭ ﺟﻮ ﻣﻨﺘﺸﺮ ﺷﺪﻩ ﺩﺭ ﺳﺎﻳﺖ ‪، S‬ﺳﻴﺴﺘﻢ ﺑﻪ ﺍﻳﺠﺎﺩ ﻧﺘﻴﺠﻪ ﺩﺭ ﺳﺎﻳﺖ ‪ SI‬ﻧﻴﺎﺯ ﺩﺍﺭﺩ‪.‬‬

‫‪132‬‬
 ‫ﭘﺎﻳﮕﺎﻩ ﻫﺎﻱ ﺩﺍﺩﻩ ﺗﻮﺯﻳﻊ ﺷﺪﻩ ﻧﺎﻫﻤﮕﻦ‬

‫ﭼﻨﺪﻳﻦ ﺑﺮﻧﺎﻣﻪ ﻛﺎﺭﺑﺮﺩﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺑﻪ ﺩﺍﺩﻩ ﺍﺯ ﻣﺠﻤﻮﻋﻪ ﭘﺎﻳﮕﺎﻩ ﻫﺎﻱ ﺩﺍﺩﻩ ﭘﻴﺶ ﻣﻮﺟﻮﺩ ﻭﺍﻗﻊ ﺷﺪﻩ ﺩﺭ‬
‫ﻣﺠﻤﻮﻋﻪ ﻧﺎﻫﻤﮕﻦ ﺳﻜﻮﻫﺎﻱ ﺳﺨﺖ ﺍﻓﺰﺍﺭ ﻭ ﻧﺮﻡ ﺍﻓﺰﺍﺭ ﻧﻴﺎﺯ ﺩﺍﺭﺩ ‪ .‬ﻣﺪﻝ ﻫﺎﻱ ﺩﺍﺩﻩ ﻣﻤﻜﻦ ﺍﺳﺖ )ﺳﻠﺴﻠﻪ‬
‫ﻣﺮﺍﺗﺒﻲ ‪،‬ﺍﺭﺗﺒﺎﻃﻲ ﻭ ﻏﻴﺮﻩ ( ﻣﺘﻔﺎﻭﺕ ﺑﺎﺷﻨﺪ‪ .‬ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﻭﺍﮔﺬﺍﺭﻱ ﺗﺮﺍﻛﻨﺶ ﻣﻤﻜﻦ ﺍﺳﺖ ﻧﺎﺳﺎﺯﮔﺎﺭ‬
‫ﺑﺎﺷﺪ‪ .‬ﻛﻨﺘﺮﻝ ﻩ ﻣﮕﺎﻫﻲ ﻣﻤﻜﻦ ﺍﺳﺖ ﻣﺒﺘﻨﻲ ﺑﺮ ﺗﻜﻨﻴﻚ ﻫﺎﻱ ﻣﺘﻔﺎﻭﺕ ) ﻗﻔﻞ ﻛﺮﺩﻥ ‪،‬ﻧﺸﺎﻥ ﺯﻣﺎﻧﻲ ﻭ‬
‫ﻏﻴﺮﻩ ( ﺑﺎﺷﺪ‪ .‬ﺟﺰﺋﻴﺎﺕ ﺳﻄﺢ ﺳﻴﺴﺘﻢ ﺗﻘﺮﻳﺒﺎ ﺑﻪ ﻃﻮﺭ ﻣﺸﺨﺺ ﻛﺎﻣﻼ ﻧﺎﺳﺎﺯﮔﺎﺭ ﺍﺳﺖ ‪.‬‬

‫ﺳﻴﺴﺘﻢ ﭼﻨﺪ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺍﻱ ﻳﻚ ﻻﻳﻪ ﻧﺮﻡ ﺍﻓﺰﺍﺭ ﺩﺭ ﺑﺎﻻﻱ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻛﻨﻮﻧﻲ ﺍﺳﺖ ﻛﻪ‬
‫ﺟﻬﺖ ﺩﺳﺘﻜﺎﺭﻱ ﺍﻃﻼﻋﺎﺕ ﺩﺭ ﭘﺎﻳﮕﺎ ﻩ ﻫﺎﻱ ﺩﺍﺩﻩ ﻧﺎﻫﻤﮕﻦ ﻃﺮﺍﺣﻲ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫ﻣﺰﺍﻳﺎ‬

‫‪ ‬ﺣﻔﻆ ﺳﺮﻣﺎﻳﻪ ﮔﺬﺍﺭﻱ ﺩﺭ ﺍﻳﻦ ﻣﻮﺍﺭﺩ‪ :‬ﺳﺨﺖ ﺍﻓﺰﺍﺭﻱ‪ ،‬ﻧﺮﻡ ﺍﻓﺰﺍﺭ ﺳﻴﺴﺘﻢ‪ ،‬ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﻛﺎﺭﺑﺮﺩﻱ‬

‫‪ ‬ﻛﻨﺘﺮﻝ ﺍﺟﺮﺍﻳﻲ ﻭ ﺍﺳﺘﻘﻼﻝ ﻣﺤﻠﻲ‬

‫‪ ‬ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ‪ DBMSs‬ﺑﺎ ﻫﺪﻑ ﺧﺎﺹ ﺭﺍ ﻣﺠﺎﺯ ﻣﻲ ﺳﺎﺯﺩ ‪.‬‬

‫‪ ‬ﺑﺴﻮﻱ ‪ DBMSs‬ﻫﻤﮕﻦ ﻣﺘﺤﺪ ﺷﺪﻩ ﮔﺎﻡ ﺑﺮﻣﻲ ﺩﺍﺭﺩ ‪.‬‬

‫– ﻳﻜﭙﺎﺭﭼﮕﻲ ﻛﺎﻣﻞ ﺑﻪ ‪ DBMSs‬ﻫﻤﮕﻦ ﺑﺎ ﻣﻮﺍﺭﺩ ﺯﻳﺮ ﻣﻮﺍﺟﻪ ﻣﻲ ﺷﻮﺩ ‪:‬‬

‫• ﺩﺷﻮﺍﺭﻱ ﻫﺎﻱ ﺗﻜﻨﻴﻜﻲ ﻭ ﻫﺰﻳﻨﻪ ﺣﻔﻆ‬

‫• ﺩﺷﻮﺍﺭﻳﻬﺎﻱ ﺳﺎﺯﻣﺎﻧﺪﻫﻲ ‪ /‬ﺳﻴﺎﺳﻲ‬

‫– ﺳﺎﺯﻣﺎﻥ ﻫﺎ ﺧﻮﺍﻫﺎﻥ ﺍﺭﺍﺋﻪ ﻛﻨﺘﺮﻝ ﺑﺮ ﺩﺍﺩﻩ ﻫﺎﻳﺸﺎﻥ ﻧﻴﺴﺘﻨﺪ ‪.‬‬

‫– ﭘﺎﻳﮕﺎﻩ ﻫﺎﻱ ﺩﺍﺩﻩ ﻣﺤﻠﻲ ﺗﻤﺎﻳﻞ ﺑﻪ ﺣﻔﻆ ﺍﺳﺘﻘﻼﻝ ﺑﻴﺸﺘﺮ ﺩﺍﺭﺩ ‪.‬‬

‫‪133‬‬
 ‫ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﺗﻌﺪﻳﻞ ﻛﻨﻨﺪﻩ‬
‫ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﺗﻌﺪﻳﻞ ﻛﻨﻨﺪﻩ‪ ،‬ﺳﻴﺴﺘﻢ ﻫﺎﻳﻲ ﻫﺴﺘﻨﺪ ﻛﻪ ﻣﻨﺎﺑﻊ ﺩﺍﺩﻩ ﻧﺎﻫﻤﮕﻦ ﻣﺘﻌﺪﺩ ﺭﺍ ﺑﺎ ﺍﺭﺍﺋﻪ ﻛﺮﺩﻥ‬
‫ﺩﻳﺪﮔﺎﻩ ﺟﻬﺎﻧﻲ ﻳﻜﭙﺎﺭﭼﻪ ﺷﺪﻩ ﻳﻜﭙﺎﺭﭼﻪ ﻣﻲ ﻛﻨﻨﺪ ﻭ ﺗﺴﻬﻴﻼﺕ ﭘﺮﺱ ﻭ ﺟﻮ ﺩﺭ ﺩﻳﺪﮔﺎﻩ ﺟﻬﺎﻧﻲ ﺭﺍ‬
‫ﺍﺭﺍﺋﻪ ﻣﻲ ﺩﻫﻨﺪ ‪ .‬ﺑﺮﺧﻼﻑ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻛﺎﻣﻼ ‪ ، fledged‬ﺑﻪ ﻃﻮﺭﻛﻠﻲ ﺗﻌﺪﻳﻞ ﻛﻨﻨﺪﻩ‬
‫ﻫﺎ ﺩﺭﺑﺎﺭﻩ ﭘﺮﺩﺍﺯﺵ ﺗﺮﺍﻛﻨﺶ ﺁﺳﻴﺐ ﻧﻤﻲ ﺑﻴﻨﻨﺪ ‪ .‬ﺍﻣﺎ ﺍﺻﻼﺣﺎﺕ ﺗﻌﺪﻳﻞ ﻛﻨﻨﺪﻩ ﻭ ﭼﻨﺪ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺍﻱ‬
‫ﮔﺎﻫﻲ ﺍﻭﻗﺎﺕ ﺑﻪ ﻃﻮﺭ ﻗﺎﺑﻞ ﺗﻌﻮﻳﻀﻲ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺷﻮﺩ ‪ .‬ﺍﺻﻄﻼﺡ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻣﺠﺎﺯﻱ ﺟﻬﺖ ﺍﺷﺎﺭﻩ ﺑﻪ‬
‫ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﭼﻨﺪ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺗﻌﺪﻳﻞ ﻛﻨﻨﺪﻩ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﻓﻬﺮﺳﺖ‬

‫ﺍﻧﻮﺍﻉ ﺧﺎﺹ ﺍﻃﻼﻋﺎﺕ ﻓﻬﺮﺳﺖ‬

‫‪ l‬ﺍﻃﻼﻋﺎﺕ ﻛﺎﺭﻣﻨﺪ ﺍﺯﻗﺒﻴﻞ ﻧﺎﻡ ‪،‬ﺷﻤﺎﺭﻩ ﺷﻨﺎﺳﺎﻳﻲ ‪ ،‬ﺍﻳﻤﻴﻞ ‪،‬ﺗﻠﻔﻦ ‪ ،‬ﺁﺩﺭﺱ ﺍﺩﺍﺭﻩ‬

‫‪ l‬ﺣﺘﻲ ﺍﻃﻼﻋﺎﺕ ﺷﺨﺼﻲ ﺍﺯ ﻣﻜﺎﻥ ﻫﺎ ﻱ ﻣﺖﻉﺩﺩ ﻗﺎﺑﻞ ﺩﺳﺘﺮﺳﻲ ﺍﺳﺖ ‪.‬‬

‫‪ ‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ ﻧﺸﺎﻥ ﻫﺎﻱ ﻛﺘﺎﺏ ﺟﺴﺘﺠﻮﮔﺮ‬

‫ﺻﻔﺤﺎﺕ ﺳﻔﻴﺪ‬

‫‪ l‬ﺛﺒﺖ ﻫﺎﻱ ﺳﺎﺯﻣﺎﻧﺪﻫﻲ ﺷﺪﻩ ﺑﺎ ﻧﺎﻡ ﻳﺎ ﺷﻨﺎﺳﺎﻳﻲ ﻛﻨﻨﺪﻩ‬

‫‪ ‬ﺑﻪ ﻣﻌﻨﻲ ﺟﺴﺘﺠﻮﻱ ﺑﻴﺸﺘﺮ ﺟﻬﺖ ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﺍﻃﻼﻋﺎﺕ ﺑﻴﺸﺘﺮ ﺩﺭﺑﺎﺭﻩ ﻳﻚ ﺛﺒﺖ‬

‫ﺻﻔﺤﺎﺕ ﺯﺭﺩ‬

‫‪ l‬ﺛﺒﺖ ﻫﺎﻱ ﺳﺎﺯﻣﺎﻧﺪﻫﻲ ﺷﺪﻩ ﺑﺎ ﻭﻳﮋﮔﻲ ﻫﺎ‬

‫‪ l‬ﺑﺮﺍﻱ ﺟﺴﺘﺠﻮﻱ ﻣﻌﻜﻮﺱ ﺟﻬﺖ ﭘﻴﺪﺍﻛﺮﺩﻥ ﺛﺒﺖ ﻫﺎﻱ ﺗﻄﺒﻴﻖ ﺩﻫﻨﺪﻩ ﻣﻘﺘﻀﻴﺎﺕ ﺧﺎﺹ‬

‫‪134‬‬
 ‫ﺭﺍﻳﺞ ﺗﺮﻳﻦ پﺭﻭﺗﻜﻞ ﺩﺳﺘﺮﺳﻲ ﻓﻬﺮﺳﺖ ﻣﻮﺭﺩ ﺍﺳﺘﻔﺎﺩﻩ ‪)LDAP :‬ﭘﺮﻭﺗﻜﻞ ﺩﺳﺘﺮﺳﻲ ﻓﻬﺮﺳﺖ‬
‫ﺳﺒﻚ ﻭﺯﻥ ( ﻛﻪ ﺳﺎﺩﻩ ﺷﺪﻩ ﺍﺯ ﭘﺮﻭﺗﻜﻞ ‪ X.500‬ﺍﻭﻟﻴﻪ‬

‫ﺳﻮﺍﻝ‪ :‬ﭼﺮﺍ ﺍﺯ ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻣﺎﻧﻨﺪ ‪ ODBC/JDBC‬ﺍﺳﺘﻔﺎﺩﻩ ﻧﻤﻲ ﻛﻨﻴﻢ ؟‬

‫ﻃﻮﺭ ﻣﻮﺍﺯﻱ ﺗﻜﺎﻣﻞ ﻳﺎﻓﺘﻪ ﺗﺎ‬

‫ﺏ‬ ‫ﭘﺎﺳﺦ‪ :‬ﭘﺮﻭﺗﻜﻞ ﻫﺎﻱ ﺳﺎﺩﻩ ﺷﺪﻩ ﺑﺮﺍﻱ ﻧﻮﻉ ﻣﺤﺪﻭﺩ ﺩﺳﺘﺮﺳﻲ ﺩﺍﺩﻩ‪،‬‬
‫‪ .ODBC/JDBC‬ﻣﻜﺎﻧﻴﺴﻢ ﺧﻮﺏ ﻧﺎﻣﮕﺬﺍﺭﻱ ﺳﻠﺴﻠﻪ ﻣﺮﺍﺗﺒﻲ ﻣﺸﺎﺑﻪ ﻓﻬﺮﺳﺖ ﻩ ﺍ ﻱ ﺳﻴﺴﺘﻢ ﻓﺎﻳﻞ ﺭﺍ ﺍﺭﺍﺋﻪ‬
‫ﻣﻲ ﺩﻫﺪ‪ .‬ﺩﺍﺩﻩ ﻣﻲ ﺗﻮﺍﻧﺪ ﺩﺭﺑﻴﻦ ﺳﺮﻭﺭﻫﺎﻱ ﻣﺘﻌﺪﺩ ﺑﺮﺍﻱ ﺑﺨﺶ ﻫﺎﻱ ﻣﺨﺘﻠﻒ ﺳﻠﺴﻠﻪ ﻣﺮﺍﺗﺐ ﺍﻓﺮﺍﺯ ﺷﻮﺩ ‪،‬‬
‫ﺩﻳﺪﮔﺎﻩ ﺧﺎﺹ ﺑﻪ ﻛﺎﺭﺑﺮ ﺭﺍ ﺍﺭﺍﺋﻪ ﻣﻲ ﺩﻫﺪ ‪.‬‬

‫ﻣﺎﻧﻨﺪ ﺳﺮﻭﺭﻫﺎﻱ ﻣﺨﺘﻠﻒ ﺑﺮﺍﻱ ‪Bell Labs Murray Hill and Bell Labs Bangalore‬‬

‫ﻓﻬﺮﺳﺖ ﻫﺎ ﻣﻤﻜﻦ ﺍﺳﺖ ﺍﺯ ﭘﺎﻳﮕﺎﻩ ﻫﺎﻱ ﺩﺍﺩﻩ ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﻜﺎﻧﻴﺴﻢ ﻗﻮﻱ ﺍﺳﺘﻔﺎﺩﻩ ﻛﻨﻨﺪ ‪.‬‬

‫ﺩﺭﺧﺘﺎﻥ ﻓﻬﺮﺳﺖ ﺗﻮﺯﻳﻊ ﺷﺪﻩ‬

‫ﺍﻃﻼﻋﺎﺕ ﺳﺎﺯﻣﺎﻧﻲ ﻣﻤﻜﻦ ﺍﺳﺖ ﺑﻪ ﺩﺭﺧﺘﺎﻥ ﻣﺘﻌﺪﺩ ﺍﻃﻼﻋﺎﺕ ﻓﻬﺮﺳﺖ ﺟﺪﺍ ﺷﻮﺩ ‪ .‬ﭘﻴﺸﻮﻧﺪ ‪RDN، DIT‬‬
‫ﺭﺍ ﺍﺭﺍﺋﻪ ﻣﻲ ﺩﻫﺪ ﺗﺎ ﺩﺭ ﺗﻤﺎﻣﻲ ﺛﺒﺖ ﻫﺎ ﺟﻬﺖ ﻛﺴﺐ ‪DN‬ﻛﻠﻲ ﺑﺮﭼﺴﺐ ﮔﺬﺍﺭﻱ ﺷﻮﺩ ‪ .‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ‬
‫ﺩﻭ‪ DITs‬ﻳﻜﻲ ﺑﺎ ﭘﻴﺸﻮﻧﺪ ‪ o=Lucent, c=USA‬ﻭ ﺩﻳﮕﺮﻱ ﺑﺎ ﭘﻴﺸﻮﻧﺪ‬
‫‪o=Lucent, c=India‬‬
‫ﺍﻏﻠﺐ ﺳﺎﺯﻣﺎﻥ ﻫﺎ ‪ DITs‬ﺭﺍ ﺑﺮﻣﺒﻨﺎﻱ ﻣﻮﻗﻌﻴﺖ ﺟﻐﺮﺍﻓﻴﺎﻳﻲ ﻳﺎ ﺑﺎ ﺳﺎﺧﺘﺎﺭ ﺳﺎﺯﻣﺎﻧﻲ ﺟﺪﺍ ﻣﻲ ﻛﻨﻨﺪ ‪ .‬ﭼﻨﺪﻳﻦ‬
‫ﺍﺟﺮﺍﻱ ‪LDAP‬ﺗﻜﺮﺍﺭ ‪) DITs‬ﺗﻜﺮﺍﺭ ﺍﺻﻠﻲ‪ -‬ﺩﻧﺒﺎﻟﻪ ﺭﻭ ﻳﺎ ﭼﻨﺪﺍﺻﻠﻲ ( ﺭﺍ ﺣﻤﺎﻳﺖ ﻣﻲ ﻛﻨﺪ ‪ .‬ﮔﺮﻫﻲ ﺩﺭﻳﻚ‬
‫‪DIT‬ﻣﻤﻜﻦ ﺍﺳﺖ ﺍﺭﺟﺎﻋﻲ ﺑﻪ ﻳﻚ ﮔﺮﻩ ﺩﺭ ‪ DIT‬ﺩﻳﮕﺮ ﺑﺎﺷﺪ ‪ .‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ ‪Bell Labs = OU‬‬
‫ﻣﻤﻜﻦ ﺍﺳﺖ ‪DIT‬ﻣﺠﺰﺍ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ ﻭ ‪ DIT‬ﺑﺮﺍﻱ ‪ o=Lucen‬ﻣﻤﻜﻦ ﺍﺳﺖ ﻭﺭﻗﻲ ﺑﺎ ‪ou=Bell Labs‬‬
‫ﻣﺤﺘﻮﻱ ﺍﺭﺟﺎﻋﻲ ﺑﺎ ‪Bell Labs DIT‬ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ‪.‬‬

‫ﺍﺭﺟﺎﻋﻲ ﻫﺎ ﻋﺎﻣﻞ ﺍﺻﻠﻲ ﺩﺭ ﻳﻜﭙﺎﺭﭼﻪ ﻛﺮﺩﻥ ﻣﺠﻤﻮﻋﻪ ﺗﻮﺯﻳﻊ ﺷﺪﻩ ﻓﻬﺮﺳﺖ ﻫﺎ ﻫﺴﺘﻨﺪ ‪ .‬ﺯﻣﺎﻧﻴﻜﻪ ﻳﻚ‬
‫ﺳﺮﻭﺭ ﺑﻪ ﭘﺮﺱ ﻭ ﺟﻮ ﻗﺎﺑﻞ ﺩﺳﺘﺮﺳﻲ ﺑﻪ ﻳﻚ ﮔﺮﻩ ﺍﺭﺟﺎﻋﻲ ﺩﺳﺖ ﻣﻲ ﻳﺎﺑﺪ ‪ ،‬ﻣﻤﻜﻦ ﺍﺳﺖ ﭘﺮﺱ ﻭ ﺟﻮ‬
‫ﺟﻠﻮﻳﻲ ﺑﻪ ‪ DIT‬ﺍﺭﺟﺎﻉ ﺩﺍﺩﻩ ﺷﺪﻩ ﻭ ﭘﺎﺳﺦ ﺑﺮﮔﺸﺘﻲ ﺑﻪ ﻣﺮﺍﺟﻌﻪ ﻛﻨﻨﺪﻩ ﺑﺎﺷﺪ ‪،‬ﻳﺎ ﺑﺮﮔﺸﺖ ﺍﺭﺟﺎﻋﻲ ﺑﻪ‬

‫‪135‬‬
‫‪ DIT‬ﺍﺭﺟﺎﻉ ﺩﺍﺩﻩ ﺷﺪﻩ )ﺑﺪﻭﻥ‬ ‫ﻣﺮﺍﺟﻌﻪ ﻛﻨﻨﺪﻩ ﺍﻱ ﺭﺍ ﺍﺭﺍﺋﻪ ﺩﻫﻴﺪ ﻛﻪ ﺑﻄﻮﺭ ﺷﻔﺎﻑ ﭘﺮﺱ ﻭ ﺟﻮﻳﻲ ﺑﻪ‬
‫ﻣﺪﺍﺧﻠﻪ ﻛﺎﺭﺑﺮ( ﻣﻲ ﻓﺮﺳﺘﺪ‪.‬‬

‫ﺳﻪ ﻓﺎﺯ ﭘﺬﻳﺮﺵ )‪(commit‬‬

‫‪ ‬ﻓﺮﺽ‪:‬‬

‫ﻫﻴﭻ ﺷﺐﻛﻪ ﺍﻱ ﺑﺨﺶ ﺑﻨﺪﻱ ﻧﺸﻮﺩ ‪.‬‬

‫ﺩﺭ ﻫﺮ ﻣﻜﺎﻧﻲ ﻛﻮﭼﻜﺘﺮﻳﻦ ﺳﺎﻳﺖ ﺑﺎﻳﺪ ﺩﺭ ﺣﺎﻝ ﻛﺎﺭ ﺑﺎﺷﺪ ‪.‬‬

‫ﺩﺭ ‪ k‬ﺳﺎﻳﺖ ﺧﻄﺎ ﺭﺥ ﺩﻫﺪ‪.‬‬

‫‪ ‬ﻓﺎﺯ ‪ : 1‬ﻛﻪ ﺷﺎﻣﻞ ﻳﻚ ﺗﺼﻤﻴﻢ ﻣﻘﺪﻣﺎﺗﻲ ﺍﺳﺖ ﻳﻌﻨﻲ ﺷﻨﺎﺳﺎﻳﻲ ‪ 2PC‬ﺩﺭ ﻓﺎﺯ ﻳﻚ‪.‬‬

‫ﻫﺮ ﺳﺎﻳﺘﻲ ﻣﻴﺘﻮﺍﻧﺪ ﻋﻤﻞ ﺧﻮﺍﻧﺪ ﺭﻭﻱ ﺩﺳﺘﻮﺭﺍﺕ ﭘﺬﻳﺮﺵ ﺷﺪﻩ ﺭﺍ ﺍﻧﺠﺎﻡ ﺩﻫﺪ ‪.‬‬

‫‪ ‬ﻓﺎﺯ ‪ : 2‬ﺍﻳﻦ ﻓﺎﺯ ﺑﻪ ﺩﻭ ﻓﺎﺯ ﺷﻜﺴﺘﻪ ﻣﻴﺸﻮﺩ‪ ،‬ﻓﺎﺯ‪ 2‬ﻭ ﻓﺎﺯ‪ 3‬ﺑﺮﺍﻱ ‪3PC‬‬

‫‪ 2PC‬ﻣﻲ ﮔﻴﺮﺩ ) ‪pre-commit‬‬ ‫ﺩﺭ ﻓﺎﺯ ﺩﻭ ﻫﻤﺎﻫﻨﮓ ﻛﻨﻨﺪﻩ ﺗﺼﻤﻴﻤﻲ ﺑﺮﺍﻱ‬

‫‪ (decision‬ﻭ ﺁﻧﺮﺍ ﺩﺭ ﭼﻨﺪﻳﻦ ﺳﺎﻳﺖ ﺫﺧﻴﺮﻩ ﻣﻴﻜﻨﺪ‪.‬‬

‫‪ commit/abort‬ﺑﻪ ﻫﻤﻪ ﻱ‬ ‫ﺩﺭ ﻓﺎﺯ ﺳﻪ ﻫﻤﺎﻫﻨﮓ ﻛﻨﻨﺪﻩ ﻳﻚ ﭘﻴﺎﻡ ﺑﺎ ﻣﺤﺘﻮﺍﻱ‬

‫ﺳﺎﻳﺖ ﻫﺎﻱ ﻣﻮﺟﻮﺩ‪.‬‬

‫‪ ‬ﺗﺤﺖ ﻧﻈﺮ ‪ ،3PC‬ﺗﺼﻤﻴﻢ ﻫﻤﺎﻫﻨﮓ ﻛﻨﻨﺪﻩ ﻣﻴﺘﻮﺍﻧﺪ ﺣﺘﻲ ﺩﺭ ﺻﻮﺭﺕ ﻭﺟﻮﺩ ﺧﻄﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺷﻮﺩ ‪.‬‬

‫ﺟﻠﻮﮔﻴﺮﻱ ﺍﺯ ﺑﻪ ﻭﺟﻮﺩ ﺁﻣﺪﻥ ﻣﺸﻜﻞ ﺑﻦ ﺑﺴﺖ‬

‫‪ ‬ﻣﻌﺎﻳﺐ‪:‬‬

‫ﺑﺮﻭﺯ‪ overheads‬ﺑﺴﻴﺎﺭ ﺑﺎﻻ‬

‫ﻓﺮﺿﻴﺎﺕ ﻣﻤﻜﻦ ﺑﺮﺍﻱ ﺑﻌﻀﻲ ﻣﻮﺍﺭﺩ ﻗﺎﺑﻞ ﺗﻮﺿﻴﺢ ﻧﺒﺎﺷﺪ ‪.‬‬

‫‪136‬‬
 ‫ﻓﺼﻞ ‪:7‬‬
‫ﭘﺮﺩﺍﺯﺵ ﭘﺮﺱ ﻭ ﺟﻮ‬

‫‪137‬‬
 ‫ﻟﻴﺴﺖ ﻣﻄﺎﻟﺐ‪:‬‬

‫ﭘﻴﺶ ﻧﻤﺎﻳﺶ‬

‫ﺍﻧﺪﺍﺯﻩ ﮔﻴﺮﻱ ﻫﺰﻳﻨﻪ ﻫﺎﻱ ﭘﺮﺱ ﻭ ﺟﻮ‬

‫ﻋﻤﻠﻴﺎﺕ ﺍﻧﺘﺨﺎﺏ‬

‫ﻣﺮﺗﺐ ﺳﺎﺯﻱ‬

‫ﺗﺎﺭﻳﺦ ﻋﻤﻠﻴﺎﺕ‬

‫ﺩﻳﮕﺮ ﻋﻤﻠﻴﺎﺕ‬

‫ﺍﺭﺯﻳﺎﺑﻲ ﻋﺒﺎﺭﺍﺕ‬

‫ﻣﺮﺍﺣﻞ ﺍﺳﺎﺳﻲ ﺩﺭ ﭘﺮﺩﺍﺯﺵ ﭘﺮﺱ ﻭ ﺟﻮ ﻫﺎ‪:‬‬

‫‪ (1‬ﺗﺠﺰﻳﻪ ﻭ ﺗﺤﻠﻴﻞ‬

‫‪ (2‬ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ‬

‫‪ (3‬ﺍﺭﺯﻳﺎﺑﻲ‬

‫‪138‬‬
 ‫ﺗﺠﺰﻳﻪ ﻭ ﺍﻧﺘﻘﺎﻝ‬
‫• ‪Query‬ﺩﺭ ﻋﻤﻠﻴﺎﺕ ﺩﺍﺧﻠﻲ ﺑﻪ ﺍﻳﻦ ﻣﻌﻨﻲ ﺳﺖ‪.‬‬

‫• ﺗﺠﺰﻳﻪ ﻛﻨﻨﺪﻩ ‪،‬ﭼﻚ ﻛﺮﺩﻥ )‪syntax‬ﭼﻚ ﻛﺮﺩﻥ ﻣﺤﺘﻮﺍﻱ ﻣﺘﻨﻲ( ‪،‬ﺗﺎﺋﻴﺪ ‪،‬ﺭﻭﺍﺑﻂ‪.‬‬

‫• ﺩﺭﺣﻘﻴﻘﺖ ﺩﺍﺧﻞ ﻳﻚ ﻋﻤﻠﻴﺎﺕ ﭘﺮﺱ ﻭ ﺟﻮ ﺗﻤﺎﻡ ﺍﻳﻦ ﻛﺎﺭﻫﺎ ﺍﻧﺠﺎﻡ ﻣﻲ ﺷﻮﺩ ﻭ ﻧﺘﻴﺠﻪ‬
‫ﻋﻤﻠﻴﺎﺕ ‪ Query‬ﻣﻮﺭﺩ ﻧﻈﺮ ﻛﺎﺭﺑﺮ ﺍﺳﺖ‪.‬‬

‫ﺍﺭﺯﻳﺎﺑﻲ‬
‫• ﺩﺭ ﺍﺟﺮﺍﻱ ﻃﺮﺡ ﭘﺮﺱ ﻭ ﺟﻮ ﻋﻤﻠﻴﺎﺕ ﺍﺭﺯﻳﺎﺑﻲ ﺑﻪ ﻃﻮﻝ ﻣﻲ ﺍﻧﺠﺎﻣﺪ ﭼﺮﺍﻛﻪ ﻃﺮﺡ ﺑﺎﻳﺪ‬
‫ﺍﺟﺮﺍ ﻭ ﭘﺎﺳﺦ ﻫﺎ ﺑﻪ ﭘﺮﺱ ﻭ ﺟﻮ ﺑﺮﮔﺮﺩﺍﻧﺪﻩ ﺷﻮﺩ‪.‬‬

‫ﻣﺮﺍﺣﻞ ﺍﺳﺎﺳﻲ ﺩﺭ ﭘﺮﺩﺍﺯﺵ ﭘﺮﺱ ﻭ ﺟﻮ ﻫﺎ‪ :‬ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ‬

‫♦ ﻳﻚ ﻋﺒﺎﺭﺕ ﻳﺎ ﺭﺍﺑﻄﻪ ﺟﺒﺮﻱ ﻣﻤﻜﻦ ﺍﺳﺖ ﺩﺍﺭﺍﻱ ﭼﻨﺪ ﻋﺒﺎﺭﺕ ﺍﺭﺯﻳﺎﺑﻲ ﺑﺎﺷﺪ‪.‬‬

‫ﻣﺜﺎﻝ‪:‬‬

‫‪E.g., σsalary<75000(∏salary(instructor)) is equivalent to‬‬

‫))‪∏salary(σsalary<75000(instructor‬‬
‫♦ ﻫﺮﻳﻚ ﺍﺯ ﺭﺍﺑﻄﻪ ﻫﺎﻱ ﺟﺒﺮﻱ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﺑﺎ ﺍﻟﮕﻮﺭﻳﺘﻢ ﻫﺎﻱ ﻣﺘﻔﺎﻭﺗﻲ ﻋﻤﻠﻴﺎﺕ ﺑﺮﺭﺳﻲ ﻭ ﺍﺭﺯﻳﺎﺑﻲ ﺭﺍ‬
‫ﺍﻧﺠﺎﻡ ﺩﻫﻨﺪ‪.‬‬

‫ﻣﺘﻘﺎﺑﻼ ﻳﻚ ﻋﺒﺎﺭﺕ ﺟﺒﺮﻱ ﻧﻴﺰ ﺍﺯ ﺟﻬﺎﺕ ﻣﺨﺘﻠﻒ ﻣﻲ ﺗﻮﺍﻧﺪ ﻣﻮﺭﺩ ﺍﺭﺯﻳﺎﺑﻲ ﻗﺮﺍﺭ ﺑﮕﻴﺮﺩ ‪.‬‬

‫ﺑﻴﺎﻥ ﻣﺸﺮﻭﺡ ﻭ ﺗﻌﻴﻴﻦ ﺍﺑﺰﺍﺭ ﺩﻗﻴﻖ ﺍﺭﺯﻳﺎﺑﻲ‪:‬‬

‫♦ ﺍﺭﺯﻳﺎﺑﻲ ﻃﺮﺡ‬

‫ﺑﻌﻨﻮﺍﻥ ﻣﺜﺎﻝ ﻣﻲ ﺗﻮﺍﻥ ﺑﺮﺍﻱ ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﻣﺪﺭﺳﺎﻥ ﺑﺎ ﺣﻘﻮﻕ ﻛﻤﺘﺮ ﺍﺯ ‪ 75000‬ﺍﺯﺷﺎﺧﺺ‬
‫)ﺷﺮﻁ( ‪ salary<75000‬ﺍﺳﺘﻔﺎﺩﻩ ﻛﺮﺩ‪.‬‬

‫‪139‬‬
‫ﻳﺎ ﻣﻲ ﺗﻮﺍﻥ ﺑﺎ ﺟﺴﺘﺠﻮ ﻭ ﺍﺳﻜﻦ ﻣﺪﺭﺳﺎﻥ ﺑﺎ ﺣﻘﻮﻕ ﺑﺎﻻﻱ ‪ 75000‬ﺭﺍ ﺣﺬﻑ ﻛﺮﺩ ﺗﺎ‬
‫ﻣﺪﺭﺳﺎﻥ ﺑﺎ ﺣﻘﻮﻕ ﻛﻤﺘﺮﺍﺯ‪ 75000‬ﻣﺸﺨﺺ ﺷﻮﻧﺪ‪.‬‬

‫♦ ‪ :Query Optimization‬ﺍﻧﺘﺨﺎﺏ ﺑﺮﻧﺎﻣﻪ ﻳﻞ ﻃﺮﺣﻲ ﺑﺎ ﻛﻤﺘﺮﻳﻦ ﻫﺰﻳﻨﻪ ﺍﺯ ﺑﻴﻦ ﺗﻤﺎﻡ ﺑﺮﻧﺎﻣﻪ‬

‫ﻫﺎﻱ ﺍﺭﺯﻳﺎﺑﻲ‬

‫ﺑﺮ ﺁﻭﺭﺩ ﻫﺰﻳﻨﻪ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺍﻃﻼﻋﺎﺕ ﺍﻣﺎﺭﻱ ﺑﺮﮔﺮﻓﺘﻪ ﺍﺯ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ‬

‫‪ n‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ ﺩﺭ ﻫﺮ ﺭﺍﺑﻄﻪ ﺗﻌﺪﺍﺩﻭﻫﺰﻳﻨﻪ ﻭ ﺍﻧﺪﺍﺯﻩ ﻫﺮﺗﺎﭘﻞ ﻭ ‪...‬‬

‫♦ ﺩﺭ ﺍﻳﻦ ﻓﺼﻞ ﻣﻲ ﺧﻮﺍﻧﻴﻢ‬

‫ﭼﮕﻮﻧﮕﻲ ﺍﻧﺪﺍﺯﻩ ﮔﻴﺮﻱ ﻫﺰﻳﻨﻪ ﻫﺎﻱ ﭘﺮﺱ ﻭ ﺟﻮ‬

‫ﺍﻟﮕﻮﺭﻳﺘﻤﻲ ﺑﺮﺍ ﺍﻧﺪﺍﺯﻩ ﮔﻴﺮﻱ ﺟﺒﺮﺭﺍﺑﻄﻪ ﺍﻱ‪.‬‬

‫ﭼﮕﻮﻧﮕﻲ ﺗﺮﻛﻴﺐ ﺍﻟﮕﻮ ﺭﻳﺘﻢ ﻫﺎ ﺑﺮﺍﻱ ﻋﻤﻠﻴﺎﺕ ﻫﺎﻱ ﻓﺮﺩﻱ ﺏ ﻣﻨﻈﻮﺭ ﺍ ﺭﺯﻳﺎﺑﻲ ﻛﺎﻣﻞ ﻳﻚ‬
‫ﻋﺒﺎﺭﺕ‪.‬‬

‫♦ ﻓﺼﻞ ﭼﻬﺎﺭﺩﻫﻢ‬

‫ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﭘﺮﺱ ﻭﺟﻮﻫﺎﻱ ﺍﻧﺠﺎﻡ ﺷﺪﻩ ﻭ ﭼﮕﻮﻧﻪ ﭘﻴﺪﺍ ﻛﻨﻴﻢ ﺭﻭﺵ ﺍﺭﺯﻳﺎﺑﻲ ﻛﻪ ﻛﻤﺘﺮﻳﻦ‬
‫ﻫﺰﻳﻨﻪ ﺭﺍ ﺩﺍﺭﺩ‪.‬‬

‫ﻧﺤﻮﻩ ﺍﻧﺪﺍﺯﻩ ﮔﻴﺮﻱ ﻫﺰﻳﻨﻪ ﺩﺭ ﭘﺮﺱ ﻭ ﺟﻮ‬

‫ﻫﺰﻳﻨﻪ ﺑﻪ ﻃﻮﺭ ﻛﻠﻲ ﻛﻞ ﺯﻣﺎﻥ ﺳﭙﺮﻱ ﺷﺪﻩ ﺑﺮﺍﻱ ﭘﺎﺳﺦ ﺑﻪ ﭘﺮﺱ ﻭ ﺟﺪ ﺍﺳﺖ‬

‫∆ ﻋﻮﺍﻣﻠﻲ ﻛﻪ ﺑﻴﺸﺘﺮﻣﻨﺠﺮ ﺑﻪ ﻫﺰﻳﻨﻪ ﻫﺎﻱ ﺯﻣﺎﻥ ﻣﻲ ﺷﻮﻧﺪ‬

‫‪ ‬ﺩﺳﺘﻴﺎﺑﻲ ﺑﻪ ﺩﻳﺴﻚ‪ ،‬ﻭﺍﺣﺪ ﭘﺮﺩﺍﺯﺵ ﻣﺮﻛﺰﻱ ﻭ ﻫﺮ ﻧﻮﻉ ﺍﺭﺗﺒﺎﻁ ﺷﺒﻜﻪ ﺍﻱ‬

‫ﻣﻌﻤﻮﻻ ﺩﺳﺘﺮﺳﻲ ﺑﻪ ﺩﻳﺴﻚ ﻫﺰﻳﻨﻪ ﻏﺎﻟﺐ ﺍﺳﺖ ﻭ ﻧﺴﺒﺘﺎ ﺁﺳﺎﻥ ﺑﺮﺁﻭﺭﺩﻣﻲ ﺷﻮﺩ ﻭ ﺍﻧﺪﺍﺯﮔﻴﺮﻱ ﺑﺎ‬
‫ﺗﻮﺟﻪ ﺑﻪ‪:‬‬

‫‪140‬‬
 ‫∆ ﺗﻌﺪﺍﺩ ﺟﺴﺘﺠﻮ* ﻣﺘﻮﺳﻂ ﻫﺰﻳﻨﻪ ﺟﺴﺘﺠﻮ‬

‫∆ ﺗﻌﺪﺍﺩ ﺑﻠﻮﻙ * ﻣﺘﻮﺳﻂ ﻫﺰﻳﻨﻪ ﺧﻮﺍﻧﺪﻩ ﺷﺪﻩ ﺑﻠﻮﻙ‬

‫∆ ﺑﺸﻤﺎﺭﻩ ﺑﻠﻮﻙ ﻧﻮﺷﺘﻪ ﺷﺪﻩ * ﻣﺘﻮﺳﻂ ﻫﺰﻳﻨﻪ ﻧﻮﺷﺘﻦ ﺑﻠﻮﻙ‬

‫‪ ‬ﻧﻜﺘﻪ‪:‬ﻫﺰﻳﻨﻪ ﺑﺮﺍﻱ ﻧﻮﺷﺘﻦ ﻳﻚ ﺑﻠﻮﻙ ﺑﻴﺸﺘﺮ ﺍﺯ ﻫﺰﻳﻨﻪ ﺍﻱ ﺑﺮﺍﻱ ﺧﻮﺍﻧﺪﻥ ﻳﻚ‬

‫ﺑﻠﻮﻙ ﺍﺳﺖ‪.‬‬

‫ﺯﻣﺎﻥ ﻧﻮﺷﺘﻦ ﻳﻚ ﺑﻠﻮﻙ ﺍﺯ ﺯﻣﺎﻥ ﺷﺮﻭﻉ ﺑﻪ ﻧﻮﺷﺘﻦ ﺗﺎ ﺣﺼﻮﻝ ﺍﻃﻤﻴﻨﺎﻥ ﺍﺯ ﺻﺤﺖ ﻧﻮﺷﺘﻦ‬
‫ﺍﻃﻼﻋﺎﺕ ﻣﻲ ﺑﺎﺷﺪ‪.‬‬

‫ﺑﺮﺍﻱ ﺳﺎﺩﮔﻲ‪ ،‬ﻓﻘﻂ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺗﻌﺪﺍﺩﻱ ﺑﻼﻙ ﻫﺎﻱ ﻣﻨﺘﻘﻞ ﺷﺪﻩ ﺍﺯ ﺩﻳﺴﻚ ﻭ ﺗﻌﺪﺍﺩ ﺍﺯ‬
‫ﺟﺴﺘﺠﻮﻫﺎ ﻫﺰﻳﻨﻪ ﻫﺎ ﺭﺍ ﻣﺤﺎﺳﺒﻪ ﻣﻲ ﻛﻨﻴﻢ ‪.‬‬

‫‪ ‬ﺯﻣﺎﻥ ﺍﻧﺘﻘﺎﻝ ﻳﻚ ﺑﻼﻙ ﻳﺎ ‪tT‬‬

‫‪ ‬ﺯﻣﺎﻥ ﺑﺮﺍﻱ ﻳﻚ ﺟﺴﺘﺠﻮ ﻳﺎ ‪tS‬‬

‫‪ ‬ﻫﺰﻳﻨﻪ ﺑﺮﺍﻱ ﺍﻧﺘﻘﺎﻝ ﺗﻌﺪﺍﺩ ‪ b‬ﺑﻼﻙ ﻭ ‪S‬ﺟﺴﺘﺠﻮ ﺧﻮﺍﻫﺪ ﺑﻮﺩ‪:‬‬

‫‪b * tT + S * tS‬‬
‫ﺑﺮﺍﻱ ﺳﺎﺩﮔﻲ ﻛﺎﺭ ﺍﺭ ﻫﺰﻳﻨﻪ ﻫﺎﻱ ﭘﺮﺩﺍﺯﺷﮕﺮ ﻣﺮﻛﺰﻱ ﺻﺮﻑ ﻧﻈﺮ ﻣﻲ ﻛﻨﻴﻢ ‪.‬‬

‫‪ ‬ﻧﻜﺘﻪ ‪ :‬ﺩﺭ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﻭﺍﻗﻌﻲ ﻫﺰﻳﻨﻪ ﻫﺎﻱ ‪ CPU‬ﺣﺘﻤﺎ ﺑﻪ ﺣﺴﺎﺏ ﻣﻲ ﺁﻳﺪ‪.‬‬

‫ﺑﺮﺍﻱ ﻧﻮﺷﺘﻦ ﺧﺮﻭﺟﻲ ﺭﻭﻱ ﺩﻳﺴﻚ ﻫﺰﻳﻨﻪ ﺍﻱ ﺭﺍ ﺩﺭ ﻓﺮﻣﻮﻝ ﻫﺰﻳﻨﻪ ﻫﺎ ﺑﻪ ﺣﺴﺎﺏ ﻧﻤﻲ ﺁﻭﺭﻳﻢ ‪.‬‬

‫ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻓﻀﺎﻱ ﺍﺿﺎﻓﻲ ﺑﺎﻓﺮ ﻣﻲ ﺗﻮﺍﻥ ﻫﺰﻳﻨﻪ ﻩ ﺍﻱ ﻭﺭﻭﺩ ﻭ ﺧﺮﻭﺝ ﺍﻃﻼﻋﺎﺕ ﺑﻪ ﺩﻳﺴﻚ ﺭﺍ‬
‫ﻛﺎﻫﺶ ﺩﺍﺩ ‪.‬‬

‫‪ ‬ﻣﻘﺪﺍﺭ ﻭﺍﻗﻌﻲ ﺣﺎﻓﻈﻪ ﺑﺎﻓﺮ ﻗﺎﺑﻞ ﺩﺳﺘﺮﺱ ﺑﺴﺘﮕﻲ ﺑﻪ ﻓﺮﺁﻳﻨﺪﻫﺎﻱ ﻫﻤﺰﻣﺎﻥ ﻭ ﺷﻨﺎﺧﺘﻪ‬
‫ﺷﺪﻩ ﺩﺭ ﺣﻴﻦ ﺍﺟﺮﺍﻱ ﺳﻴﺴﺘﻢ ﻋﺎﻣﻞ ﺩﺍﺭﺩ ‪.‬‬

‫‪141‬‬
 ‫‪ ‬ﺍﻏﻠﺐ ﺑﺪﺗﺮﻳﻦ ﺑﺮﺁﻭﺭﺩ ﻣﻤﻜﻦ ﺭﺍ ﺩﺭ ﻧﻈﺮ ﻣﻲ ﮔﻴﺮﻳﻢ ﻛﻪ ﺣﺪﺍﻗﻞ ﻣﻘﺪﺍﺭ ﺣﺎﻓﻈﻪ‬
‫ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﺑﺮﺍﻱ ﻋﻤﻠﻴﺎﺕ ﺩﺭ ﺩﺳﺘﺮﺱ ﺍﺳﺖ‬

‫ﺩﺍﺩﻩ ﻫﺎﻱ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﻣﻤﻜﻦ ﺍﺳﺖ ﺩﺭ ﺣﺎﻝ ﺣﺎﺿﺮ ﺩﺭ ﺑﺎﻓﺮ ﻗﺮﺍﺭ ﺑﮕﻴﺮﻧﺪ ﻭ ﺍﺯ ﺧﻮﺍﻧﺪﻥ ﻳﺎ ﻧﻮﺷﺘﻦ‬
‫ﺭﻭﻱ ﺩﻳﺴﻚ ﺍﺟﺘﻨﺎﺏ ﻛﻨﻨﺪ‬

‫‪ ‬ﺍﻣﺎ ﻫﺎﺭﺩ ﺩﻳﺴﻚ ﺯﻣﺎﻥ ﺭﺍ ﺑﺮﺍﻱ ﺑﺮﺁﻭﺭﺩ ﻫﺰﻳﻨﻪ ﻫﺎﺩﺭ ﻧﻈﺮ ﻣﻲ ﮔﻴﺮﺩ ‪.‬‬

‫ﻋﻤﻠﻴﺎﺕ ﺍﻧﺘﺨﺎﺑﻲ‬

‫ﺍﺳﻜﻦ ﻓﺎﻳﻞ‪:‬‬

‫ﺍ‬ ‫‪ ‬ﺍﻟﮕﻮﺭﻳﺘﻢ‪)1‬ﺟﺴﺘﺠﻮﻱ ﺧﻄﻲ ( ﺗﻤﺎﻡ ﻓﺎﻳﻞ ﺭﺍ ﻣﻮﺭﺩ ﺟﺴﺘﺠﻮ ﻗﺮﺍﺭ ﺩﺍﺩﻩ ﻭ ﺗﻤﺎﻡ ﺭﻛﻮﺭﺩﻫﺎ ﺭﺍ ﺏ‬
‫ﺷﺮﻁ ﺗﻌﻴﻴﻦ ﺷﺪﻩ ﺗﺴﺖ ﻣﻲ ﻛﻨﺪ ﺗﺎ ﺷﺮﻁ ﺩﺭﻫﺮ ﺭﻛﻮﺭﺩ ﺻﺪﻕ ﻛﻨﺪ ﻭ ﺍﻥ ﺭﺍ ﻧﻤﺎﻳﺶ ﻣﻲ ﺩﻫﺪ‬

‫‪Cost estimate = br block transfers + 1 seek‬‬

‫‪Br‬‬
‫ﺗﻌﺪﺍﺩ ﺑﻼﻙ ﻫﺎﻳﻲ ﻛﻪ ﺷﺮﻁ ﺩﺭ ﺭﻛﻮﺭﺩﻫﺎﻱ ﺁﻥ ﺻﺪﻕ ﻛﺮﺩﻩ ﺍﺳﺖ ﺍﺯ ﺭﺍﺑﻄﻪ ‪r‬‬

‫ﻣﺘﻮﻗﻒ‬ ‫ﺍﮔﺮ ﺍﻧﺘﺨﺎﺏ ﺩﺭ ﻳﻚ ﻭﻳﮋﮔﻲ ﻛﻠﻴﺪﻱ ﺍﺳﺖ‪ ،‬ﻣﻲ ﺗﻮﺍﻧﺪ ﺩﺭ ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﺭﻛﻮﺭﺩ ﺭﺍ‬
‫ﻛﻨﺪ‬

‫‪cost = (br /2) block transfers + 1 seek ‬‬

‫ﺟﺴﺘﺠﻮﻱ ﺧﻄﻲ ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﻪ ﻛﺎﺭ ﺑﺮﺩﻩ ﻣﻲ ﺷﻮﺩ ﺑﺎ ﺻﺮﻑ ﻧﻈﺮ ﺩﺍﺷﺘﻦ ﺍﺯ ‪:‬‬

‫‪ ‬ﺷﺮﺍﻳﻂ ﺍﻧﺘﺨﺎﺏ‬

‫‪ ‬ﺗﺮﺗﻴﺐ ﺭﻛﻮﺭﺩﻫﺎ ﺩﺭ ﻓﺎﻳﻞ‬

‫‪ ‬ﺩﺭ ﺩﺳﺘﺮﺱ ﺑﻮﺩﻥ ﻳﻚ ﺷﺎﺧﺺ‬

‫‪ ‬ﺟﺴﺘﺠﻮﻱ ﺩﻭﺩﻭﻳﻲ ﺑﺮﺍﻱ ﺩﺍﺩﻩ ﻫﺎﻳﻲ ﻛﻪ ﺑﺼﻮﺭﺕ ﻣﺮﺗﺐ ﺫﺧﻴﺮﻩ ﻧﺸﺪﻩ ﺍﻧﺪ ﻣﻌﻨﺎﻳﻲ ﻧﺨﻮﺍﻫﺪ‬
‫ﺩﺍﺷﺖ ﺑﺠﺰ ﺯﻣﺎﻧﻲ ﻛﻪ ﺍﻳﻨﺪﻛﺲ ﻫﺎ ﻗﺎﺑﻞ ﺩﺳﺘﺮﺳﻲ ﺑﺎﺷﻨﺪ ‪.‬‬
‫‪142‬‬
‫‪Selections Using Indices‬‬

‫♦ ‪ : Index scan‬ﺍﻟﮕﻮﺭﻳﺘﻢ ﻫﺎﻱ ﺟﺴﺘﺠﻮ ﻛﻪ ﺍﺯ ﺷﺎﺧﺺ )ﻋﻨﻮﺍﻥ( ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﻛﻨﻨﺪ‪.‬‬

‫‪ ‬ﺷﺮﺍﻳﻂ ﺍﻧﺘﺨﺎﺏ ﺑﺮﺍﻱ ﺟﺴﺘﺠﻮ ﺑﺎﻳﺪ ﺑﺮ ﺭﻭﻱ ﻛﻠﻴﺪ ﺷﺎﺧﺺ ﺑﺎﺷﺪ ‪.‬‬

‫♦ )‪ : A2 (primary index, equality on key‬ﺑﺎﺯﻳﺎﺑﻲ ﺗﻨﻬﺎ ﻳﻚ ﺭﻛﻮﺭﺩ ﻛﻪ ﺷﺮﻁ ﺯﻳﺮ‬

‫ﺩﺭ ﺁﻥ ﺻﺎﺩﻕ ﺍﺳﺖ‪:‬‬

‫)‪Cost = (hi + 1) * (tT + tS‬‬

‫♦ )‪ : A3 (primary index, equality on non key‬ﺑﺎﺯﻳﺎﺑﻲ ﭼﻨﺪ ﺭﻛﻮﺭﺩ‪ ،‬ﻛﻪ‬

‫ﺭﻛﻮﺭﺩﻫﺎ ﺑﺼﻮﺭﺕ ﻣﺘﻮﺍﻟﻲ ﺧﻮﺍﻫﻨﺪ ﺑﻮﺩ ‪.‬‬

‫‪ :Let b ‬ﺗﻌﺪﺍ ﺑﻠﻮﻙ ﻫﺎﻳﻲ ﻛﻪ ﺭﻛﻮﺭﺩﻫﺎﻱ ﺁﻥ ﻣﻄﺎﺑﻘﺖ ﺩﺍﺭﺩ )ﺷﺮﻁ ﺩﺭ ﺁﻥ ﻫﺎ‬

‫ﺻﺎﺩﻕ ﺑﻮﺩﻩ ﺍﺳﺖ (‪:‬‬

‫‪Cost = hi * (tT + tS) + tS + tT * b‬‬

‫‪ :A4 (secondary index, equality on nonkey). ‬ﺑﺎﺯﻳﺎﺑﻲ ﻳﻚ ﺗﻚ ﺭﻛﻮﺭﺩ ﺩﺭ‬

‫ﺻﻮﺭﺗﻴﻜﻪ ﻛﻠﻴﺪ ﺟﺴﺘﺠﻮ ﻛﻠﻴﺪ ﻛﺎﻧﺪﻳﺪ ﺍﺳﺖ‬

‫)‪Cost = (hi + 1) * (tT + tS‬‬

‫‪ ‬ﺑﺎﺯﻳﺎﺑﻲ ﻳﻚ ﺗﻚ ﺭﻛﻮﺭﺩ ﺩﺭ ﺻﻮﺭﺗﻴﻜﻪ ﻛﻠﻴﺪ ﺟﺴﺘﺠﻮ ﻛﻠﻴﺪ ﻛﺎﻧﺪﻳﺪ ﻧﻴﺴﺖ ‪.‬‬

‫‪ ‬ﻫﺮﻳﻚ ﺍﺯ ‪ n‬ﺭﻛﻮﺭﺩﻱ ﻛﻪ ﺗﻄﺒﻴﻖ ﺩﺍﺭﻧﺪ ﻣﻤﻜﻦ ﺍﺳﺖ ﺩﺭ ﺑﻼﻙ ﻫﺎﻱ ﻣﺨﺘﻠﻔﻲ‬

‫ﺑﺎﺷﻨﺪ‪.‬‬

‫)‪Cost = (hi + n) * (tT + tS‬‬

‫‪ ‬ﻣﻤﻜﻦ ﺍﺳﺖ ﻫﺰﻳﻨﻪ ﺑﺴﻴﺎﺭ ﺑﺎﻻﻳﻲ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ ‪.‬‬

‫‪143‬‬
 ‫ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﺍﺩﻏﺎﻣﻲ ﺧﺎﺭﺟﻲ‬

‫ﺍﺟﺮﺍﻱ ﺗﻜﻨﻴﻚ ﻣﺮﺗﺐ ﺳﺎﺯﻱ‪:‬‬

‫‪ .1‬ﺩﺭ ﺣﺎﻟﺖ ﺍﻭﻟﻴﻪ ‪ i‬ﺻﻔﺮ ﺩﺭ ﻧﻈﺮ ﮔﺮﻓﺘﻪ ﺷﻮﺩ‪.‬‬

‫‪ .2‬ﻣﺮﺍﺣﻞ ﺯﻳﺮ ﺗﺎ ﺍﻧﺘﻬﺎ ﺗﻜﺮﺍﺭ ﺷﻮﺩ‪:‬‬

‫‪ .a‬ﺧﻮﺍﻧﺪﻥ ﺑﻼﻙ ﺭﺍﺑﻄﻪ ﺍﻱ ﺍﺯ ﺣﺎﻓﻈﻪ‬
‫‪ .b‬ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﺩﺭ ﺑﻼﻙ ﻫﺎﻱ ﺣﺎﻓﻈﻪ‬
‫‪ .c‬ﻧﻮﺷﺘﻦ ﺩﺍﺩﻩ ﻫﺎﻱ ﻣﺮﺗﺐ ﺷﺪﻩ ﺑﺮﺍﻱ ﺍﺟﺮﺍ‬

‫‪ .3‬ﻣﻘﺪﺍﺭ ﻧﻬﺎﻳﻲ ‪ i‬ﺑﻪ ‪ N‬ﺗﻌﻠﻖ ﮔﻴﺮﺩ‪.‬‬

‫ﺍﺟﺮﺍﻱ ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﺍﺩﻏﺎﻣﻲ ) ‪ n‬ﺭﺍﻩ ﺍﺩﻏﺎﻡ(‬

‫ﻓﺮﺽ ﻣﻲ ﻛﻨﻴﻢ ‪: N < M‬‬

‫‪ N .1‬ﺑﻼﻙ ﺣﺎﻓﻈﻪ ﺑﺮﺍﻱ ﺑﺎﻓﺮ ﻭﺭﻭﺩﻱ ﻭ ﻳﻚ ﺑﻼﻙ ﺑﺮﺍﻱ ﺑﺎﻓﺮ ﺧﺮﻭﺟﻲ ﺩﺭ ﻧﻈﺮ‬
‫ﮔﺮﻓﺘﻪ ﻣﻲ ﺷﻮﺩ‪ .‬ﺍﻭﻟﻴﻦ ﺑﻼﻙ ﻫﺮ ﺻﻔﺤﻪ ﺑﺎﻓﺮ ﺧﻮﺍﻧﺪﻩ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪ .2‬ﺗﻜﺮﺍﺭ‬

‫‪ .1‬ﺍﻧﺘﺨﺎﺏ ﺍﻭﻟﻴﻦ ﺭﻛﻮﺭﺩ ﻣﺮﺗﺐ ﺷﺪﻩ ﺩﺭ ﻫﻤﻪ ﺻﻔﺤﺎﺕ ﺑﺎﻓﺮ ‪.‬‬

‫‪ .2‬ﻧﻮﺷﺘﻦ ﺭﻛﻮﺭﺩ ﺩﺭ ﺑﺎﻓﺮ ﺧﺮﻭﺟﻲ‪.‬‬

‫‪ .3‬ﺣﺬﻑ ﺭﻛﻮﺭﺩ ﺍﺯ ﺑﺎﻓﺮ ﻭﺭﻭﺩﻱ ﺍﻟﺒﺘﻪ ﺩﺭ ﺻﻮﺭﺗﻴﻜﻪ‪ ،‬ﺻﻔﺤﻪ ﺑﺎﻓﺮ ﺧﺎﻟﻲ ﺷﺪﻩ ﺑﺎﺷﺪ‬
‫ﺳﭙﺲ ﺑﻼﻙ ﺑﻌﺪﻱ ﺍﺟﺮﺍ ﺷﺪﻩ ﺩﺭ ﺑﺎﻓﺮ ﺧﻮﺍﻧﺪﻩ ﺷﺪﻩ ﺑﺎﺷﺪ ‪.‬‬

‫‪ .3‬ﺗﺎ ﺍﻳﻨﻜﻪ ﻫﻤﻪ ﺻﻔﺤﺎﺕ ﺑﺎﻓﺮ ﻭﺭﻭﺩﻱ ﺧﺎﻟﻲ ﺷﻮﺩ‪.‬‬

‫‪144‬‬
 ‫ﺍﮔﺮ ‪ N ≥ M‬ﭼﻨﺪ ﻣﺮﺣﻠﻪ ﺑﺎﻳﺪ ﺍﻧﺠﺎﻡ ﺷﻮﺩ‪:‬‬

‫ﺩﺭ ﻫﺮ ﮔﺬﺭﮔﺎﻩ ﻳﺎ ﭘﺎﺱ‪‌ ،‬ﮔﺮﻭﻩ ﻫﺎﻱ ﺑﻪ ﻫﻢ ﭘﻴﻮﺳﺘﻪ ﺍﺯ ‪ M-1‬ﺍﺟﺮﺍ ﻭ ﺑﺎ ﻫﻢ ﺍﺩﻏﺎﻡ ﻣﻲ‬

‫ﺷﻮﻧﺪ‪.‬‬

‫ﺩﺭ ﻫﺮ ﭘﺎﺱ ﻓﺎﻛﺘﻮﺭﻱ ﺍﺯ ‪ M‬ﺑﺎ ﺍﺟﺮﺍﻱ ‪ M-1‬ﺗﻮﻟﻴﺪ ﻣﻲ ﺷﻮﺩ ﻭ ﺍﺟﺮﺍ ﺑﺎ ﻓﺎﻛﺘﻮﺭﻱ ﻛﻪ‬
‫ﺗﻮﻟﻴﺪ ﺷﺪﻩ ﺍﺩﺍﻣﻪ ﭘﻴﺪﺍ ﻣﻲ ﻛﻨﺪ‪ ) .‬ﻣﻘﺪﺍﺭ ‪ M‬ﺑﺮﺍﻱ ﻣﺮﺣﻠﻪ ﺑﻌﺪ ﻳﻜﻲ ﻛﻢ ﻣﻲ ﺷﻮﺩ ‪(.‬‬

‫‪ Θ‬ﺑﻌﻨﻮﺍﻥ ﻣﺜﺎﻝ‪ ،‬ﺍﮔﺮ ‪ M=11‬ﺑﺎﺷﺪ‪ ،‬ﻭ ‪ 90‬ﺍﺟﺮﺍ ﺍﻧﺠﺎﻡ ﺷﻮﺩ‪ ،‬ﻳﻚ ﭘﺎﺱ ﺗﻌﺪﺍﺩ‬
‫ﺍﺟﺮﺍﻫﺎ ﺭﺍ ﺑﻪ ‪ 9‬ﻛﺎﻫﺶ ﺩﺍﺩﻩ‪ ،‬ﻛﻪ ﻫﺮ ﻛﺪﺍﻡ ‪ 10‬ﺑﺮﺍﺑﺮ ﺍﺟﺮﺍﻱ ﺍﻭﻟﻴﻪ ﺍﺳﺖ‪.‬‬

‫ﭘﺎﺱ ﻫﺎ ﭘﺸﺖ ﺳﺮﻫﻢ ﻭ ﻣﻜﺮﺭﺍ ﺍﺟﺮﺍ ﺷﺪﻩ ﺗﺎ ﻫﻤﻪ ﺑﺎ ﻫﻢ ﺍﺩﻏﺎﻡ ﺷﻮﻧﺪ ‪.‬‬

‫ﻣﺜﺎﻝ‪ :‬ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﺧﺎﺭﺟﻲ ﺗﻮﺳﻂ ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﺍﺩﻏﺎﻣﻲ‬

‫ﺑﺮﺁﻭﺭﺩ ﻫﺰﻳﻨﻪ‪:‬‬

‫ﺍﺟﺮﺍﻱ ﺍﺩﻏﺎﻡ ﻫﺮ ﺑﻠﻮﻙ ﻣﻨﺠﺮ ﺑﻪ ﺗﻌﺪﺍﺩ ﺯﻳﺎﺩﻱ ﺟﺴﺘﺠﻮ ﻣﻴﺸﻮﺩ ‪.‬‬

‫‪145‬‬
‫‪ ‬ﺩﺭ ﺍﺯﺍﻱ ﺍﺳﺘﻔﺎﺩﻩ ‪ bb‬ﺑﻠﻮﻙ ﻫﺎﻱ ﺑﺎﻓﺮ ﺩﺭ ﻫﺮ ﺍﺟﺮﺍ )ﺧﻮﺍﻧﺪﻥ ﻭ ﻧﻮﺷﺘﻦ ‪bb‬‬
‫ﺑﻠﻮﻙ ﺩﺭ ﻳﻚ ﺯﻣﺎﻥ (‪.‬‬

‫ﺍﺩﻏﺎﻡ ﺍﺟﺮﺍ ﺷﺪﻩ ﺩﺭ ﻳﻚ ﭘﺎﺱ‪M/bb–1 :‬‬

‫ﺗﻌﺪﺍﺩ ﻛﻞ ﺍﺩﻏﺎﻡ ﺩﺭ ﻫﺮ ﭘﺎﺱ ‪log M/bb–1(br/M)‬‬

‫ﺑﻠﻮﻙ ﻫﺎﻱ ﻣﻨﺘﻘﻞ ﺷﺪﻩ ﺑﺮﺍﻱ ﺍﺟﺮﺍﻱ ﺍﻭﻟﻴﻪ ﻫﺮ ﭘﺎﺱ ‪ 2br‬ﺍﺳﺖ ‪.‬‬

‫ﺩﺭ ﺁﺧﺮﻳﻦ ﭘﺎﺱ ﻫﺰﻳﻨﻪ ﻧﻮﺷﺘﻦ ﺣﺴﺎﺏ ﻧﻤﻲ ﺷﻮﺩ ‪.‬‬

‫‪ ‬ﻣﺎ ﺍﺯ ﻫﺰﻳﻨﻪ ‪ Write‬ﻧﻬﺎﻳﻲ ﺑﺮﺍﻱ ﺗﻤﺎﻡ ﻋﻤﻠﻴﺎﺕ ﻫﺎ ﺻﺮﻑ ﻧﻈﺮ ﻣﻲ ﻛﻨﻴﻢ‬

‫ﭼﺮﺍ ﻛﻪ ﻣﻤﻜﻦ ﺍﺳﺖ ﺧﺮﻭﺟﻲ ﺑﻪ ﻋﻤﻠﻴﺎﺕ ﺩﻳﮕﺮﻱ ﻓﺮﺳﺘﺎﺩﻩ ﺷﺪﻩ ﺑﺎﺷﺪ‬
‫ﻭ ﻫﻴﭻ ﮔﻮﻧﻪ ﻋﻤﻠﻴﺎﺕ ‪ write‬ﺭﻭﻱ ﺩﻳﺴﻚ ﺍﻧﺠﺎﻡ ﻧﺸﻮﺩ‪.‬‬

‫ﺑﻨﺎﺑﺮﺍﻳﻦ ﺗﻌﺪﺍﺩ ﻛﻞ ﻧﻘﻞ ﻭﺍﻧﺘﻘﺎﻻﺕ ﺑﺮﺍﻱ ﺍﻳﻦ ﻧﻮﻉ ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﺧﻮﺍﻫﺪ ﺑﻮﺩ ‪.‬‬
‫)‪br ( 2 log M/bb–1 (br / M) + 1‬‬
‫ﻫﺰﻳﻨﻪ ﻫﺎﻱ ﺟﺴﺘﺠﻮ‪:‬‬

‫ﺩﺭ ﻃﻮﻝ ﺍﺟﺮﺍ ‪ :‬ﺩﺭ ﻫﺮ ﺍﺟﺮﺍﻳﻚ ﺟﺴﺘﺠﻮ ﺑﺮﺍﻱ ﺧﻮﺍﻧﺪﻥ ﻭ ﻳﻚ ﺟﺴﺘﺠﻮ ﺑﺮﺍﻱ ﻧﻮﺷﺘﻦ‬
‫ﻭﺟﻮﺩ ﺩﺍﺭﺩ‬

‫‪2 br / M‬‬

‫ﻃﻮﻝ ﻓﺎﺯ ﺍﺩﻏﺎﻡ ﻧﻴﺎﺯ ﺩﺍﺭﺩ ﺑﻪ ‪ 2 br / bb‬ﺟﺴﺘﺠﻮ ﺑﺮﺍﻱ ﻫﺮ ﺍﺩﻏﺎﻡ ﺩﺭ ﻫﺮﭘﺎﺱ )ﺑﺠﺰ‬
‫ﻣﺮﺣﻠﻪ ﻧﻬﺎﻳﻲ ﻛﻪ ﻧﻴﺎﺯﻱ ﺑﻪ ﻧﻮﺷﺘﻦ ﻧﻴﺴﺖ (‬

‫ﻣﻘﺪﺍﺭ ﻧﻬﺎﻳﻲ ﺟﺴﺘﺠﻮ ﺧﻮﺍﻫﺪ ﺷﺪ ‪:‬‬

‫)‪2 br / M + br / bb (2 logM/bb–1(br / M) -1‬‬

‫‪146‬‬
 ‫ﺍﻟﮕﻮﺭﻳﺘﻢ ﺑﻠﻮﻙ ﻫﺎﻱ ﺣﻠﻘﻪ ﻫﺎﻱ ﺗﻮﺩﺭﺗﻮ‬

‫‪s‬‬ ‫‪r‬‬ ‫ﻣﺤﺎﺳﺒﻪ ﺗﺘﺎ‬

‫‪for each tuple tr in r do begin‬‬

‫‪for each tuple ts in s do begin‬‬
‫‪test pair (tr,ts) to see if they satisfy the join condition θ‬‬

‫‪if they do, add tr • ts to the result.‬‬

‫‪end‬‬
‫‪end‬‬
‫ﻋﻨﻮﺍﻥ ﺍﺭﺗﺒﺎﻁ ﺩﺭﻭﻧﻲ ﺩﺭﻳﻚ ﭘﻴﻮﻧﺪ ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﻧﺪ‪.‬‬
‫‪ r‬ﺑﻪ ﻋﻨﻮﺍﻥ ﺍﺭﺗﺒﺎﻁ ﺑﻴﺮﻭﻧﻲ ﻭ ‪ s‬ﺏ‬

‫ﺑﺪﻭﻥ ﺩﺭﻧﻈﺮﮔﺮﻓﺘﻦ ﺷﺎﺧﺺ ﻫﺎ ﺑﺎ ﻫﺮ ﺷﺮﺍﻳﻄﻲ ﭘﺬﻳﺮﺵ ﻣﻲ ﺗﻮﺍﻧﺪ ﺍﻧﺠﺎﻡ ﮔﻴﺮﺩ‪.‬‬

‫ﺩﺭ ﺍﺭﺗﺒﺎﻃﺎﺕ ﺩﻭﻃﺮﻓﻪ ﻫﺰﻳﻨﻪ ﻫﺎ ﺑﺎﻳﺪ ﺑﺮﺭﺳﻲ ﺷﻮﺩ‪.‬‬

‫ﻫﺰﻳﻨﻪ ﺑﺮﺍﻱ ﺑﺪﺗﺮﻳﻦ ﺣﺎﻟﺖ ﺑﺮﺁﻭﺭﺩ ﺷﺪﻩ ﺍﺳﺖ ﻛﻪ ﻓﻀﺎﻱ ﻣﻮﺟﺪ ﺩﺭ ﺣﺎﻓﻈﻪ ﺗﻨﻬﺎ ﺑﺮﺍﻱ ﻳﻚ ﺑﻼﻙ‬
‫ﺍﺯ ﻫﺮ ﺭﺍﺑﻄﻪ ﻛﺎﻓﻲ ﺍﺳﺖ‪.‬‬
‫‪nr ∗ bs + br‬‬
‫‪nr + br‬‬
‫ﺭﺍﺑﻄﻪ ﻛﻮﭼﻜﻲ ﻛﻪ ﺑﻪ ﻃﻮﺭﻛﺎﻣﻞ ﺩﺭﺣﺎﻓﻈﻪ ﻗﺮﺍﺭ ﺩﺍﺭﺩ ﺑﻪ ﻋﻨﻮﺍﻥ ﺭﺍﺑﻄﻪ ﺩﺭﻭﻧﻲ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ‬
‫ﺷﻮﺩ‪.‬‬

‫‪ ‬ﻛﺎﻫﺶ ﻫﺰﻳﻨﻪ ﺑﻼﻙ ﺍﻧﺘﻘﺎﻟﻲ‪br + bs :‬‬

‫ﺑﺮﺁﻭﺭﺩ ﻫﺰﻳﻨﻪ ﺑﺎﻓﺮﺽ ﺑﺪﺗﺮﻳﻦ ﺣﺎﻟﺖ ﺩﺭﺩﺳﺘﺮﺱ ﺑﻮﺩﻥ ﺣﺎﻓﻈﻪ‬

‫‪ ‬ﺩﺍﻧﺶ ﺁﻣﻮﺯﺍﻥ ﺑﻪ ﻋﻨﻮﺍﻥ ﺭﺍﺑﻄﻪ ﺧﺎﺭﺟﻲ‬

‫‪147‬‬
 ‫‪5000 ∗ 400 + 100 = 2,000,100 block transfers, n‬‬
‫‪5000 + 100 = 5100 seeks n‬‬
‫‪ ‬ﻣﺪﺗﻲ ﻛﻪ ﺭﺍﺑﻄﻪ ﺧﺎﺭﺟﻲ ﺑﻪ ﻃﻮﻝ ﻣﻲ ﺍﻧﺠﺎﻣﺪ‬

‫‪10000 ∗ 100 + 400 = 1,000,400 block transfers and n‬‬

‫‪10,400 seeks‬‬
‫ﺍﮔﺮ ﺍﺭﺗﺒﺎﻁ ﺩﺍﻧﺸﺠﻮﻳﻲ ﺑﻪ ﻃﻮﺭ ﻛﺎﻣﻞ ﺩﺭ ﺣﺎﻓﻈﻪ ﻗﺮﺍﺭ ﺑﮕﻴﺮ ﺩ ﻫﺰﻳﻨﻪ ﺑﺮﺁﻭﺭﺩ ﺷﺪﻩ ﺑﺮﺍﻱ‬
‫ﺍﻧﺘﻘﺎﻝ‪500‬ﺑﻼﻙ ﺧﻮﺍﻫﺪ ﺑﻮﺩ‪.‬‬

‫ﺍﻟﮕﻮﺭﻳﺘﻢ ‪ Block nested-loops‬ﺍﻧﺘﺨﺎﺏ ﺑﺮﺗﺮﻱ ﺍﺳﺖ‪.‬‬

‫ﻧﻮﻉ ﺩﻳﮕﺮ ﺍﺯ ﺣﻠﻘﻪ ﻫﺎﻱ ﺗﻮ ﺩﺭ ﺗﻮ ﭘﻴﻮﺳﺘﻦ ﺍﺳﺖ ﻛﻪ ﺩﺭ ﺁﻥ ﻫﺮ ﺑﻠﻮﻙ ﺍﺯ ﺭﺍﺑﻄﻪ ﺩﺭﻭﻧﻲ ﺑﺎ‬
‫ﻫﺮ ﺑﻠﻮﻙ ﺍﺯ ﺭﺍﺑﻄﻪ ﺑﻴﺮﻭﻧﻲ ﻳﻚ ﺯﻭﺝ ﻣﻲ ﺷﻮﻧﺪ‪.‬‬

‫‪for each block Br of r do begin‬‬

‫‪for each block Bs of s do begin‬‬
‫‪for each tuple tr in Br do begin‬‬
‫‪for each tuple ts in Bs do begin‬‬
‫‪Check if (tr,ts) satisfy the join condition‬‬
‫‪if they do, add tr • ts to the result.‬‬
‫‪end‬‬
‫‪end‬‬
‫‪end‬‬
‫‪end‬‬
‫ﺑﺮﺁﻭﺭﺩ ﺑﺪﺗﺮﻳﻦ ﺷﺮﺍﻳﻂ ﺑﻪ ﺍﻳﻦ ﺷﻜﻞ ﺍﺳﺖ ‪:‬‬

‫‪br ∗ bs + br block transfers + 2 * br seeks‬‬

‫ﻫﺮﺑﻠﻮﻙ ﺩﺭﺭﺍﺑﻄﻪ ﺩﺭﻭﻧﻲ ‪s‬ﻛﻪ ﺑﻪ ﺍﺯﺍﻱ ﻫﺮ ﺑﻠﻮﻙ ﺭﺍﺑﻄﻪ ﺑﻴﺮﻭﻧﻲ ﻳﻜﺒﺎﺭ ﺧﻮﺍﻧﺪﻩ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫‪148‬‬
 ‫ﺑﻬﺘﺮﻳﻦ ﺑﺮﺁﻭﺭﺩ ﺧﻮﺍﻫﺪ ﺑﻮﺩ ‪:‬‬

‫‪br + bs block transfers + 2 seeks‬‬

‫ﺭﺍﻩ ﻫﺎﻳﻲ ﺑﺮﺍﻱ ﺑﻬﺒﻮﺩ ﺣﻠﻘﻪ ﻫﺎﻱ ﺗﻮ ﺩﺭ ﺗﻮ ﻭ ﺑﻠﻮﻙ ﺍﻟﮕﻮﺭﻳﺘﻢ ﻫﺎﻱ ﺣﻠﻘﻪ ﺗﻮ ﺩﺭ ﺗﻮ‪:‬‬

‫‪ l‬ﺩﺭ ﺑﻠﻮﻙ ﺣﻠﻘﻪ ﺗﻮ ﺩﺭ ﺗﻮ‪ ،‬ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ‪ M — 2‬ﺑﻠﻮﻙ ﺩﻳﺴﻚ ﻭﺍﺣﺪ ﺭﻭﺍﺑﻂ ﺑﻴﺮﻭﻧﻲ‬
‫ﻣﺴﺪﻭﺩ ﻣﻲ ﺷﻮﺩ‪ ،‬ﻛﻪ ﺩﺭ ﺁﻥ ‪ = M‬ﺍﻧﺪﺍﺯﻩ ﺑﻠﻮﻙ ﻫﺎﻱ ﺣﺎﻓﻈﻪ؛ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺩﻭ ﺑﻠﻮﻙ ﺑﺎﻗﻲ‬
‫ﻣﺎﻧﺪﻩ ﺑﺮﺍﻱ ﺑﺎﻓﺮ ﺭﺍﺑﻄﻪ ﺩﺭﻭﻧﻲ ﻭ ﺧﺮﻭﺟﻲ‬

‫‪Cost = br / (M-2) ∗ bs + br block transfers +2 br / (M-2) seeks‬‬

‫‪ l‬ﺍﮔﺮ ﻣﺸﺨﺼﻪ ﻫﺎﻱ ﭘﻴﻮﻧﺪ ﺷﻜﻠﻲ ﺍﺯ ﻳﻚ ﻛﻠﻴﺪ ﻳﺎ ﺭﺍﺑﻄﻪ ﺩﺍﺧﻠﻲ ﺑﺎﺷﺪ ﺩﺍﺧﻠﻲ ﺗﺮﻳﻦ ﺣﻠﻘﻪ‬
‫ﺩﺭ ﺍﺑﺘﺪﺍﻱ ﻛﺎﺭ ﻣﺘﻮﻗﻒ ﻣﻴﺸﻮﺩ‪.‬‬

‫‪ l‬ﺣﺮﻛﺖ)ﺍﺟﺮﺍﻱ( ﻣﺘﻨﺎﻭﺏ ﺣﻠﻘﻪ ﺩﺭﻭﻧﻲ ﺭﻭ ﺑﻪ ﺟﻠﻮ ﻭ ﻋﻘﺐ‪ ،‬ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺑﻠﻮﻙ ﻫﺎﻱ ﺑﺎﻗﻲ‬
‫ﻣﺎﻧﺪﻩ ﺩﺭ ﺑﺎﻓﺮ )ﺑﺎ ﺟﺎﻳﮕﺰﻳﻨﻲ ‪.(LRU‬‬

‫‪ ‬ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺷﺎﺧﺺ ﺩﺭ ﺍﺭﺗﺒﺎﻁ ﺩﺭﻭﻧﻲ ﺍﮔﺮ ﺩﺭ ﺩﺳﺘﺮﺱ ﺍﺳﺖ )ﺍﺳﻼﻳﺪ ﺑﻌﺪﻱ(‪.‬‬

‫ﭘﻴﻮﻧﺪ ﺣﻠﻘﻪ ﻫﺎﻱ ﺗﻮ ﺩﺭ ﺗﻮ‬

‫‪ n‬ﻣﺮﺍﺟﻌﺎﺕ ﺑﻪ ﺷﺎﺧﺺ ﻣﻲ ﺗﻮﺍﻧﺪ ﺟﺎﻳﮕﺰﻳﻦ ﺍﺳﻜﻦ ﻫﺎﻱ ﻓﺎﻳﻞ ﺷﻮﺩ ﺍﮔﺮ‬

‫‪ l‬ﭘﻴﻮﻧﺪ ‪ equi-join‬ﻳﺎ ‪ natural join‬ﺍﺳﺖ ﻭ‬

‫‪ l‬ﻣﺸﺨﺼﻪ ﭘﻴﻮﻧﺪ ﻗﺎﺑﻞ ﺩﺳﺘﺮﺱ ﺑﻮﺩﻥ ﻳﻚ ﺷﺎﺧﺺ ﺭﻭﻱ ﺍﺭﺗﺒﺎﻁ ﺩﺭﻭﻧﻲ ﺍﺳﺖ ‪.‬‬

‫‪ ‬ﻣﻲ ﺗﻮﺍﻥ ﻳﻚ ﺷﺎﺧﺺ ﻓﻘﻂ ﺑﺮﺍﻱ ﻣﺤﺎﺳﺒﻪ ﭘﻴﻮﻧﺪ ﻫﺎ ﺍﻳﺠﺎﺩ ﻛﺮﺩ‪.‬‬

‫‪ n‬ﺑﺮﺍﻱ ﻫﺮ ﺗﺎﭘﻞ ‪ TR‬ﺩﺭ ﺭﺍﺑﻄﻪ ﻱ ﺑﻴﺮﻭﻧﻲ ‪ ،R‬ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺷﺎﺧﺺ ﺑﺮﺍﻱ ﻣﺮﺍﺟﻌﻪ ﺑﻪ ﺗﺎﭘﻞ ﻫﺎ ﺩﺭ ‪s‬‬
‫ﻛﻪ ﺷﺮﺍﻳﻂ ﭘﻴﻮﺳﺘﻦ ﺑﻪ ﺗﺎﭘﻞ ‪ TR‬ﺑﺮﺁﻭﺭﺩ ﺷﺪﻩ ﺍﺳﺖ ‪.‬‬

‫‪149‬‬
‫‪ n‬ﺑﺪﺗﺮﻳﻦ ﺣﺎﻟﺖ ‪ :‬ﻓﻀﺎﻳﺎﻓﺮ ﻓﻘﻂ ﺑﻪ ﺍﻧﺪﺍﺯﻩ ﻳﻚ ﺻﻔﺤﻪ ﺍﺯ ‪ r‬ﺑﺎﺷﺪﻭﺑﺮﺍﻱ ﻫﺮﺗﺎﭘﻞ ﺩﺭ ‪ r‬ﻣﺎ ﺑﻪ ﺷﺎﺧﺺ‬
‫ﻣﺮﺍﺟﻌﻪ ﻛﻨﻴﻢ‪.‬‬

‫‪Cost of the join: br (tT + tS) + nr ∗ c n‬‬

‫‪ l‬ﺟﺎﻳﻲ ﻛﻪ ‪ c‬ﻫﺴﺖ ﻫﺰﻳﻨﻪ ﺷﺎﺧﺺ ﻭ ﻭﺍﻛﺸﻲ ﻫﻤﻪ ﻣﻄﺎﺑﻖ ﺗﺎﻳﻞ ‪ s‬ﺑﺮﺍﻱ ﻳﻚ ﺗﺎﭘﻞ ﻳﺎ ‪r‬‬
‫ﺩﺭﻧﻈﺮ ﮔﺮﻓﺘﻪ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪ C‬ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﺮﺁﻭﺭﺩ ﻫﺰﻳﻨﻪ ﺷﻮﺩ ﺍﺯ ﺗﻨﻬﺎ ﻳﻚ ﺍﻧﺘﺨﺎﺏ ﺭﻭﻱ ﺗﺎﭘﻞ ‪ s‬ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺷﺮﺍﻳﻂ ﭘﻴﻮﻧﺪ‪.‬‬

‫‪ n‬ﺍﮔﺮﺷﺎﺧﺺ ﻫﺎﻱ ﻗﺎﺑﻞ ﺩﺳﺘﺮﺱ ﺗﺎﭘﻞ ‪ r‬ﻭ ‪ s‬ﺩﺍﺭﻱ ﻭﻳﮋﮔﻲ ﻫﺎﻱ ﭘﻴﻮﻧﺪ ﺑﺎﺷﻨﺪ ﺍﺯ ﺍﺭﺗﺒﺎﻁ ﺑﺎ ﺗﺎﭘﻞ‬
‫ﻫﺎﻱ ﻛﻤﺘﺮﻱ ﺑﻪ ﻋﻨﻮﺍﻥ ﺍﺭﺗﺒﺎﻁ ﺑﻴﺮﻭﻧﻲ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﻛﻨﻨﺪ ‪.‬‬

‫ﻣﺜﺎﻝ ﺍﺯ ﻫﺰﻳﻨﻪ ﻫﺎﻱ ﭘﻴﻮﻧﺪ ﺣﻠﻘﻪ ﻫﺎﻱ ﺗﻮ ﺩﺭ ﺗﻮ‬

‫ﻣﺤﺎﺳﺒﻪ ﻣﺜﺎﻝ ﺩﺍﻧﺸﺠﻮﻳﻲ ‪،‬ﺩﺍﻧﺸﺠﻮ ﺑﻪ ﻋﻨﻮﺍﻥ ﺭﺍﺑﻄﻪ ﺧﺎﺭﺟﻲ ﺩﺭﻧﻈﺮ ﮔﺮﻓﺘﻪ ﻣﻲ ﺷﻮﺩ ‪.‬‬

‫‪ID‬ﺑﺎ ﺳﻪ ﺷﺎﺧﺺ ﻛﻪ ﺩﺭ ﻫﺮ‬ ‫ﺍﺟﺎﺯﻩ ﺩﻫﻴﺪ ‪B+‬ﺑﻪ ﻋﻨﻮﺍﻥ ﻛﻠﻴﺪ ﺍﺻﻠﻲ ﮔﺮﻓﺘﻪ ﺷﻮﺩ ﺭﻭﻱ ﻓﻴﻠﺪ‬
‫ﻣﺮﺣﻠﻪ ‪ 20‬ﻭﺭﻭﺩﻱ ﺩﺍﺭﺩ‪.‬‬

‫ﺍﺯ ﺁﻧﺠﺎﻛﻪ ﺩﺳﺘﺮﺳﻲ ﺑﻪ ‪10000‬ﺗﺎﭘﻞ ﺑﺎ ﺍﺭﺗﻔﺎﻉ ﺩﺭﺧﺖ ‪ 4‬ﺯﻣﺎﻧﺒﺮ ﺍﺳﺖ ﻭ ﺑﻴﺸﺘﺮ ﺩﺳﺘﺮﺳﻲ ﺑﻪ‬
‫ﺩﺍﺩﻩ ﺍﺻﻠﻲ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﺍﺳﺖ‪.‬‬

‫ﺗﺎﭘﻞ‪ 5000‬ﺩﺍﻧﺸﺠﻮﻳﻲ ﺩﺭ ﻧﻈﺮ ﮔﺮﻓﺘﻪ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪: Cost of block nested loops join‬‬

‫‪400*100 + 100 = 40,100 block transfers + 2 * 100 = 200 ‬‬

‫‪seeks‬‬
‫‪ ‬ﺑﺎ ﻓﺮﺽ ﺑﺪﺗﺮﻳﻦ ﺣﺎﻟﺖ ﺣﺎﻓﻈﻪ‬

‫‪ ‬ﺷﺎﻳﺪ ﺑﺎ ﺣﺎﻓﻈﻪ ﺑﻴﺸﺘﺮ ﺑﻪ ﻃﻮﺭ ﻗﺎﺑﻞ ﺗﻮﺟﻬﻲ‬

‫‪: Cost of indexed nested loops join‬‬

‫‪150‬‬
‫‪100 + 5000 * 5 = 25,100 block transfers and seeks. ‬‬
‫‪ ‬ﻫﺰﻳﻨﻪ ﻫﺎﻱ ‪CPU‬ﺑﻪ ﺗﻈﺮ ﻣﻴﺮﺳﺪ ﻛﻪ ﻛﻤﺘﺮﺍﺯ ﭘﻴﻮﻧﺪ ﺑﻠﻮﻙ ﺣﻠﻘﻪ ﻫﺎﻱ ﺗﻮ ﺩﺭ ﺗﻮﺳﺖ ‪.‬‬

‫‪: Merge - Join‬‬

‫ﻣﺮﺗﺐ ﻛﺮﺩﻥ ﻫﺮﺩﻭﺍﺭﺗﺒﺎﻁ ﺑﺮ ﺍﺳﺎﺱ ﻓﻴﻠﺪ ﭘﻴﻮﻧﺪ)ﺍﮔﺮ ﻗﺒﻼ ﻣﺮﺗﺐ ﻧﺸﺪﻩ ﺑﺎﺷﺪ(‪.‬‬

‫ﺍﺩﻏﺎﻡ ﺭﻭﺍﺑﻂ ﻣﺮﺗﺐ ﺷﺪﻩ ﺑﺮﺍﻱ ﭘﻴﻮﻧﺪ ﺁﻧﻬﺎ‪:‬‬

‫‪ .1‬ﮔﺎﻡ ﭘﻴﻮﻧﺪ ﺷﺒﻴﻪ ﺑﻪ ﺍﺩﻏﺎﻡ ﺩﺭ ﺍﻟﮕﻮﺭﻳﺘﻢ ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﺍﺩﻏﺎﻣﻲ ﺍﺳﺖ ‪.‬‬

‫‪ .2‬ﺍﺧﺘﻼﻑ ﺍﺻﻠﻲ ﺩﺭ ﻫﻨﺪﻝ ﻛﺮﺩﻥ ﻣﻘﺎﺩﻳﺮ ﺷﺒﻴﻪ ﺑﻪ ﻫﻢ ﺩﺭ ﻓﺎﻳﻞ ﭘﻴﻮﺳﺘﮕﻲ ﺍﺳﺖ‪ ،‬ﻫﺮﺟﻔﺖ‬
‫ﺩﺭ ﻓﺎﻳﻞ ﭘﻴﻮﺳﺘﮕﻲ ﺑﺎﻳﺪ ﺩﺍﺭﺍﻱ ﻣﻘﺎﺩﻳﺮ ﻳﻜﺴﺎﻥ ﺑﺎﺵﻧﺪ ‪.‬‬

‫‪ ‬ﺍﻳﻦ ﺭﻭﺵ ﻣﻲ ﺗﻮﺍﻧﺪ ﺍﺳﺘﻔﺎﺩﻩ ﺷﻮﺩ ﻓﻘﻂ ﺑﺮﺍﻱ ﭘﻴﻮﻧﺪﻫﺎﻱ ﻃﺒﻴﻌﻲ ﻭ ﭘﻴﻮﻧﺪﻫﺎﻱ ﻣﻌﺎﺩﻝ ﺑﺎﺷﻨﺪ ‪.‬‬

‫‪ ‬ﻫﺮ ﺑﻠﻮﻙ ﻧﻴﺎﺯ ﺑﻪ ﺗﻨﻬﺎ ﻳﻚ ﺑﺎﺭ ﺧﻮﺍﻧﺪﻩ ﺷﺪﻥ ﺩﺍﺭﺩ )ﻓﺮﺽ ﺗﻤﺎﻡ ﺗﺎﭘﻞ ﻫﺎﺑﺮﺍﻳﮕﺮﻓﺘﻦ ﻫﺮ‬
‫ﻣﻘﺪﺍﺭﻣﺸﺨﺺ ﺍﺯ ﻓﻴﻠﺪ ﭘﻴﻮﺳﺘﮕﻲ ﺩﺭﺣﺎﻓﻈﻪ ﻗﺮﺍﺭ ﻣﻲ ﮔﻴﺮﺩ (‪.‬‬

‫‪ ‬ﺑﻨﺎﺑﺮﺍﻳﻦ ﻫﺰﻳﻨﻪ ﺍﻳﻦ ﺍﻟﮕﻮﺭﻳﺘﻢ‪:‬‬

‫‪br + bs block transfers + br / bb + bs / bb seeks‬‬
‫)ﺑﻌﻼﻭﻩ ﻫﺰﻳﻨﻪ ﻣﺮﺗﺐ ﺷﺪﻥ ﺩﺭﺻﻮﺭﺗﻲ ﻛﻪ ﻣﺮﺗﺐ ﻧﺒﺎﺷﺪ ( ‪.‬‬

‫‪151‬‬
 ‫‪ :hybrid merge-join ‬ﺍﮔﺮﺭﺍﺑﻄﻪ ﺍﻱ ﻣﺮﺗﺐ ﺷﺪﻩ ﻭ ﺍﺯ ﺳﻮﻱ ﺩﻳﮕﺮ ‪secondary B+-‬‬
‫‪tree index‬ﺭﻭﻱ ﻓﺎﻳﻞ ﭘﻴﻮﺳﺘﮕﻲ ﺍﺳﺖ‪.‬‬

‫ﺍﺩﻏﺎﻡ ﺭﺍﺑﻄﻪ ﻣﺮﺗﺐ ﺷﺪﻩ ﺑﺎ ﺑﺮگ ﻧﻮﺷﺘﻪ ﻫﺎﻱ ‪.B+-tree‬‬

‫ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﻧﺘﻴﺠﻪ ﺑﺮﺭﻭﻱ ﺍﺩﺭﺳﻲ ﺍﺯ ﺗﺎﭘﻞ ﺭﺍﺑﻄﻪ ﻣﺮﺗﺐ ﻧﺸﺪﻩ ‪.‬‬

‫ﺍﺳﻜﻦ ﺍﺭﺗﺒﺎﻁ ﻫﺎﻱ ﻧﺎﻣﺮﺗﺐ ﺑﺮﺍﻱ ﻧﻈﻢ ﺁﺩﺭﺱ ﻓﻴﺰﻳﻜﻲ ﻭ ﺍﺩﻏﺎﻡ ﺑﺎ ﻧﺘﻴﺠﻪ ﻗﺒﻠﻲ‪ ،‬ﺑﺮﺍﻱ‬
‫ﺟﺎﻳﮕﺰﻳﻨﻲ ﺁﺩﺭﺱ ﺗﻮﺳﻂ ﺗﺎﭘﻞ ﻭﺍﻗﻌﻲ‪.‬‬

‫‪ ‬ﺍﺳﻜﻦ ﻣﺘﻮﺍﻟﻲ ﻛﺎﺭﺍﻳﻲ ﺑﻴﺸﺘﺮﻱ ﻧﺴﺒﺖ ﻣﺮﺍﺟﻌﻪ ﺗﺼﺎﺩﻓﻲ ﺩﺍﺭﺩ ‪.‬‬

‫‪:Hash - Join‬‬

‫‪Hash-Join :‬ﺍﻟﮕﻮﺭﻳﺘﻢ‬

‫ﻣﺤﺎﺳﺒﻪ ‪ hash-join‬ﺩﺭ ‪ r‬ﻭ ‪ s‬ﺑﻪ ﺷﺮﺡ ﺯﻳﺮ ﺍﺳﺖ‪:‬‬

‫‪152‬‬
‫‪ .1‬ﭘﺎﺭﺗﻴﺸﻦ ﺑﻨﺪﻱ ﺭﺍﺑﻄﻪ ‪ s‬ﺍﺯ ﺗﺎﺑﻊ ﻫﺶ ‪ h‬ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﻛﻨﺪ ‪ .‬ﺩﺭ ﻫﻨﮕﺎﻡ ﭘﺎﺭﺗﻴﺸﻦ ﺑﻨﺪﻱ ﻳﻚ‬
‫ﺭﺍﺑﻄﻪ‪ ،‬ﻳﻚ ﺑﻠﻮﻙ ﺍﺯ ﺣﺎﻓﻈﻪ ﺑﻌﻨﻮﺍﻥ ﺑﺎﻓﺮ ﺧﺮﻭﺟﻲ ﺑﺮﺍﻱ ﻫﺮ ﭘﺎﺭﺗﻴﺸﻦ ﻣﺤﻔﻮﻅ ﻣﻲ ﺑﺎﺷﺪ ‪.‬‬

‫‪ .2‬ﭘﺎﺭﺗﻴﺸﻦ ﺑﻨﺪﻱ ‪ r‬ﻧﻴﺰ ﺑﻪ ﻫﻤﻴﻦ ﺗﺮﺗﻴﺐ‪.‬‬

‫‪ .3‬ﺑﺮﺍﻱ ﻫﺮ ‪:i‬‬

‫‪ hash‬ﺑﺮ ﺭﻭﻱ ﺣﺎﻓﻈﻪ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ‬ ‫• ﻓﺮﺍﺧﻮﺍﻧﻲ ‪ s‬ﺑﻪ ﺣﺎﻓﻈﻪ ﻭ ﺍﻳﺠﺎﺩ ﺷﺎﺧﺺ‬
‫ﻭﻳﮋﮔﻲ)ﻓﻴﻠﺪ( ﭘﻴﻮﺳﺘﮕﻲ‪.‬‬

‫ﺍﻳﻦ ﺷﺎﺧﺺ ‪ hash‬ﻳﻚ ﺗﺎﺑﻊ ﻣﺘﻔﺎﻭﺕ ‪ hash‬ﻗﺒﻞ ﺍﺯ ﻳﻚ ‪ h‬ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫• ﻓﺮﺍﺧﻮﺍﻧﻲ ‪tuble‬ﻫﺎ ﺩﺭ ‪ ri‬ﺍﺯ ﺩﻳﺴﻚ ﻳﻜﻲ ﻳﻜﻲ ﺍﻧﺠﺎﻡ ﻣﻲ ﺷﻮﺩ ‪ .‬ﺑﺮﺍﻱ ﻫﺮ ‪tuple tr‬‬
‫ﻣﺤﻠﻲ ﻣﻄﺎﺑﻖ ﺑﺎ ‪tuple ts‬ﺩﺭ ‪ si‬ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺷﺎﺧﺺ ‪hash‬ﺣﺎﻓﻈﻪ ﻗﺮﺍﺭ ﻣﻲ ﮔﻴﺮﺩ ‪.‬‬
‫ﺧﺮﻭﺟﻲ ﺗﺮﻛﻴﺒﻲ ﺍﺯ ﻭﻳﮋﮔﻲ ﻫﺎﻱ )ﻓﻴﻠﺪﻫﺎﻱ( ﺁﻧﻬﺎﺳﺖ‪.‬‬

‫ﺍﺭﺗﺒﺎﻁ ‪ s‬ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﺩ‪build input :‬‬

‫ﺭﺍﺑﻄﻪ ‪ r‬ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﺩ‪probe input :‬‬

‫ﻣﻘﺪﺍﺭ ‪ n‬ﻭ ﺗﺎﺑﻊ ﻫﺶ ‪h‬ﻃﻮﺭﻱ ﺍﻧﺘﺨﺎﺏ ﻣﻲ ﺷﻮﻧﺪ ﻛﻪ ‪ si‬ﺑﺎﻳﺪ ﺩﺭ ﺣﺎﻓﻈﻪ ﻣﺘﻨﺎﺳﺐ ﻗﺮﺍﺭ ﺑﮕﻴﺮﺩ‪.‬‬

‫• ﻣﻌﻤﻮﻻ ‪n‬ﺍﻧﺘﺨﺎﺏ ﻣﻲ ﺷﻮﺩ ﺍﺯ ‪ bs/M * f‬ﺟﺎﻫﺎﻳﻲ ﻛﻪ ‪ f‬ﺩﺭ ﻓﺮﻣﻮﻝ ﻗﺮﺍﺭ ﺩﺍﺭﺩ ﻳﻚ‬
‫”‪ “fudge factor‬ﺍﺳﺖ ﻛﻪ ﻋﺪﺩ ﺭﻧﺪﻱ ﺣﺪﻭﺩ ‪ 1.2‬ﻣﻲ ﺑﺎﺷﺪ‪.‬‬

‫• ﭘﺮﻭﺏ)ﻛﺎﻭﺷﮕﺮ( ﭘﺎﺭﺗﻴﺸﻦ ﻫﺎﻱ ﺭﺍﺑﻄﻪ ‪ si‬ﻧﻴﺎﺯﻱ ﺑﻪ ﺣﺎﻓﻈﻪ ﻣﻨﺎﺳﺐ ﻧﺪﺍﺭﺩ‪.‬‬

‫ﭘﺎﺭﺗﻴﺸﻦ ﺑﻨﺪﻱ‬ ‫‪ M‬ﺍﺯ ﺣﺎﻓﻈﻪ ﺑﺎﺷﺪ‪ ،‬ﻧﻴﺎﺯ ﺑﻪ‬ ‫ﺍﮔﺮ ﺗﻌﺪﺍﺩ ﭘﺎﺭﺗﻴﺸﻦ ‪ n‬ﺑﻴﺸﺘﺮ ﺍﺯ ﺗﻌﺪﺍﺩ ﺻﻔﺤﺎﺕ‬
‫ﺑﺎﺯﮔﺸﺘﻲ ﺍﺳﺖ‪.‬‬

‫• ‪instead of partitioning n ways, use M – 1 partitions for s‬‬

‫• ﺍﺯ ‪ M – 1‬ﭘﺎﺭﺗﻴﺸﻦ ﺑﺮﺍﻱ ‪ s‬ﺑﻪ ﺟﺎﻱ ‪ n‬ﭘﺎﺭﺗﻴﺸﻦ ﺑﻨﺪﻱ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﻛﻨﻴﻢ ‪.‬‬

‫• ﻋﻼﻭﻩ ﺑﺮ ‪ M – 1‬ﭘﺎﺭﺗﻴﺸﻦ ﺍﺯ ﻳﻚ ﺗﺎﺑﻊ ‪ hash‬ﻧﻴﺰ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﻛﻨﻴﻢ‪.‬‬

‫‪153‬‬
 ‫• ﺍﺳﺘﻔﺪﻩ ﺍﺯ ﻫﻤﺎﻥ ﺭﻭﺵ ﭘﺎﺭﺗﻴﺸﻦ ﺑﻨﺪﻱ ﺑﺮﺍﻱ ‪.r‬‬

‫‪:Handling of Overflows‬‬
‫ﺍﮔﺮﺑﺮﺧﻲ ﭘﺎﺭﺗﻴﺸﻦ ﻫﺎ ﺗﺎﭘﻞ ﻫﺎﻱ ﺑﻴﺸﺘﺮﻱ ﻧﺴﺒﺖ ﺑﻪ ﺑﺮﺧﻲ ﺩﻳﮕﺮ ﺩﺍﺷﺘﻪ ﺑﺎﺷﻨﺪ ﭘﺎﺭﺗﻴﺸﻦ‬
‫ﻧﺎﻣﺘﻮﺍﺯﻥ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪ Hash-table overflow‬ﺩﺭﭘﺎﺭﺗﻴﺸﻦ ‪ si‬ﺭﺥ ﻣﻲ ﺩﻫﺪ ﺍﮔﺮ ‪ si‬ﺩﺭﺣﺎﻓﻈﻪ ﻣﺘﻨﺎﺳﺐ ﻗﺮﺍﺭ‬

‫ﻧﮕﻴﺮﺩ ﻛﻪ ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﻪ ﺍﻳﻦ ﺩﻻﻳﻞ ﺑﺎﺷﺪ ‪.‬‬

‫ﺗﻌﺪﺍﺩﻱ ﺍﺯ ﺗﺎﭘﻞ ﻫﺎﻱ ‪ s‬ﺩﺭ ﻓﻴﻠﺪ ﭘﻴﻮﺳﺘﮕﻲ ﺩﺍﺭﺍﻱ ﺍﺭﺯﺵ ﻳﻜﺴﺎﻥ ﻫﺴﺘﻨﺪ ‪.‬‬

‫ﺗﺎﺑﻊ ﻫﺶ ﺑﺪ‪.‬‬

‫‪ Overflow resolution‬ﻣﻲ ﺗﻮﺍﻧﺪ ﺩﺭﻓﺎﺯ ﺳﺎﺧﺖ ﺍﺗﻔﺎﻕ ﺑﻴﺎﻓﺘﺪ )ﺳﺮﺭﻳﺰ(‪.‬‬

‫ﭘﺎﺭﺗﻴﺸﻦ ‪ si‬ﺑﻴﺸﺘﺮ ﺑﺎ ﺗﻮﺍﺑﻊ ﻫﺶ ﻣﺘﻔﺎﻭﺗﻲ ﺗﻘﺴﻴﻢ ﺑﻨﺪﻱ ﺷﺪﻩ‪.‬‬

‫ﭘﺎﺭﺗﻴﺸﻦ ‪ ri‬ﺑﺎﻳﺪ ﺗﻘﺴﻴﻢ ﺑﻨﺪﻱ ﻳﻜﺴﺎﻥ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ‪.‬‬

‫‪) Overflow avoidance‬ﺍﺟﺘﻨﺎﺏ ﺍﺯ ﺳﺮﺭﻳﺰ ( ﺍﻧﺠﺎﻡ ﺑﺎ ﺩﻗﺖ ﭘﺎﺭﺗﻴﺸﻦ ﺑﻨﺪﻱ ﺑﺮﺍﻱ‬

‫ﺟﻠﻮﮔﻴﺮﻱ ﺍﺯ ﺳﺮﺭﻳﺰ ﺩﺭ ﻃﻮﻝ ﻓﺎﺯ ﺳﺎﺧﺖ‪.‬‬

‫ﺑﻌﻨﻮﺍﻥ ﻣﺜﺎﻝ ﭘﺎﺭﺗﻴﺸﻦ ﺳﺎﺧﺖ ﺍﺭﺗﺒﺎﻁ ﺩﺍﺧﻠﻲ ﺑﺎ ﺑﺴﻴﺎﺭﻱ ﺍﺯ ﭘﺎﺭﺗﻴﺸﻦ ﻫﺎ ﻭ‬

‫ﺳﭙﺲ ﺗﺮﻛﻴﺐ ﺁﻧﻬﺎ‪.‬‬

‫ﺩﻭ ﺭﻭﻳﻜﺮﺩ ﺷﻜﺖ ﺑﺎﺗﻌﺪﺍﺩ ﺯﻳﺎﺩﻱ ﻣﻮﺍﺭﺩ ﺗﻜﺮﺍﺭﻱ‪.‬‬

‫‪ :Fallback option‬ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺣﻠﻘﻪ ﻫﺎﻱ ﭘﻴﻮﻧﺪﻱ ﺗﻮﺩﺭﺗﻮ ﺑﺮﺍﻱ ﭘﺎﺭﺗﻴﺸﻦ ﻫﺎﻱ‬

‫‪.overflow‬‬

‫‪154‬‬
 ‫ﻫﺰﻳﻨﻪ ﻫﺎﻱ ‪: Hash-Join‬‬

‫‪ ‬ﺍﮔﺮ ﭘﺎﺭﺗﻴﺸﻦ ﺑﻨﺪﻱ ﺑﺎﺯﮔﺸﺘﻲ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﻧﺒﺎﺷﺪ ﻫﺰﻳﻨﻪ ‪ hash join‬ﺧﻮﺍﻫﺪ ﺑﻮﺩ‪.‬‬

‫‪3(br + bs) +4 ∗ nh block transfers +‬‬

‫‪2( br / bb + bs / bb) seeks‬‬

‫‪ ‬ﺍﮔﺮ ﭘﺎﺭﺗﻴﺸﻦ ﺑﻨﺪﻱ ﺑﺎﺯﮔﺸﺘﻲ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﺑﺎﺷﺪ‪:‬‬

‫‪ s‬ﺑﺮﺍﻱ ﺣﺪﺍﻗﻞ ‪ M‬ﺑﻼﻙ ﺍﺯﻫﺮ ﭘﺎﺭﺗﻴﺸﻦ‬ ‫ﺗﻌﺪﺍﺩﭘﺎﺱ ﻫﺎﻱ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﺑﺮﺍﻱ ﭘﺎﺭﺗﻴﺸﻦ ﺳﺎﺧﺖ ﺍﺭﺗﺒﺎﻁ‬
‫ﺧﻮﺍﻫﺪ ﺑﻮﺩ‪:‬‬

‫‪logM/bb–1(bs/M)‬‬

‫‪ ‬ﺍﻧﺘﺨﺎﺏ ﺭﺍﺑﻄﻪ ﻛﻮﭼﻜﺘﺮ ﺑﻪ ﻋﻨﻮﺍﻥ ﺭﺍﺑﻄﻪ ﺳﺎﺧﺖ ﻭ ﺳﺎﺯ ﺑﻬﺘﺮﻳﻦ ﺍﻧﺘﺨﺎﺏ ﺍﺳﺖ‪.‬‬

‫‪ ‬ﺑﺮﺁﻭﺭﺩ ﻫﺰﻳﻨﻪ ﻧﻬﺎﻳﻲ‪:‬‬

‫‪2(br + bs) logM/bb–1(bs/M) + br + bs block transfers+‬‬
‫‪br / bb + bs / bb) logM/bb–1(bs/M)  seeks‬‬

‫‪ ‬ﺍﮔﺮﺗﻤﺎﻡ ﻭﺭﻭﺩﻱ ﻫﺎﻱ ﺳﺎﺧﺖ ﻭﺳﺎﺯ ﺑﺘﻮﺍﻧﺪ ﺩﺭ ﺣﺎﻓﻈﻪ ﺍﺻﻠﻲ ﻧﮕﻬﺪﺍﺭﻱ ﺷﻮﺩ ﻧﻴﺎﺯﻱ ﺑﻪ ﻫﻴﭻ‬
‫ﭘﺎﺭﺗﻴﺸﻦ ﺑﻨﺪﻱ ﻧﻴﺴﺖ‪.‬‬

‫‪ ‬ﻫﺰﻳﻨﻪ ﺑﺮﺍﻱ ‪ br + bs.‬ﻗﺎﺑﻞ ﻛﺎﻫﺶ ﺩﺍﺩﻥ ﺍﺳﺖ‪.‬‬

‫ﻣﺜﺎﻝ ﺑﺮﺍﻱ ﻣﺤﺎﺳﺒﻪ ﻫﺰﻳﻨﻪ ﻫﺎﻱ ‪Hash-Join‬‬

‫ﻓﺮﺽ ﻣﻲ ﻛﻨﻴﻢ ﺣﺎﻓﻈﻪ ﺍﻱ ﺑﺎ ﺳﺎﻳﺰ ‪ 20‬ﺑﻼﻙ ﺩﺍﺭﻳﻢ‪.‬‬

‫‪binstructor= 100 and bteaches = 400.‬‬

‫‪155‬‬
‫ﺳﺎﺯﻧﺪﻩ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ‪ ،build input‬ﭘﺎﺭﺗﻴﺸﻦ ﻫﺎﻱ ﺭﺍ ﺑﻪ ‪ 5‬ﻗﺴﻤﺖ ﺑﺎ ﺳﺎﻳﺰ ‪ 20‬ﺑﻼﻛﻲ‬
‫ﺗﻘﺴﻴﻢ ﻣﻴﻜﻨﺪ‪ ،‬ﻛﻪ ﭘﺎﺭﺗﻴﺸﻦ ﻫﺎ ﻣﻲ ﺗﻮﺍﻧﺪ ﺩﺭ ﻳﻚ ﮔﺬﺭﮔﺎﻩ ﻳﺎ ﭘﺎﺱ ﺍﺟﺮﺍ ﺷﻮﻧﺪ‪.‬‬

‫ﻛﻪ ﺍﻟﺒﺘﻪ ﻣﻴﺘﻮﺍﻥ ﭘﺎﺭﺗﻴﺸﻦ ﻫﺎ ﺭﺍ ﺑﺎ ﺳﺎﻳﺰ ﻫﺮ ﻳﻚ ‪ ، 80‬ﺩﺭ ﻳﻚ ﮔﺬﺭﮔﺎﻩ ﺍﺟﺮﺍ ﻧﻤﻮﺩ‪.‬‬

‫ﺑﻨﺎﺑﺮﺍﻳﻦ ﻫﺰﻳﻨﻪ ﻧﻬﺎﻳﻲ ﺑﺮﺍﺑﺮ ﺍﺳﺖ ﺑﺎ‪:‬‬

‫‪seeks 3(100 + 400) = 1500 block transfers +2( 100/3 + 400/3) = 336‬‬

‫‪Hybrid Hash–Join:‬‬

‫ﺑﻬﺘﺮﻳﻦ ﺍﺳﺘﻔﺎﺩﻩ ﺯﻣﺎﻧﻲ ﺍﺳﺖ ﻛﻪ ﺣﺎﻓﻈﻪ ﺑﻪ ﺍﻧﺪﺍﺯﻩ ﻛﺎﻓﻲ ﺑﺰﺭگ ﺍﺳﺖ ﻭ ﺳﺎﺧﺖ ﻭ ﺳﺎﺯ ﻭﺭﻭﺩﻱ ﻫﺎ‬
‫ﺑﺰﺭﮔﺘﺮ ﺍﺯ ﺣﺎﻓﻈﻪ ﺍﺳﺖ‪.‬‬

‫ﻭﻳﮋﮔﻲ ﺍﺻﻠﻲ ﺍﻳﻦ ﺍﻟﮕﻮﺭﻳﺘﻢ‪:‬‬

‫ﻧﮕﻬﺪﺍﺭﻱ ﺍﻭﻟﻴﻦ ﭘﺎﺭﺗﻴﺸﻦ ﺳﺎﺧﺖ ﺩﺭ ﺣﺎﻓﻈﻪ‬

‫‪ ‬ﺑﻌﻨﻮﺍﻥ ﻣﺜﺎﻝ‪ ،‬ﺣﺎﻓﻈﻪ ﺍﻱ ﺑﺎ ‪ 25‬ﺑﻼﻙ‪ ،‬ﻛﻪ ﺳﺎﺯﻧﺪﻩ ﻣﻴﺘﻮﺍﻧﺪ ﺁﻧﺮﺍ ﺑﻪ ‪ 5‬ﭘﺎﺭﺗﻴﺸﻦ ‪ 20‬ﺑﻼﻛﻲ‬
‫ﺗﻘﺴﻴﻢ ﺑﻨﺪﻱ ﻛﻨﺪ‪.‬‬

‫♦ ﺗﻘﺴﻴﻢ ﺑﻨﺪﻱ ﺣﺎﻓﻈﻪ‪:‬‬

‫‪ ‬ﺍﻭﻟﻴﻦ ﭘﺎﺭﺗﻴﺸﻦ ‪ 20‬ﺑﻠﻮﻙ ﺍﺯ ﺣﺎﻓﻈﻪ ﺭﺍ ﺍﺷﻐﺎﻝ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫‪ 4‬ﭘﺎﺭﺗﻴﺸﻦ ﺩﻳﮕﺮ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ‬ ‫‪ ‬ﻳﻚ ﺑﻠﻮﻙ ﺑﺮﺍﻱ ﻭﺭﻭﺩﻱ ﻭﻫﺮﺑﻠﻮﻙ ﺑﺮﺍﻱ ﺑﺎﻓﺮ‬
‫ﺷﻮﺩ ‪.‬‬

‫‪ ‬ﻧﻜﺘﻪ ﺁﻣﻮﺯﺵ ‪ :‬ﺍﻳﻨﻜﻪ ﭘﺎﺭﺗﻴﺸﻦ ﻫﺎ ﺑﻪ ﺻﻮﺭﺕ ﻳﻜﺴﺎﻥ ﺑﻪ ‪ 5‬ﭘﺎﺭﺗﻴﺸﻦ ﺑﺎ ﺳﺎﻳﺰ ‪ 80‬ﺗﻘﺴﻴﻢ ﺷﺪﻩ‬
‫ﺍﻧﺪ‪.‬‬

‫‪156‬‬
 ‫♦ ﺑﺮﺍﻱ ﺍﻭﻟﻴﻦ ﺑﺎﺭ ﺍﺯ ﺭﻭﺵ ﺟﺴﺘﺠﻮ ﺑﻪ ﺟﺎﻱ ﻧﻮﺷﺘﻦ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪ ‬ﻫﺰﻳﻨﻪ ‪ 3(80 + 320) + 20 +80 = 1300‬ﺑﻼﻙ ﺍﻧﺘﻘﺎﻟﻲ ‪،Hybrid hash-join‬‬

‫ﻣﻴﺘﻮﺍﻧﺪ ﺑﻪ ﺍﻧﺪﺍﺯﻩ ‪1500‬ﻫﺶ ﭘﻴﻮﻧﺪﻱ ﺑﺎﺷﺪ‪.‬‬

‫‪ Hybrid hash-join ‬ﺑﻴﺸﺘﺮ ﻣﻮﺭﺩ ﺍﺳﺘﻔﺎﺩﻩ ﻗﺮﺍﺭ ﻣﻲ ﮔﻴﺮﺩ ﺍﮔﺮ‪M >> :‬‬

‫ﻋﻤﻠﻴﺎﺕ ﺩﻳﮕﺮ‪:‬‬

‫‪ :Duplicate elimination ‬ﺟﻬﺖ ﭘﻴﺎﺩﻩ ﺳﺎﺯﻱ ﻫﺶ ﻫﺎ ﻭ ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﻫﺎﺳﺖ ‪.‬‬

‫ﺩﺭ ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﺩﻭﻧﺴﺨﻪ ﺍﻱ‪ ،‬ﻫﺮ ﻧﺴﺨﻪ ﺍﺯ ﻳﻚ ﻣﺠﻤﻮﻋﻪ ﻗﺎﺑﻞ ﺣﺬﻑ ﻣﻲ ﺑﺎﺷﺪ ‪.‬‬

‫ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ‪ :‬ﻧﺴﺨﻪ ﻫﺎ ﺣﻴﻦ ﺍﺟﺮﺍ ﻗﺎﺑﻞ ﺣﺬﻑ ﺷﺪﻥ ﻫﺴﺘﻨﺪ‪ ،‬ﻛﻪ ﺑﻪ ﺧﻮﺑﻲ ﻣﺮﺍﺣﻞ‬
‫ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﺩﺭ ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﺍﺩﻏﺎﻣﻲ ﺍﺳﺖ ‪.‬‬

‫‪ Hashing‬ﻣﺎﻧﻨﺪ ﻧﺴﺨﻪ ﻫﺎﻱ ﻣﺸﺎﺑﻪ ﺩﺭ ﻳﻚ ﺑﺎﻛﺖ ﻫﺴﺘﻨﺪ ‪.‬‬

‫‪) Aggregation. ‬ﻣﺠﺘﻤﻊ ﺑﻮﺩﻥ( ‪ :‬ﻣﻲ ﺗﻮﺍﻧﺪ ﺗﻮﺳﻂ ﻳﻚ ﺭﻭﺵ ﻣﺸﺎﺑﻪ ﺭﻭﺵ ﭘﻴﺎﺩﻩ ﺳﺎﺯﻱ‬
‫ﺷﻮﺩ‪.‬‬

‫ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﻭ ‪ Hashing‬ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﻣﻮﺭﺩ ﺍﺳﺘﻔﺎﺩﻩ ﺗﺎﭘﻞ ﻫﺎﻳﻲ ﻛﻪ ﺩﺭﻭﻥ ﻳﻚ ﮔﺮﻭﻩ‬
‫ﻣﺸﺎﺑﻪ ﻫﺴﺘﻨﺪ‪ ،‬ﺑﺎﺷﻨﺪ‪ ،‬ﻭ ﻣﺠﺘﻤﻊ ﺑﻮﺩﻥ ﺗﻮﺍﺑﻊ ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﺮﺍﻱ ﻫﺮ ﮔﺮﻭﻩ ﺩﺭ ﻧﻈﺮ ﮔﺮﻓﺘﻪ‬
‫ﺷﻮﺩ‪.‬‬

‫ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ‪ :‬ﺗﺮﻛﻴﺐ ﺗﺎﭘﻞ ﻫﺎ ﺩﺭ ﻳﻚ ﮔﺮﻭﻩ ﻣﺸﺎﺑﻪ‪ ،‬ﻃﻲ ﺍﺟﺮﺍﻱ ﻋﻤﻠﻴﺎﺕ ﺻﻮﺭﺕ ﻣﻲ‬
‫ﮔﻴﺮﺩ‪ ،‬ﻛﻪ ﺍﻳﻦ ﺑﻮﺳﻴﻠﻪ ﻣﻴﺰﺍﻥ ﺑﺮﺁﻭﺭﺩ ﺷﺪﻩ ﺍﺯ ﻣﺠﺘﻤﻊ ﺑﻮﺩﻥ ﺁﻧﻬﺎ ﺗﻌﻌﻴﻦ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪157‬‬
 ‫‪ ‬ﺑﺮﺍﻱ ﺷﻤﺎﺭﺵ‪ ،‬ﺣﺪﺍﻗﻞ‪ ،‬ﺣﺪﺍﻛﺜﺮ‪ ،‬ﻣﺠﻤﻮﻉ‪ :‬ﻣﻴﺰﺍﻥ ﻣﺠﺘﻤﻊ ﺑﻮﺩﻥ ﻧﮕﺪﺍﺭﻱ ﺷﺪﻩ‬
‫ﺑﺮﻭﻱ ﺗﺎﭘﻞ ﻫﺎﺳﺖ‪.‬‬

‫‪ ‬ﺑﺮﺍﻱ ﻣﻴﺎﻧﮕﻴﻦ‪ :‬ﻣﻘﺎﺩﻳﺮ ﺷﻤﺎﺭﺵ ﺷﺪﻩ ﺟﻤﻊ ﺷﺪﻩ ﻭ ﺑﺮ ﺗﻌﺪﺍﺩ ﺷﻤﺎﺭﺵ ﻫﺎ ﺗﻘﺴﻴﻢ‬
‫ﻣﻲ ﺷﻮﺩ‪.‬‬

‫ﺍﺭﺯﻳﺎﺑﻲ ﺁﺯﻣﺎﻳﺸﺎﺕ‪:‬‬

‫‪ ‬ﺍﻟﮕﻮﺭﻳﺘﻢ ﻫﺎﻱ ﻛﻪ ﺗﺎ ﺍﻛﻨﻮﻥ ﺩﻳﺪﻳﻢ‪ ،‬ﺑﺮﺍﻱ ﺍﻧﺠﺎﻡ ﻋﻤﻠﻴﺎﺕ ﻫﺎﻱ ﻓﺮﺩﻱ ﺍﺳﺖ‪.‬‬

‫‪ ‬ﺟﺎﻳﮕﺰﻳﻦ ﻫﺎﻳﻲ ﺑﺮﺍﻱ ﺍﺭﺯﻳﺎﺑﻲ ﻛﻞ ﻋﺒﺎﺭﺕ ﻳﻚ ﺩﺭﺧﺖ ‪:‬‬

‫‪ Materialization‬ﻳﺎ ﺗﺤﻘﻖ‪ :‬ﺗﻮﻟﻴﺪ ﻧﺘﺎﻳﺞ ﺍﺯ ﻋﺒﺎﺭﺗﻲ ﻛﻪ ﻭﺭﻭﺩﻱ ﺭﻭﺍﺑﻂ ﻭ ﻳﺎ ﺩﺭ ﺣﺎﻝ‬

‫ﻣﺤﺎﺳﺒﻪ‪ ،‬ﺍﺳﺖ ﺁﻥ ﺭﺍ ﺑﺮ ﺭﻭﻱ ﺩﻳﺴﻚ ﺫﺧﻴﺮﻩ ﻭ ﺗﻜﺮﺍﺭ ﻛﻨﻴﺪ ‪.‬‬

‫‪ Pipelining‬ﻳﺎ ﻟﻮﻟﻪ ﮔﺬﺍﺭﻱ ‪ :‬ﺭﻭﻱ ﺗﺎﭘﻞ ﻫﺎﻱ ﻋﻤﻠﻴﺎﺕ ﭘﺪﺭ ﺣﺘﻲ ﺑﻌﻨﻮﺍﻥ ﻳﻚ‬
‫ﻋﻤﻠﻴﺎﺕ ﺩﺭ ﺣﺎﻝ ﺍﺟﺮﺍ ﮔﺬﺭ ﺩﻫﻴﺪ ‪.‬‬

‫‪ ‬ﻣﺎ ﺩﺭﺑﺎﺭﻩ ﺁﻟﺘﺮﻧﺎﺗﻴﻮ ﻫﺎﻱ ﺑﺎﻻ ﺑﺎ ﺟﺰﻳﻴﺎﺕ ﺑﻴﺸﺘﺮ ﺩﺭ ﺣﺎﻝ ﻣﻄﺎﻟﻌﻪ ﻫﺴﺘﻴﻢ ‪.‬‬

‫‪:Materialization‬‬

‫ﺍﺭﺯﻳﺎﺑﻲ ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ ‪ :‬ﺍﺭﺯﻳﺎﺑﻲ ﻳﻚ ﻋﻤﻠﻴﺎﺕ ﺩﺭ ﻳﻚ ﺯﻣﺎﻥ‪ ،‬ﺩﺭ ﭘﺎﻳﻴﻦ ﺗﺮﻳﻦ ﺳﻄﺢ ﺷﺮﻭﻉ ﻣﻲ ﺷﻮﺩ ‪.‬‬
‫ﻧﺘﺎﻳﺞ ﻣﻴﺎﻧﻲ ﻣﺤﻘﻖ ﺷﺪﻩ ﺭﺍ ﺑﺮﺍﻱ ﺭﻭﺍﺑﻂ ﻣﻮﻗﺖ ﺗﺎ ﺍﺭﺯﻳﺎﺑﻲ ﻋﻤﻠﻴﺎﺕ ﺳﻄﺢ ﺑﻌﺪﻱ ﺍﺳﺘﻔﺎﺩﻩ ﻛﻨﻴﺪ ‪.‬‬

‫ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ‪ ،‬ﺩﺭ ﺷﻜﻞ ﺯﻳﺮ‪ ،‬ﻣﺤﺎﺳﺒﻪ ﻭ ﺫﺧﻴﺮﻩ ﻛﻨﻴﺪ‪:‬‬

‫) ‪σ building ="Watson" (department‬‬

‫‪158‬‬
 ‫ﺳﭙﺲ ﺑﺎ ﺩﺳﺘﻮﺭﺍﺕ ﺭﺍﻫﻨﻤﺎ ﻋﻤﻠﻴﺎﺕ ﺍﻟﺤﺎﻕ ﺭﺍ ﻣﺤﺎﺳﺒﻪ ﻭ ﺫﺧﻴﺮﻩ ﻛﻨﻴﺪ ﻭ ﺩﺭ ﻧﻬﺎﻳﺖ ﻋﻤﻠﻴﺎﺕ‬
‫ﭘﺮﺗﻮ)ﭘﺮﻭﺟﻜﺸﻦ( ﺭﺍ ﺭﻭﻱ ﻛﻠﻤﻪ ‪name‬ﻣﺤﺎﺳﺒﻪ ﻛﻨﻴﺪ‪.‬‬

‫‪ n‬ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ ﻫﻤﻴﺸﻪ ﻗﺎﺑﻞ ﺍﺟﺮﺍﺳﺖ‬

‫‪ n‬ﻫﺰﻳﻨﻪ ﻧﻮﺷﺘﻦ ﻧﺘﺎﻳﺞ ﺭﻭﻱ ﺩﻳﺴﻚ ﻭ ﺧﻮﺍﻧﺪﻥ ﻣﺠﺪﺩ ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﺴﻴﺎﺭ ﺑﺎﻻ ﺑﺎﺷﺪ‬

‫ﻓﺮﻣﻮﻝ ﻫﺰﻳﻨﻪ ﻫﺎﻱ ﻣﺎ ﺑﺮﺍﻱ ﺍﻧﺠﺎﻡ ﻋﻢﻟﻴﺎﺕ ﺑﺎ ﭼﺸﻢ ﭘﻮﺷﻲ ﺍﺯ ﻫﺰﻳﻨﻪ ﻧﻮﺷﺘﻦ ﻧﺘﺎﻳﺞ ﺭﻭﻱ ﺩﻳﺴﻚ ‪:‬‬

‫‪ ‬ﻫﺰﻳﻨﻪ ﻛﻞ = ﻣﺠﻤﻮﻉ ﻫﺰﻳﻨﻪ ﻫﺎﻱ ﻋﻤﻠﻴﺎﺗﻬﺎﻱ ﻓﺮﺩﻱ ‪ +‬ﻫﺰﻳﻨﻪ ﻧﻮﺷﺘﻦ ﻧﺘﺎﻳﺞ‬

‫ﻣﻴﺎﻧﻲ ﺑﺮﻭﻱ ﺩﻳﺴﻚ‬

‫‪ n‬ﺩﻭ ﺑﺎﻓﺮ‪ :‬ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺩﻭ ﺑﺎﻓﺮ ﺧﺮﻭﺟﻲ ﺑﺮﺍﻱ ﻫﺮ ﻳﻚ ﺍﺯ ﻋﻤﻠﻴﺎﺕ‪ ،‬ﻭﻗﺘﻲ ﻛﻪ ﻳﻜﻲ ﭘﺮ ﺍﺳﺖ ﻭ ﻧﻮﺷﺘﻦ‬
‫ﺁﻥ ﺑﺮ ﺭﻭﻱ ﺩﻳﺴﻚ ﺩﺭ ﺣﺎﻟﻲ ﻛﻪ ﺑﺎﻓﺮ ﺩﻳﮕﺮ ﺍ ﺩﺭ ﺣﺎﻝ ﭘﺮ ﺷﺪﻥ ﺍﺳﺖ‬

‫‪ l‬ﺍﺟﺎﺯﻩ ﻣﻲ ﺩﻫﺪ ﺗﺎ ﻫﻤﭙﻮﺷﺎﻧﻲ ﺑﻴﻦ ﻧﻮﺷﺘﻦ ﻭ ﻣﺤﺎﺳﺒﻪ ﺍﻳﺠﺎﺩ ﺷﻮﺩ ﻭ ﺯﻣﺎﻥ ﺍﺟﺮﺍ ﺭﺍ‬
‫ﻛﺎﻫﺶ ﻣﻲ ﺩﻫﺪ‪.‬‬

‫‪:Pipelining‬‬

‫‪ :Pipelined evaluation‬ﺍﺭﺯﻳﺎﺑﻲ ﭼﻨﺪﻳﻦ ﻋﻤﻠﻴﺎﺕ ﺑﻪ ﻃﻮﺭ ﻫﻤﺰﻣﺎﻥ‪ ،‬ﻭ ﺍﻧﺘﻘﺎﻝ ﻧﺘﺎﻳﺞ ﺣﺎﺻﻞ ﺍﺯ‬
‫ﻳﻚ ﻋﻤﻠﻴﺎﺕ ﺑﻪ ﭘﺎﺱ ﺑﻌﺪﻱ‪.‬‬

‫ﺑﻌﻨﻮﺍﻥ ﻣﺜﺎﻝ‪ ،‬ﺩﺭ ﺩﺭﺧﺖ ﻋﺒﺎﺭﺕ ﻗﺒﻠﻲ‪ ،‬ﻧﺘﻴﺠﻪ ﺍﻱ ﺫﺧﻴﺮﻩ ﻧﺸﺪﻩ‪،‬‬

‫‪159‬‬
 ‫ﺑﺠﺎﻱ ﭘﻴﻮﻧﺪ ﻣﺴﺘﻘﻴﻢ ﺗﺎﭘﻞ ﻫﺎ ﻛﻪ ﺑﻄﻮﺭ ﻣﺸﺎﺑﻪ ﻧﺘﻴﺠﻪ ﭘﺎﺱ ﺗﺎﭘﻞ ﻫﺎ ﺭﻭﻱ ﻃﺮﺡ ﺫﺧﻴﺮﻩ ﻧﻤﻲ‬
‫ﻛﻨﺪ‪.‬‬

‫‪ ‬ﺗﺤﻘﻴﻘﺎﺕ ﺍﺭﺯﺍﻥ ﺗﺮ ﻧﺸﺎﻥ ﻣﻲ ﺩﻫﺪ ﻛﻪ ‪ :‬ﻧﻴﺎﺯﻱ ﺑﻪ ﺫﺧﻴﺮﻩ ﺍﺭﺗﺒﺎﻁ ﻣﻮﻗﺖ ﺭﻭﻱ ﺩﻳﺴﻚ ﻧﻴﺴﺖ‬

‫‪ ‬ﺍﻣﻜﺎﻥ ‪ Pipelining‬ﻫﺴﺖ ﺍﻣﺎ ﻧﻪ ﻫﻤﻴﺸﻪ ﺑﻌﻨﻮﺍﻥ ﻣﺜﺎﻝ ﻣﺮﺗﺐ ﺳﺎﺯﻱ ‪hash-join.‬‬

‫‪ ‬ﺑﺮﺍﻱ ‪ pipelining‬ﻣﻮﺛﺮ ﺍﺳﺖ ﺍﺳﻔﺎﺩﻩ ﺍﺯ ﺍﻟﮕﻮﺭﻳﺘﻢ ﻫﺎﻱ ﺍﺭﺯﻳﺎﺑﻲ ﻛﻪ ﻫﺮ ﺗﺎﭘﻞ ﺧﺮﻭﺟﻲ ﺑﻪ‬
‫ﻋﻨﻮﺍﻥ ﺗﺎﭘﻞ ﻭﺭﻭﺩﻱ ﻋﻤﻠﻴﺎﺕ ﺩﺭﻳﺎﻓﺖ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫‪ Pipelines ‬ﺑﻪ ﺩﻭ ﺭﻭﺵ ﻗﺎﺑﻞ ﺍﺟﺮﺍﺳﺖ ‪ :‬ﻫﺪﺍﻳﺖ ﺗﻘﺎﺿﺎ ) ‪ (demand driven‬ﻭ‬

‫ﻫﺪﺍﻳﺖ ﺗﻮﻟﻴﺪ) ‪(producer driven‬‬

‫‪ ‬ﻫﺪﺍﻳﺖ ﺗﻘﺎﺿﺎ ﻳﺎ ﺍﺭﺯﻳﺎﺑﻲ ‪lazy‬‬

‫ﺳﻴﺴﺘﻢ ﺩﺭﺧﻮﺍﺳﺖ ﺗﺎﭘﻞ ﺑﻌﺪﻱ ﺭﺍ ﺍﺯﻋﻤﻠﻴﺎﺕ ﺳﻄﺢ ﺑﺎﻻﺗﺮ ﺗﻜﺮﺍﺭ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫)ﻋﻤﻠﻴﺎﺕ ﻓﺮﺯﻧﺪ ( ﺑﺮﺍﻱ‬ ‫ﻫﺮﻳﻚ ﺍﺯ ﺩﺭﺧﻮﺍﺳﺖ ﻫﺎﻱ ﺗﺎﭘﻞ ﺑﻌﺪﻱ ﺍﺯ ﺯﻳﺮ ﻋﻤﻠﻴﺎﺕ ﻫﺎ‬
‫ﺧﺮﻭﺟﻲ ﺗﺎﭘﻞ ﺑﻌﺪﺏ ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﺍﺳﺖ‪.‬‬

‫ﺩﺭﺑﻴﻦ ﻓﺮﺍﺧﻮﺍﻧﻲ ﻫﺎ )ﺗﻤﺎﺱ ﻫﺎ ‪(calls‬ﻋﻤﻠﻴﺎﺕ ﺿﻤﻦ ﺣﻔﻆ ‪ state‬ﺷﻨﺎﺳﺎﻳﻲ ﻣﻲ‬

‫ﻛﻨﺪ ﭼﻪ ﭼﻴﺰﻱ ﺭﺍ ﺑﻪ ﻋﻤﻠﻴﺎﺕ ﺑﻌﺪﻱ ﺑﺮﮔﺮﺩﺍﻧﺪ‬

‫‪ ‬ﺩﺭ ‪producer-driven or eager pipelining‬‬

‫ﺍﭘﺮﺍﺗﻮﺭ ﻫﺎ ﺑﺎ ﺍﺷﺘﻴﺎﻕ ﺗﺎﭘﻞ ﻫﺎ ﺭﺍ ﺗﻮﻟﻴﺪ ﻣﻲ ﻛﻨﻨﺪ ﻭ ﺍﻧﻬﺎ ﺭﺍ ﺑﻪ ﻭﺍﻟﺪﻫﺎﻱ ﺧﻮﺩ ﺍﻧﺘﻘﺎﻝ ﻣﻲ‬
‫ﺩﻫﻨﺪ ‪.‬‬

‫‪ n‬ﺩﺭﺍﻳﻦ ﺑﻴﻦ ﺍﭘﺮﺍﺗﻮﺭﻫﺎ ﺑﺎﻓﺮﻫﺎ ﺭﺍ ﻛﻪ ﺑﻌﻨﻮﺍﻥ ﻓﺮﺯﻧﺪ ﺩﺭ ﻧﻈﺮ ﮔﺮﻓﺘﻪ ﺍﻧﺪ ﻧﮕﻬﺪﺍﺭﻱ‬
‫ﻣﻲ ﻛﻨﻨﺪ ‪.‬‬

‫‪160‬‬
 ‫‪ n‬ﺍﮔﺮﺑﺎﻓﺮﭘﺮﺑﺎﺷﺪ ‪ ،‬ﻓﺮﺯﻧﺪ ﻣﻨﺘﻈﺮﺧﺎﻟﻲ ﺷﺪﻥ ﺑﺎﻓﺮ ﻣﻲ ﺷﻮﺩ ﻭﭘﺲ ﺍﺯ ﺁﻥ ﺗﺎﭘﻞ‬
‫ﺑﻴﺸﺘﺮﻱ ﺗﻮﻟﻴﺪ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫ﻋﻤﻠﻴﺎﺕ ﺯﻣﺎﻧﺒﻨﺪﻱ ﺳﻴﺴﺘﻢ ﻓﻀﺎ ﺩﺭ ﺑﺎﻓﺮ ﺧﺮﻭﺟﻲ ﻭ ﭘﺮﺩﺍﺯﺵ ﺗﺎﭘﻞ ﻫﺎﻱ ﺑﻴﺸﺘﺮ ﻭﺭﻭﺩﻱ‬
‫ﺭﺍ ﻣﻲ ﺗﻮﺍﻧﺪ ﻣﺪﻳﺮﻳﺖ ﻛﻨﺪ‬

‫‪ ‬ﻣﺪﻝ ﻫﺎﻱ ﺍﺯ ﺍﻳﻦ ﺭﻭﺵ ‪ pull :‬ﻭ ‪push‬‬

‫‪161‬‬
 ‫ﻓﺼﻞ ‪:8‬‬
‫ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﭘﺮﺱ ﻭ ﺟﻮ‬

‫‪162‬‬
 ‫ﻣﻘﺪﻣﻪ‬ ‫•‬
‫ﺗﻐﻴﻴﺮ ﺷﻜﻞ ﺑﻴﺎﻥ ﺭﺍﺑﻄﻪ ﺍﻱ‬ ‫•‬
‫ﻛﺎﺗﺎﻟﻮگ ﺍﻃﻼﻋﺎﺕ ﺑﺮﺍﻱ ﺗﺨﻤﻴﻦ ﻫﺰﻳﻨﻪ‬ ‫•‬
‫ﺍﻃﻼﻋﺎﺕ ﺁﻣﺎﺭﻱ ﺑﺮﺍﻱ ﺗﺨﻤﻴﻦ ﻫﺰﻳﻨﻪ‬ ‫•‬
‫ﺑﻬﺎﻱ ﭘﺎﻳﻪ ﻱ ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ‬ ‫•‬
‫ﺑﺮﻧﺎﻣﻪ ﻧﻮﻳﺴﻲ ﭘﻮﻳﺎ ﺑﺮﺍﻱ ﺍﻧﺘﺨﺎﺏ ﻃﺮﺡ ﻫﺎﻱ ﺍﺭﺯﻳﺎﺑﻲ‬ ‫•‬
‫ﭼﺸﻢ ﺍﻧﺪﺍﺯﻫﺎﻱ ﻋﻤﻠﻲ‬ ‫•‬

‫‪163‬‬
 ‫ﻣﻘﺪﻣﻪ‬

‫ﺭﺍﻩ ﻫﺎﻱ ﻣﺘﻨﺎﻭﺏ ﺑﺮﺍﻱ ﺍﺭﺯﻳﺎﺑﻲ ﺟﺴﺘﺠﻮﻱ ﻣﻔﺮﻭﺽ ‪:‬‬

‫• ﺑﻴﺎﻥ ﻫﺎﻱ ﻣﺸﺎﺑﻪ‬
‫• ﺍﻟﮕﻮﺭﻳﺘﻢ ﻫﺎﻱ ﻣﺘﻔﺎﻭﺕ ﺑﺮﺍﻱ ﻫﺮ ﻋﻤﻠﻜﺮﺩ‬

‫ﻳﻚ ﻃﺮﺡ ﺍﺭﺯﻳﺎﺑﻲ ﺑﻪ ﻃﻮﺭ ﺩﻗﻴﻖ ﺗﻌﺮﻳﻒ ﻣﻲ ﻛﻨﺪ ﻛﻪ ﺑﺮﺍﻱ ﻫﺮ ﻋﻤﻠﻜﺮﺩ ﭼﻪ ﺍﻟﮕﻮﺭﻳﺘﻤﻲ ﺍﺳﺘﻔﺎﺩﻩ ﺷﺪﻩ ﻭ‬
‫ﻫﻤﭽﻨﻴﻦ ﭼﮕﻮﻧﮕﻲ ﺍﺟﺮﺍﻱ ﻫﻤﺎﻫﻨﮓ ﻋﻤﻠﻜﺮﺩ ﻫﺎ ﺭﺍ ﺑﻴﺎﻥ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫ﻫﻤﭽﻨﻴﻦ ﺑﺎﻳﺪ ﺩﺍﻧﺴﺖ ﻛﻪ ﭼﮕﻮﻧﻪ ﻃﺮﺡ ﻫﺎﻱ ﺍﺟﺮﺍﻱ ﺟﺴﺘﺠﻮ ﺭﺍ ﺑﺮ ﺭﻭﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻱ ﻣﻮﺭﺩ ﻧﻈﺮ ﺧﻮﺩ‪،‬‬
‫ﻣﻼﺣﻈﻪ ﻛﺮﺩ‪.‬‬

‫‪164‬‬
 ‫ﻫﺰﻳﻨﻪ ﻫﺎﻱ ﻣﺨﺘﻠﻒ ﺑﻴﻦ ﻃﺮﺡ ﻫﺎﻱ ﺍ ﺭﺯﻳﺎﺑﻲ ﺑﺮﺍﻱ ﻳﻚ ﺟﺴﺘﺠﻮ )‪ (query‬ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﻫﻨﮕﻔﺖ ﺷﻮﻧﺪ ‪.‬‬
‫ﮔﺎﻡ ﻫﺎ ﺩﺭ ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﺟﺴﺘﺠﻮﻱ ﻣﺒﺘﻨﻲ ﺑﺮ ﻫﺰﻳﻨﻪ‬ ‫‪‬‬
‫‪ .1‬ﺗﻮﻟﻴﺪ ﺍﺟﺮﺍﻫﺎﻱ ﻣﺸﺎﺑﻪ ﻣﻨﻄﻘﻲ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻗﻮﺍﻧﻴﻦ ﺗﺸﺎﺑﻪ‬
‫‪ .2‬ﻳﺎﺩﺩﺍﺷﺖ ﻛﺮﺩﻥ ﻧﺘﻴﺠﻪ ﻱ ﺍﺟﺮﺍﻫﺎ ﺑﺮﺍﻱ ﺑﺪﺳﺖ ﺁﻭﺭﺩﻥ ﻃﺮﺡ ﻫﺎﻱ ﺟﺴﺘﺠﻮﻱ ﻣﺘﻨﺎﻭﺏ‬
‫‪ .3‬ﺍﻧﺘﺨﺎﺏ ﺍﺭﺯﺍﻧﺘﺮﻳﻦ ﻃﺮﺡ ﺑﺮ ﭘﺎﻳﻪ ﻱ ﻫﺰﻳﻨﻪ ﻱ ﺗﺨﻤﻴﻨﻲ‬
‫ﺗﺨﻤﻴﻦ ﻃﺮﺡ ﻣﺒﺘﻨﻲ ﺑﺮ ﻫﺰﻳﻨﻪ ﺷﺎﻣﻞ ‪:‬‬ ‫‪‬‬
‫• ﺍﻃﻼﻋﺎﺕ ﺁﻣﺎﺭﻱ ﺩﺭﺑﺎﺭﻩ ﻱ ﺭﺍﺑﻄﻪ ﻫﺎ ﺳﺖ‪ ،‬ﻣﺜﺎﻝ ‪ :‬ﺗﻌﺪﺍﺩ ﭼﻨﺪﺗﺎﻳﻲ ﻫﺎ‪ ،‬ﺗﻌﺪﺍﺩ ﻣﻘﺎﺩﻳﺮ ﺑﺮﺗﺮ ﺑﺮﺍﻱ ﻫﺮ‬
‫ﺧﺎﺻﻴﺖ‬
‫• ﺗﺨﻤﻴﻦ ﺍﻣﺎﺭﻱ ﺑﺮﺍﻱ ﻧﺘﺎﻳﺞ ﻣﺘﻮﺳﻂ ﺑﺮﺍﻱ ﻣﺤﺎﺳﺒﻪ ﻱ ﻫﺰﻳﻨﻪ ﻱ ﺍﺟﺮﺍ ﻫﺎﻱ ﭘﻴﭽﻴﺪﻩ‬
‫• ﻣﺤﺎﺳﺒﻪ ﻱ ﻫﺰﻳﻨﻪ ﻱ ﻓﺮﻣﻮﻟﻪ ﻛﺮﺩﻥ ﺍﻟﮕﻮﺭﻳﺘﻢ ﻫﺎ‪ ،‬ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺁﻣﺎﺭ‬

‫ﺗﻮﻟﻴﺪ ﺍﺟﺮﺍﻫﺎﻱ ﺗﺸﺎﺑﻪ‬

‫ﺗﻐﻴﻴﺮ ﺷﻜﻞ ﺍﺟﺮﺍﻫﺎﻱ ﺭﺍﺑﻄﻪ ﺍﻱ‪:‬‬

‫ﺩﻭ ﺍﺟﺮﺍﻱ ﺟﺒﺮ ﺭﺍﺑﻄﻪ ﺍﻱ ﺭﺍ ﻣﺸﺎﺑﻪ ﻣﻲ ﺗﻮﺍﻥ ﮔﻔﺖ ﺍﮔﺮ ﺩﻭ ﺍﺟﺮﺍ ﭼﻨﺪﺗﺎﻳﻲ ﻫﺎﻱ ﻣﺴﺎﻭﻱ ﺑﺮ ﺭﻭﻱ ﭘﺎﻳﮕﺎﻩ‬
‫ﺻﻮﺭﺕ ﻣﺠﻤﻮﻋﻪ ﻫﺎﻱ ﭼﻨﺪ ﺗﺎﻳﻲ‬ ‫ﺩﺍﺩﻩ ﻱ ﻗﺎﻧﻮﻧﻲ ﺗﻮﻟﻴﺪ ﻛﻨﻨﺪ ‪.‬ﺩﺭ‪ ،SQL‬ﻭﺭﻭﺩﻱ ﻫﺎ ﻭ ﺧﺮﻭﺟﻲ ﻫﺎ ﺑﻪ‬
‫ﻫﺴﺘﻨﺪ‪.‬ﻗﻮﺍﻧﻴﻦ ﺗﺸﺎﺑﻪ ﺑﻴﺎﻥ ﻣﻲ ﻛﻨﺪ ﻛﻪ ﺍﺟﺮﺍﻫﺎﻱ ﺩﻭ ﻓﺮﻡ ﺑﺎ ﻫﻢ ﺑﺮﺍﺑﺮﻧﺪ ﻭ ﻣﻲ ﺗﻮﺍﻥ ﺍﺟﺮﺍﻱ ﻓﺮﻡ ﺍﻭﻟﻲ ﺭﺍ‬
‫ﺑﺎ ﺩﻭﻣﻲ ﺟﺎﺑﺠﺎ ﻛﺮﺩ‪.‬‬
‫ﻗﻮﺍﻧﻴﻦ ﺗﺸﺎﺑﻪ‪:‬‬

‫‪ .1‬ﺍﺭﺗﺒﺎﻁ ﺑﻴﻦ ﻋﻤﻠﻜﺮﺩﻫﺎﻱ ﺍﻧﺘﺨﺎﺏ ) ‪ ( selection‬ﺭﺍ ﻣﻲ ﺗﻮﺍﻥ ﺍﺯ ﺑﻴﻦ ﺑﺮﺩ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ‬
‫‪ selection‬ﻫﺎﻱ ﻣﺴﺘﻘﻞ ﻣﺘﻮﺍﻟﻲ‪.‬‬

‫)) ‪σ θ ∧θ ( E ) = σ θ (σ θ ( E‬‬
‫‪1‬‬ ‫‪2‬‬ ‫‪1‬‬ ‫‪2‬‬

‫‪ .2‬ﻋﻤﻠﻜﺮﺩﻫﺎﻱ ‪ selection‬ﺭﺍ ﻣﻲ ﺗﻮﺍﻥ ﺑﺎ ﻫﻢ ﻣﺒﺎﺩﻟﻪ ﻛﺮﺩ )ﺟﺎﺑﺠﺎ ﻛﺮﺩ(‪.‬‬

‫‪165‬‬
‫)) ‪σ θ (σ θ ( E )) = σ θ (σ θ ( E‬‬
‫‪1‬‬ ‫‪2‬‬ ‫‪2‬‬ ‫‪1‬‬

‫‪ .3‬ﻓﻘﻂ ﺁﺧﺮﻳﻦ ﻋﻤﻠﻜﺮﺩ ﺩﺭ ﺗﻮﺍﻟﻲ ‪ ،‬ﻣﻮﺭﺩ ﻧﻴﺎﺯ ﺍﺳﺖ ﻭ ﺑﻘﻴﻪ ﺭﺍ ﻣﻲ ﺗﻮﺍﻥ ﺣﺬﻑ ﻛﺮﺩ ‪.‬‬

‫) ‪Π L1 (Π L2 ( (Π Ln ( E )))) = Π L1 ( E‬‬

‫‪ Selection .4‬ﻫﺎ ﺭﺍ ﻣﻲ ﺗﻮﺍﻥ ﺑﺎ ﻫﻢ ﺗﺮﻛﻴﺐ ﻛﺮﺩ ﺑﺎ ﭘﻴﻮﺳﺘﻦ ﺗﺘﺎ ‪.‬‬

‫‪ .5‬ﻋﻤﻠﻜﺮﺩﻫﺎﻱ ﺗﺮﻛﻴﺐ ﺗﺘﺎ ﻗﺎﺑﻞ ﻣﺒﺎﺩﻟﻪ )ﺟﺎﺑﺠﺎ ﻛﺮﺩﻥ( ﻫﺴﺘﻨﺪ‪.‬‬

‫‪ (a) .6‬ﺗﺮﻛﻴﺐ ﻃﺒﻴﻌﻲ ﻋﻤﻠﻜﺮﺩﻫﺎ ﺷﺮﻛﺖ ﭘﺬﻳﺮ ﻫﺴﺘﻨﺪ ‪.‬‬

‫)‪ (b‬ﺗﺮﻛﻴﺐ ﺗﺘﺎ ﺑﻪ ﺷﺮﺡ ﺯﻳﺮ ﺷﺮﻛﺖ ﭘﺬﻳﺮ ﺍﺳﺖ‬

‫‪ θ2‬ﺷﺎﻣﻞ ﺧﺎﺻﻴﺖ ﻫﺎﻱ ‪ E2‬ﻭ ‪ E3‬ﺍﺳﺖ‬

‫ﺗﺼﻮﻳﺮﻱ ﺍﺯ ﻧﻤﺎﻳﺶ ﻗﻮﺍﻧﻴﻦ ﺗﺸﺎﺑﻪ‬

‫‪166‬‬
 ‫‪ .7‬ﻋﻤﻠﻜﺮﺩ ﺍﻧﺘﺨﺎﺏ ) ‪ ( selection‬ﺗﻘﺴﻴﻢ ﻣﻲ ﺷﻮﺩ ﺑﻪ ﺑﻴﺶ ﺍﺯ ﻋﻤﻠﻜﺮﺩ ﺗﺮﻛﻴﺐ ﺗﺘﺎ‪ ،‬ﺩﺭ ﺯﻳﺮ ﺩﻭ‬
‫ﻣﻮﺭﺩ ﺍﺯ ﺷﺮﺍﻳﻂ ﺫﻛﺮ ﺷﺪﻩ‪:‬‬
‫‪ (a‬ﺯﻣﺎﻧﻲ ﻛﻪ ﺗﻤﺎﻣﻲ ﺧﺼﻮﺻﻴﺎﺕ ﺩﺭ ‪ θ0‬ﺁﻣﺪﻩ ‪،‬ﺗﻨﻬﺎ ﺧﺼﻮﺻﻴﺎﺕ ﻳﻜﻲ ﺍﺯ ﺍﺟﺮﺍﻫﺎ )‪ (E1‬ﻣﻲ‬
‫ﭘﻴﻮﻧﺪﺩ‪.‬‬

‫‪ (b‬ﺯﻣﺎﻧﻴﻜﻪ ‪ θ 1‬ﺗﻤﺎﻡ ﺧﺼﻮﺻﻴﺎﺕ ‪ E1‬ﺭﺍ ﻭ ‪ θ2‬ﺗﻤﺎﻡ ﺧﺼﻮﺻﻴﺎﺕ ‪ E2‬ﺭﺍ ﺷﺎﻣﻞ ﻣﻲ ﺷﻮﺩ‪:‬‬

‫‪ .8‬ﻃﺮﺡ ﺗﻘﺴﻴﻢ ﻋﻤﻠﻜﺮﺩ ﻫﺎﻱ ﻓﺮﺍﺗﺮ ﺍﺯ ﺗﺘﺎ ﺑﻪ ﺻﻮﺭﺕ ﺯﻳﺮ ﻣﺘﺼﻞ ﻣﻲ ﺷﻮﻧﺪ ‪:‬‬
‫‪ (a‬ﺍﮔﺮ ﺗﺘﺎ ﺗﻨﻬﺎ ﺍﺯ ‪ L1 ∪ L2‬ﭘﺪﻳﺪ ﺁﻳﺪ‪:‬‬

‫‪∏L ∪ L ( E1‬‬ ‫‪θ‬‬ ‫‪E2 ) =( ∏L ( E)1‬‬ ‫‪θ‬‬ ‫) ‪( ∏L ( E2‬‬

‫‪1‬‬ ‫‪2‬‬ ‫‪1‬‬ ‫‪2‬‬

‫ﺑﺎﺷﺪ ﺁﻧﮕﺎﻩ‪:‬‬ ‫‪θ‬‬ ‫‪ (b‬ﺍﮔﺮ ﺍﺯ ﭘﻴﻮﺳﺘﻦ ‪E2 E1‬‬

‫• ﺑﺎﻳﺪ ﺍﺟﺎﺯﻩ ﺩﺍﺩ ‪ L1‬ﻭ ‪ L2‬ﺑﻪ ﺗﺮﺗﻴﺐ ﺑﻪ ﺻﻮﺭﺕ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺧﺼﻮﺻﻴﺎﺕ ‪ E1‬ﻭ ‪ E2‬ﺑﺎﺷﻨﺪ‪.‬‬
‫• ﺑﺎﻳﺪ ﺍﺟﺎﺯﻩ ﺩﺍﺩ ﺗﺎ ‪ L3‬ﺧﺼﻮﺻﻴﺘﻲ ﺍﺯ ‪ E1‬ﺑﺎﺷﺪ ﻛﻪ ﺩﺭ ﺷﺮﺍﻳﻂ ﺗﺘﺎ ﺑﭙﻴﻮﻧﺪﺩ ﺍﻣﺎ ﺩﺭ ‪L1 ∪ L2‬‬
‫ﻧﺒﺎﺷﺪ‪.‬‬

‫‪167‬‬
‫• ﺑﺎﻳﺪ ﺍﺟﺎﺯﻩ ﺩﺍﺩ ﺗﺎ ‪ L4‬ﺧﺼﻮﺻﻴﺘﻲ ﺍﺯ ‪ E2‬ﺑﺎﺷﺪ ﻛﻪ ﺩﺭ ﺷﺮﺍﻳﻂ ﺗﺘﺎ ﺑﭙﻴﻮﻧﺪﺩ ﺍﻣﺎ ﺩﺭ ‪L1 ∪ L2‬‬
‫ﻧﺒﺎﺷﺪ‪.‬‬

‫‪ .9‬ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﻋﻤﻠﻜﺮﺩ ﻫﺎﻱ ﺍﺗﺼﺎﻝ ﻭ ﺗﻘﺎﻃﻌﻲ ﺭﺍ ﻣﻲ ﺗﻮﺍﻥ ﺟﺎﺑﺠﺎ ﻛﺮﺩ ‪.‬‬

‫‪E1 ∪ E2 = E 2 ∪ E1‬‬
‫‪E1 ∩ E2 = E 2 ∩ E1‬‬
‫ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺍﺗﺼﺎﻝ ﻫﺎ ﻭ ﺗﻘﺎﻃﻊ ﻫﺎ ﺷﺮﻛﺖ ﭘﺬﻳﺮﻧﺪ ‪.‬‬ ‫‪.10‬‬
‫)‪(E1 ∪ E2) ∪ E3 = E1 ∪ (E2 ∪ E3‬‬
‫)‪(E1 ∩ E2) ∩ E3 = E1 ∩ (E2 ∩ E3‬‬
‫ﻋﻤﻠﻜﺮﺩ ‪ selection‬ﻓﺮﺍﺗﺮ ﺍﺯ ∩ ‪ ∪,‬ﻭ ‪ -‬ﺗﻌﻤﻴﻢ ﭘﻴﺪﺍ ﻣﻲ ﻛﻨﺪ‬ ‫‪.11‬‬
‫)‪σθ (E1 – E2) = σθ (E1) – σθ(E2‬‬
‫ﻭ ﻫﻤﭽﻨﻴﻦ ‪:‬‬

‫‪σθ (E1 – E2) = σθ(E1) – E2‬‬

‫ﻃﺮﺡ ﻋﻤﻠﻜﺮﺩ ﺗﻮﺯﻳﻊ ﻓﺮﺍﺗﺮ ﺍﺯ ﺍﺗﺼﺎﻝ ﺍﺳﺖ‬ ‫‪.12‬‬
‫))‪ΠL(E1 ∪ E2) = (ΠL(E1)) ∪ (ΠL(E2‬‬
‫ﺗﻤﺮﻳﻦ‪ :‬ﻗﻮﺍﻧﻴﻦ ﺗﺸﺎﺑﻬﻲ ﺍﻳﺠﺎﺩ ﻛﻨﻴﺪ‪:‬‬

‫‪ o‬ﮔﺮﻭﻫﻲ ﺑﺎ ﻋﻤﻠﻜﺮﺩ ﺍﺟﺘﻤﺎﻉ‬

‫‪ o‬ﺧﺎﺭﺟﻲ ﺗﺮﻳﻦ ﻋﻤﻠﻜﺮﺩ ﺍﺗﺼﺎﻝ ﺭﺍ ﺭﻫﺎ ﻛﻨﻴﺪ‬

‫ﻣﺜﺎﻟﻲ ﺑﺮﺍﻱ ﺗﻐﻴﻴﺮ ﺷﻜﻞ )ﺩﮔﺮ ﮔﻮﻧﻲ(‪Pushing Selections:‬‬

‫‪ :Query‬ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﻧﺎﻣﻬﺎﻱ ﺗﻤﺎﻣﻲ ﺁﻣﻮﺯﻧﺪﮔﺎﻥ ﺩﺭﺑﺨﺶ ﻣﻮﺳﻴﻘﻲ ‪ ،‬ﻫﻤﺮﺍﻩ ﺑﺎ ﻋﻨﻮﺍﻥ ﺩﺭﺳﻬﺎﻳﻲ ﻛﻪ‬ ‫‪‬‬
‫ﺁﻧﻬﺎ ﺁﻣﻮﺯﺵ ﻣﻲ ﺩﻫﻨﺪ‪.‬‬

‫‪168‬‬
‫”‪Πname, title(σdept_name= “Music‬‬
‫))))‪Πcourse_id, title (course‬‬ ‫‪(teaches‬‬ ‫‪(instructor‬‬

‫ﺗﻐﻴﻴﺮ ﺷﻜﻞ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻗﺎﻧﻮﻥ ‪.7(a‬‬ ‫‪‬‬

‫))‪Πname, title((σdept_name= “Music”(instructor‬‬
‫)))‪Πcourse_id, title (course‬‬ ‫‪(teaches‬‬

‫ﺍﺟﺮﺍﻱ ‪ selection‬ﺩﺭ ﺍﻭﻟﻴﻦ ﺯﻣﺎﻧﻲ ﻛﻪ ﺍﻧﺪﺍﺯﻩ ﻱ ﺭﺍﺑﻄﻪ ﻱ ﻣﺘﺼﻞ ﺷﺪﻩ ﺗﺎ ﺣﺪ ﻣﻤﻜﻦ ﻛﻮﭼﻚ ﺷﺪﻩ‬ ‫‪‬‬
‫ﺑﺎﺷﺪ ﺍﻣﻜﺎﻥ ﭘﺬﻳﺮ ﺍﺳﺖ‪.‬‬

‫ﻣﺜﺎﻟﻲ ﺑﺮﺍﻱ ﺗﻐﻴﻴﺮ ﺷﻜﻞ ﻣﻀﺎﻋﻒ‬

‫‪ :Query‬ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﻧﺎﻡ ﺗﻤﺎﻡ ﺍﻣﻮﺯﻧﺪﻩ ﻫﺎ ﺩﺭ ﺑﺨﺶ ﻣﻮﺳﻴﻘﻲ ﻛﻪ ﺩﺭ ﺳﺎﻝ‪ 2009‬ﺗﺪﺭﻳﺲ ﻛﺮﺩﻩ ﺍﻧﺪ ‪.‬‬ ‫‪‬‬
‫ﻫﻤﺮﺍﻩ ﺑﺎ ﻧﺎﻡ ﺩﺭﻭﺳﻲ ﻛﻪ ﺁﻧﻬﺎ ﺩﺭﺱ ﺩﺍﺩﻩ ﺍﻧﺪ‪.‬‬
‫‪Πname, title(σdept_name= “Music”∧year = 2009‬‬
‫))))‪Πcourse_id, title (course‬‬ ‫‪(teaches‬‬ ‫‪(instructor‬‬
‫ﺟﺎﺑﺠﺎﻳﻲ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺍﺗﺼﺎﻝ ﺷﺮﻛﺖ ﭘﺬﻳﺮﻱ )ﻗﺎﻧﻮﻥ ‪.(6a‬‬ ‫‪‬‬
‫‪Πname, title(σdept_name= “Music”∧gear = 2009‬‬
‫)))‪Πcourse_id, title (course‬‬ ‫)‪teaches‬‬ ‫‪((instructor‬‬

‫ﻓﺮﻡ ﺩﻭﻣﻲ ﺩﺭ ﻓﺮﺻﺘﻲ ﺑﺎ ﺑﻜﺎﺭﮔﻴﺮﻱ ﻗﺎﻧﻮﻥ "ﺍﺟﺮﺍﻱ ﺍﻧﺘﺨﺎﺑﻬﺎﻱ ﺯﻭﺩﺗﺮ" ‪،‬پﺩﻳﺪ ﻣﻲ ﺁﻳﺪ ‪.‬‬ ‫‪‬‬
‫)‪σ year = 2009 (teaches) σdept_name = “Music” (instructor‬‬
‫ﺗﻐﻴﻴﺮ ﺷﻜﻞ ﭼﻨﺪﮔﺎﻧﻪ‬

‫‪169‬‬
 Pushing Project:‫ﻣﺜﺎﻟﻲ ﺑﺮﺍﻱ ﺗﻐﻴﻴﺮ ﺷﻜﻞ‬

:‫ﺷﺎﻣﻞ‬ 
teaches) Πname, title(σdept_name= “Music” (instructor)
Πcourse_id, title (course))))

: ‫ﻫﻨﮕﺎﻣﻴﻜﻪ ﻣﺎ ﻣﺤﺎﺳﺒﻪ ﻣﻲ ﻛﻨﻴﻢ‬ 

teaches) (σdept_name = “Music” (instructor

: ‫ﻣﺎ ﺭﺍﺑﻄﻪ ﺍﻱ ﺭﺍ ﻛﻪ ﺭﻭﻱ ﺍﻟﮕﻮﻱ ﺯﻳﺮ ﻓﺮﺍﻫﻢ ﻣﻲ ﻛﻨﻴﻢ‬

(ID, name, dept_name, salary, course_id, sec_id, semester, year)

‫ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﻛﻨﻨﺪﻭ ﺣﺬﻑ ﻛﺮﺩﻥ‬8b ‫ﻭ‬8a ‫ ﺍﺯ ﻗﻮﺍﻧﻴﻦ ﻣﺸﺎﺑﻪ ﺳﺎﺯﻱ‬push ‫ﻃﺮﺡ ﻫﺎﻱ‬ 
:‫ﺧﺼﻮﺻﻴﺎﺕ ﻏﻴﺮ ﺿﺮﻭﺭﻱ ﺍﺯ ﻧﺘﺎﻳﺞ ﻣﺘﻮﺳﻂ ﺩﺭ ﺯﻳﺮ ﻧﺸﺎﻥ ﺩﺍﺩﻩ ﺷﺪﻩ‬

teaches)) Πname, title(Πname, course_id (σdept_name= “Music” (instructor)

Πcourse_id, title (course))))
170
 ‫ﺍﺟﺮﺍﻱ ﻃﺮﺡ ﺑﻪ ﺳﺎﺩﮔﻲ ﻛﺎﻫﺶ ﺍﻧﺪﺍﺯﻩ ﻱ ﺭﺍﺑﻄﻪ ﻱ ﻣﺘﺼﻞ ﺷﺪﻩ ﺍﺳﺖ ‪.‬‬ ‫‪‬‬
‫(‬ ‫ﻣﺜﺎﻟﻲ ﺑﺮﺍﻱ ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﺍﺗﺼﺎﻝ )‪Join Ordering‬‬
‫• ﺑﺮﺍﻱ ﺗﻤﺎﻣﻲ ﺭﻭﺍﺑﻂ ‪ r1‬ﻭ ‪ r2‬ﻭ ‪: r3‬‬

‫)‪r3‬‬ ‫‪(r2‬‬ ‫‪r3 = r1‬‬ ‫)‪r2‬‬ ‫‪(r1‬‬

‫‪ r3‬ﻛﺎﻣﻼ ﺑﺰﺭگ ﺑﻮﺩﻩ ﻭ ‪ r2 r1‬ﻛﻮﭼﻚ ﺑﺎﺷﺪ‪ ،‬ﺍﻧﮕﺎﻩ ﻣﺎ ﺭﺍﺑﻄﻪ ﻱ ﺯﻳﺮ ﺭﺍ ﺍﻧﺘﺨﺎﺏ‬ ‫• ﺍﮔﺮ ‪r2‬‬
‫ﻣﻲ ﻛﻨﻴﻢ‪:‬‬

‫‪r3‬‬ ‫)‪r2‬‬ ‫‪(r1‬‬

‫ﺑﺎ ﺍﻳﻦ ﻭﺟﻮﺩ ﻣﺎ ﺭﺍﺑﻄﻪ ﻱ ﻛﻮﭼﻜﺘﺮﻱ ﺭﺍ ﻣﺤﺎﺳﺒﻪ ﻭ ﺫﺧﻴﺮﻩ ﻣﻲ ﻛﻨﻴﻢ ‪.‬‬

‫(‬ ‫ﻣﺜﺎﻟﻲ ﺑﺮﺍﻱ ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﺍﺗﺼﺎﻝ )‪join Ordering‬‬

‫• ﺑﺼﻮﺭﺕ ﺯﻳﺮ ﺑﻴﺎﻥ ﻣﻲ ﺷﻮﺩ‪.‬‬

‫)‪teaches) Πname, title(σdept_name= “Music” (instructor‬‬

‫))))‪Πcourse_id, title (course‬‬
‫• ﺍﺑﺘﺪﺍ ﻣﺤﺎﺳﺒﻪ‬

‫‪Πcourse_id, title (course) compute teaches‬‬

‫ﺍﻧﺠﺎﻡ ﻣﻲ ﮔﻴﺮﺩ ﻭ ﺳﭙﺲ ﻧﺘﻴﺠﻪ ﻱ ﺍﺗﺼﺎﻝ ﺍﻧﺠﺎﻡ ﻣﻲ ﮔﻴﺮﺩ ﺑﺎ‬

‫)‪σdept_name= “Music” (instructor‬‬

‫ﺍﻣﺎ ﻧﺘﻴﺠﻪ ﻱ ﺍﻭﻟﻴﻦ ﺍﺗﺼﺎﻝ ﻣﺎﻧﻨﺪ ﻳﻚ ﺭﺍﺑﻄﻪ ﻱ ﺑﺰﺭگ ﺧﻮﺍﻫﺪ ﺷﺪ ‪.‬‬

‫• ﺍﻣﺎ ﺑﺨﺶ ﻛﻮﭼﻜﻲ ﺍﺯ ﺁﻣﻮﺯﺵ ﺩﺍﻧﺸﮕﺎﻩ ﻣﺎﻧﻨﺪ ﺑﺨﺶ ﻣﻮﺳﻴﻘﻲ ﺍﺳﺖ ‪.‬‬
‫‪ o‬ﺑﻬﺘﺮ ﺍﺳﺖ ﺍﻳﻨﮕﻮﻧﻪ ﻣﺤﺎﺳﺒﻪ ﺭﺍ ﺍﻧﺠﺎﻡ ﺩﺍﺩ‪:‬‬

‫)‪teaches σdept_name= “Music” (instructor‬‬

‫ﺷﻤﺎﺭﺵ ﺑﻴﺎﻥ ﻫﺎﻱ ﻣﺘﺸﺎﺑﻪ‬

‫‪171‬‬
 ‫• ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﺟﺴﺘﺠﻮﻫﺎ ﺍﺯ ﻗﻮﺍﻧﻴﻦ ﺗﺸﺎﺑﻪ ﺑﺮﺍﻱ ﺑﻴﺎﻧﻬﺎﻱ ﺗﻮﻟﻴﺪ ﺭﺩﻩ ﺑﻨﺪﻱ ﺷﺪﻩ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ‬
‫ﻛﻨﺪ‪.‬‬
‫• ﺗﻮﻟﻴﺪ ﺗﻤﺎﻣﻲ ﺑﻴﺎﻥ ﻫﺎﻱ ﻣﺘﺸﺎﺑﻪ ﺑﺎ ﻣﻮﺍﺭﺩ ﺯﻳﺮ ﺍﻧﺠﺎﻡ ﻣﻲ ﮔﻴﺮﺩ ‪:‬‬
‫ﺗﻜﺮﺍﺭ‬ ‫‪o‬‬
‫ﺑﻜﺎﺭﮔﻴﺮﻱ ﺗﻤﺎﻣﻲ ﻗﻮﺍﻧﻴﻦ ﺗﺸﺎﺑﻪ ﺑﻜﺎﺭﮔﺮﻓﺘﻪ ﺷﺪﻩ ﺭﻭﻱ ﻫﺮﻳﻚ ﺍﺯ ﺯﻳﺮ ﺑﻴﺎﻧﻬﺎ‪.‬‬ ‫‪‬‬
‫‪ ‬ﺍﺿﺎﻓﻪ ﻛﺮﺩﻥ ﺑﻴﺎﻧﻬﺎﻱ ﺗﻮﻟﻴﺪ ﺷﺪﻩ ﻱ ﺟﺪﻳﺪ ﺑﻪ ﻣﺠﻤﻮﻋﻪ ﻱ ﺑﻴﺎﻧﻬﺎﻱ ﻣﺘﺸﺎﺑﻪ ‪.‬‬
‫‪ ‬ﺗﺎ ﺯﻣﺎﻧﻴﻜﻪ ﻫﻴﭻ ﺑﻴﺎﻥ ﻣﺘﺸﺎﺑﻪ ﺟﺪﻳﺪ ﻱ ﻧﻤﺎﻧﺪ ‪.‬‬
‫ﺩﻭﻣﻌﺒﺮ‬ ‫‪o‬‬
‫‪‬‬
‫ﻃﺮﺡ ﺗﻮﻟﻴﺪ ﺑﻬﻴﻨﻪ ﺷﺪﻩ ﺑﺮ ﭘﺎﻳﻪ ﻱ ﻗﻮﺍﻧﻴﻦ ﺗﻐﻴﻴﺮ ﺷﻜﻞ ﺍﺳﺖ ‪.‬‬
‫‪selection‬ﻫﺎ ‪projection ،‬ﻫﺎ ﻭ ‪join‬ﻫﺎ‬ ‫‪ ‬ﻣﻮﺍﺭﺩ ﺧﺎﺹ ﺑﺮﺍﻱ ﺟﺴﺘﺠﻮ ﻓﻘﻂ ﺑﺎ‬
‫ﺍﻧﺠﺎﻡ ﻣﻲ ﮔﻴﺮﺩ‪.‬‬
‫ﭘﻴﺎﺩﻩ ﺳﺎﺯﻱ ﺗﻐﻴﻴﺮ ﺷﻜﻞ ﺑﺮﭘﺎﻳﻪ ﻱ ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ‬

‫• ﻓﺎﺻﻠﻪ ﻱ ﻧﻴﺎﺯﻣﻨﺪﻱ ﻫﺎ ﺑﺎ ﺍﺷﺘﺮﺍﻙ ﮔﺬﺍﺭﻱ ﺯﻳﺮ ﺑﻴﺎﻧﻬﺎﻱ ﺭﺍﻳﺞ ﻛﺎﻫﺶ ﭘﻴﺪﺍ ﻣﻲ ﻛﻨﺪ ‪:‬‬
‫‪ o‬ﺯﻣﺎﻧﻴﻜﻪ ‪ E1‬ﺍﺯ ‪ E2‬ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻗﻮﺍﻧﻴﻦ ﺗﺸﺎﺑﻪ ﺗﻮﻟﻴﺪ ﺷﺪﻩ ‪ ،‬ﻣﻌﻤﻮﻻ ﺩﺭ ﺳﻄﺢ ﺑﺎﻻ‬
‫ﺍﻳﻦ ﺩﻭ ﺑﺎﻫﻢ ﻣﺘﻤﺎﻳﺰ ﻫﺴﺘﻨﺪ ‪ .‬ﺯﻳﺮﺩﺭﺧﺖ ﻫﺎﻱ ﭘﺎﻳﻴﻦ ﺑﺎ ﻫﻢ ﺑﺮﺍﺑﺮﻧﺪ ﻭ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﺍﺯ‬
‫‪ pointer‬ﻫﺎﻱ ﻣﺸﺘﺮﻙ ﺍﺳﺘﻔﺎﺩﻩ ﻛﻨﻨﺪ‪.‬‬
‫‪ ‬ﻣﺜﺎﻝ ‪ :‬ﺯﻣﺎﻧﻴﻜﻪ ﺍﺗﺼﺎﻝ ﺟﺎﺑﺠﺎﻳﻲ ﭘﺬﻳﺮ ﺑﻜﺎﺭﮔﺮﻓﺘﻪ ﺷﺪﻩ‬

‫‪ o‬ﺯﻳﺮ ﺑﻴﻨﻬﺎﻱ ﻣﺴﺎﻭﻱ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﺩﺭ ﺯﻣﺎﻧﻬﺎﻱ ﮔﻮﻧﺎﮔﻮﻥ ﺗﻮﻟﻴﺪ ﺷﻮﻧﺪ‬

‫‪ ‬ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﻧﺴﺨﻪ ﻫﺎﻱ ﺯﻳﺮ ﺑﻴﺎﻧﻬﺎ ﻭ ﺍﺷﺘﺮﺍﻙ ﻳﻚ ﻛﭙﻲ ‪.‬‬
‫• ﺯﻣﺎﻥ ﻧﻴﺎﺯﻣﻨﺪﻳﻬﺎ ﺭﺍ ﻣﻲ ﺗﻮﺍﻥ ﻛﺎﻫﺶ ﺩﺍﺩ ﺑﺎ ﻋﺪﻡ ﺗﻮﻟﻴﺪ ﺗﻤﺎﻣﻲ ﺑﻴﺎﻧﻬﺎ ‪.‬‬
‫‪ o‬ﺑﺮﻧﺎﻣﻪ ﻧﻮﻳﺴﻲ ﭘﻮﻳﺎ‬

‫‪172‬‬
 ‫‪ ‬ﻣﺎ ﻳﻚ ﻣﻮﺭﺩ ﺧﺎﺹ ﺍﺯ ﺑﺮﻧﺎﻣﻪ ﻧﻮﻳﺴﻲ ﭘﻮﻳﺎ ﺑﺮﺍﻱ ﺍﺗﺼﺎﻝ ﺑﺎ ﻣﺮﺗﺐ ﺳﺎﺯﻱ ﺑﻬﻴﻨﻪ ﺭﺍ‬
‫ﺑﺮﺭﺳﻲ ﻣﻲ ﻛﻨﻴﻢ‪.‬‬
‫ﺗﺨﻤﻴﻦ ﻫﺰﻳﻨﻪ‬

‫ﻧﻴﺎﺯ ﺑﻪ ﺁﻣﺎﺭ ﺭﺍﺑﻄﻪ ﻫﺎﻱ ﻭﺭﻭﺩﻱ ﻫﺎ ‪ ،‬ﻣﺜﻞ‪ :‬ﺗﻌﺪﺍﺩ ﭼﻨﺪﺗﺎﻳﻲ ﻫﺎ‪ ،‬ﺍﻧﺪﺍﺯﻩ ﻱ ﭼﻨﺪﺗﺎﻳﻲ ﻫﺎ ‪ ،‬ﻭﺭﻭﺩﻱ ﻫﺎ ﻣﻲ‬
‫ﺗﻮﺍﻧﻨﺪ ﻧﺘﻴﺠﻪ ﻱ ﺯﻳﺮ ﺑﻴﺎﻧﻬﺎ ﺷﻮﻧﺪ‪:‬‬

‫‪ o‬ﻧﻴﺎﺯ ﺑﻪ ﺗﺨﻤﻴﻦ ﺁﻣﺎﺭﻫﺎﻱ ﻧﺘﺎﻳﺞ ﺍﺟﺮﺍ‬

‫‪ o‬ﺍﻧﺠﺎﻡ ﻧﻴﺎﺯﻫﺎﻱ ﺁﻣﺎﺭ ﺍﺿﺎﻓﻲ‬

‫ﻣﺜﻞ‪ :‬ﺗﻌﺪﺍﺩ ﺍﺭﺯﺷﻬﺎﻱ ﺍﻣﺘﻴﺎﺯﺍﺕ ﺑﺮﺍﻱ ﻳﻚ ﺧﺼﻮﺻﻴﺖ ‪.‬‬

‫ﺍﻧﺘﺨﺎﺏ ﻃﺮﺡ ﻫﺎﻱ ﺍﺭﺯﻳﺎﺑﻲ‬

‫ﺑﻴﺸﺘﺮﻳﻦ ﺑﺮﺭﺳﻲ ﻓﻌﻞ ﻭﺍﻧﻔﻌﺎﻻﺕ ﺗﻜﻨﻴﻚ ﻫﺎﻱ ﺍﺭﺯﻳﺎﺑﻲ ﺯﻣﺎﻧﻲ ﺍﻧﺠﺎﻡ ﻣﻲ ﮔﻴﺮﺩ ﻛﻪ ﻃﺮﺡ ﻫﺎﻱ ﺍﺭﺯﻳﺎﺑﻲ‬
‫ﺍﻧﺘﺨﺎﺏ ﻣﻲ ﺷﻮﻧﺪ ‪ .‬ﺍﻧﺘﺨﺎﺏ ﺍﺭﺯﺍﻧﺘﺮﻳﻦ ﺍﻟﮕﻮﺭﻳﺘﻢ ﺑﺮﺍﻱ ﻫﺮ ﻋﻤﻠﻜﺮﺩ ﺑﺼﻮﺭﺕ ﻣﺴﺘﻘﻞ ﻣﻤﻜﻦ ﺍﺳﺖ ﺑﺮﺍﻱ‬
‫ﻫﻤﻪ ﻱ ﺍﻟﮕﻮﺭﻳﺘﻢ ﻫﺎ ﺛﻤﺮﺑﺨﺶ ﻧﺒﺎﺷﺪ ‪ .‬ﺟﺴﺘﺠﻮﻱ ﻛﺎﺭﺑﺮﺩﻱ ﺑﻬﻴﻨﻪ ﺷﺪﻩ ﻋﻨﺎﺻﺮ ﺯﻳﺮ ﺭﺍ ﺗﺮﻛﻴﺐ ﻣﻲ ﻛﻨﺪ ‪:‬‬

‫‪ (1‬ﺟﺴﺘﺠﻮﻱ ﺗﻤﺎﻡ ﻃﺮﺡ ﻫﺎ ﻭ ﺍﻧﺘﺨﺎﺏ ﺑﻬﺘﺮﻳﻦ ﻃﺮﺡ ﺩﺭ ﺭﻭﺷﻲ ﺑﻪ ﺳﺒﻚ ﻣﺒﺘﻨﻲ ﺑﺮ ﻫﺰﻳﻨﻪ ‪.‬‬
‫‪ (2‬ﺑﻬﺮﻩ ﮔﻴﺮﻱ ﺍﺯ ﻛﺸﻒ ﻛﻨﻨﺪﻩ ﻫﺎ ﺑﺮﺍﻱ ﺍﻧﺘﺨﺎﺏ ﻳﻚ ﺑﺮﻧﺎﻣﻪ‪.‬‬
‫ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﻣﺒﻨﻲ ﺑﺮ ﻫﺰﻳﻨﻪ‬

‫‪. . . rn.‬‬ ‫‪r2‬‬ ‫ﺗﻔﻜﺮ ﺩﺭ ﻛﺸﻒ ﺑﻬﺘﺮﻳﻦ ﻣﻨﻈﻢ ﺳﺎﺯﻱ ﺍﺗﺼﺎﻝ ﺑﺮﺍﻱ ‪r1‬‬

‫!)‪ . (2(n – 1))!/(n – 1‬ﻣﻨﻈﻢ ﺳﺎﺯﻱ ﻫﺎﻱ ﺍﺗﺼﺎﻝ ﮔﻮﻧﺎﮔﻮﻥ ﺑﺮﺍﻱ ﺑﻴﺎﻥ ﺑﺎﻻ ‪ .‬ﺯﻣﺎﻧﻴﻜﻪ ‪ n=7‬ﻭﻋﺪﺩ‬
‫ﺑﺮﺍﺑﺮﺳﺖ ﺑﺎ ‪ 665280 ,‬ﻭ ﺑﺎ ‪ n=10‬ﻋﺪﺩﺑﺰﺭﮔﺘﺮ ﺍﺯ ‪ 176‬ﺑﻴﻠﻴﻮﻥ ﻣﻲ ﺷﻮﺩ ‪ .‬ﻧﻴﺎﺯﻱ ﺑﻪ ﺗﻮﻟﻴﺪ ﻫﻤﻪ‬
‫ﻱ ﻣﻨﻈﻢ ﺳﺎﺯﻱ ﻫﺎﻱ ﺍﺗﺼﺎﻻﺕ ﻧﻴﺴﺖ ‪ .‬ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺑﺮﻧﺎﻣﻪ ﻧﻮﻳﺴﻲ ﭘﻮﻳﺎ‪ ،‬ﻛﻢ ﻫﺰﻳﻨﻪ ﺗﺮﻳﻦ ﻣﻨﻈﻢ ﺳﺎﺯﻱ‬
‫ﻣﺤﺎﺳﺒﻪ ﺷﺪﻩ ﺍﺳﺖ ‪.‬ﺗﻨﻬﺎ ﻓﻘﻂ ﻳﻜﺒﺎﺭ ﻭ‬ ‫ﺍﺗﺼﺎﻝ ﺑﺮﺍﻱ ﻫﺮ ﺯﻳﺮ ﻣﺠﻤﻮﻋﻪ ﺍﺯ }‪{r1, r2, . . . rn‬‬
‫ﺫﺧﻴﺮﻩ ﺷﺪﻩ ﺑﺮﺍﻱ ﺍﺳﺘﺎﺩﻩ ﻱ ﺑﻌﺪﻱ ﺍﺳﺖ‪.‬‬

‫ﺑﺮﺍﻱ ﻳﺎﻓﺘﻦ ﺑﻬﺘﺮﻳﻦ ﺍﺗﺼﺎﻝ ﺑﺮﺍﻱ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺍﺭﺗﺒﺎﻃﺎﺕ ‪ n‬ﻫﺎ ‪:‬‬

‫‪173‬‬
‫• ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﺑﻬﺘﺮﻳﻦ ﻃﺮﺡ ﺑﺮﺍﻱ ﻣﺠﻤﻮﻋﻪ ‪ S‬ﻛﻪ ﺍﺭﺗﺒﺎﻃﺎﺕ ‪ n‬ﻫﺎ ﺍﺳﺖ‪ ،‬ﺷﺎﻣﻞ ﺗﻤﺎﻣﻲ ﻃﺮﺡ ﻫﺎﻱ‬
‫ﻣﻤﻜﻦ ﺑﻪ ﺷﻜﻞ ‪ (S – S1) S1‬ﻛﻪ ﺩﺭﺁﻥ ‪ S1‬ﻫﺮ ﺯﻳﺮ ﻣﺠﻤﻮﻋﻪ ﻱ ﻏﻴﺮ ﺧﺎﻟﻲ ﺍﺯ ‪ S‬ﺍﺳﺖ‪.‬‬
‫• ﻣﺤﺎﺳﺒﻪ ﻱ ﺑﺎﺯﮔﺸﺘﻲ ﻫﺰﻳﻨﻪ ﻫﺎ ﺑﺮﺍﻱ ﺍﺗﺼﺎﻝ ﺯﻳﺮﻣﺠﻤﻮﻋﻪ ﻫﺎﻱ ‪ S‬ﺍﺳﺖ ﻛﻪ ﺑﺮﺍﻱ ﭘﻴﺪﺍ ﻛﺮﺩﻥ‬
‫ﻫﺰﻳﻨﻪ ﻱ ﻫﺮ ﻃﺮﺡ ﺍﻧﺠﺎﻡ ﻣﻲ ﮔﻴﺮﺩ‪ .‬ﺍﻧﺘﺨﺎﺏ ﺍﺭﺯﺍﻧﺘﺮﻳﻦ ﺩﺭ ‪ 2n – 2‬ﻫﺎﻱ ﻣﺘﻨﺎﻭﺏ‪.‬‬
‫• ﻣﻮﺭﺩ ﻣﺒﻨﺎ ﺑﺮﺍﻱ ﺑﺎﺯﮔﺸﺖ‪ :‬ﺍﺭﺗﺒﺎﻁ ﻣﻨﻔﺮﺩ ﻃﺮﺡ ﺩﺳﺘﺮﺳﻲ ‪.‬‬
‫• ﺑﻜﺎﺭﮔﻴﺮﻱ ﺗﻤﺎﻣﻲ ‪ selection‬ﻫﺎ ﺭﻭﻱ ‪ Ri‬ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺑﻬﺘﺮﻳﻦ ﺍﻧﺘﺨﺎﺏ ﺩﺭ ﻓﻬﺮﺳﺖ ﻫﺎﻱ‬
‫ﺭﻭﻱ ‪.Ri‬‬

‫• ﺯﻣﺎﻧﻴﻜﻪ ﻃﺮﺡ ﺑﺮﺍﻱ ﻫﺮ ﺯﻳﺮ ﻣﺠﻤﻮﻋﻪ ﻱ ﻣﺤﺎﺳﺒﻪ ﺷﺪﻩ ‪ ،‬ﺫﺧﻴﺮﻩ ﻭ ﺍﺳﺘﻔﺎﺩﻩ ﻣﺠﺪﺩ ﻣﻲ ﺷﻮﺩ ﻭ‬
‫ﺩﻭﺑﺎﺭﻩ ﺧﻮﺍﺳﺘﻪ ﻣﻲ ﺷﻮﺩ ﺑﻪ ﺟﺎﻱ ﻣﺤﺎﺳﺒﻪ ﻣﺠﺪﺩ ﺁﻥ ‪.‬‬

‫ﺍﻟﮕﻮﺭﻳﺘﻢ ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﺗﺮﺗﻴﺐ ﺍﺗﺼﺎﻝ‬

‫ﺑﺮﺧﻲ ﺍﺯ ﺗﺒﺪﻳﻼﺕ ﺑﺮﺍﻱ ﺍﺟﺎﺯﻩ ﻓﻬﺮﺳﺖ ﻫﺎﻱ ﺣﻠﻘﻪ ﻫﺎﻱ ﺗﻮﺩﺭﺗﻮ ﺍﺗﺼﺎﻝ ﻣﻲ ﻳﺎﺑﻨﺪ ﺭﻭﻱ ﺍﺭﺗﺒﺎﻃﺎﺗﻲ ﻛﻪ‬
‫‪ Selection‬ﻫﺎﻳﻲ ﺩﺍﺭﺩ‪.‬‬
‫ﺩﺭﺧﺖ ﻫﺎﻱ ﺍﺗﺼﺎﻝ ﻋﻤﻖ ﭼﭗ‬

‫ﺩﺭ ﺩﺭﺧﺘﻬﺎﻱ ﺍﺗﺼﺎﻝ ﻋﻤﻖ ﭼﭗ ﻭﺭﻭﺩﻱ ﺳﻤﺖ ﺭﺍﺳﺖ ﺑﺮﺍﻱ ﻫﺮ ﺍﺗﺼﺎﻝ ﻳﻚ ﺭﺍﺑﻄﻪ ﺍﺳﺖ‪ ،‬ﻧﻪ ﻧﺘﻴﺠﻪ ﺍﻱ ﺍﺯ‬
‫ﺍﺗﺼﺎﻝ ﻣﺘﻮﺳﻂ‪.‬‬

‫‪174‬‬
 ‫ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﻣﺒﺘﻨﻲ ﺑﺮ ﻗﻮﺍﻧﻴﻦ ﺗﺸﺎﺑﻪ‬

‫ﻗﻮﺍﻧﻴﻦ ﺗﺸﺎﺑﻪ ﻓﻴﺰﻳﻜﻲ ﺍﺟﺎﺯﻩ ﻣﻲ ﺩﻫﻨﺪ ﻛﻪ ﻃﺮﺡ ﺟﺴﺘﺠﻮﻱ ﻣﻨﻄﻘﻲ ﺑﺮﮔﺮﺩﺍﻧﺪﻩ ﺷﺪﻩ ﺑﻪ ﺑﻬﻴﻨﻪ ﺳﺎﺯ ﻣﻮﺛﺮ‬
‫ﺑﺮ ﻣﺒﻨﺎﻱ ﻗﻮﺍﻧﻴﻦ ﺗﺸﺎﺑﻪ ﻭﺍﺑﺴﺘﻪ ﺍﺳﺖ ﺑﻪ‪:‬‬

‫ﻧﻤﺎﻳﺶ ﻓﻀﺎﻱ ﻛﺎﺭﺍﻣﺪ ﺑﻴﺎﻧﻬﺎﻳﻲ ﻛﻪ ﻣﻤﺎﻧﻌﺖ ﻣﻲ ﻛﻨﺪ ﺍﺯ ﺳﺎﺧﺘﻦ ﻛﭙﻲ ﻫﺎﻱ ﭼﻨﺪﮔﺎﻧﻪ ﺑﺮ‬
‫ﺭﻭﻱ ﺯﻳﺮ ﺑﻴﺎﻧﻬﺎ‪.‬‬
‫ﺗﻜﻨﻴﻚ ﻫﺎﻱ ﻛﺎﺭﺁﻣﺪ ﺑﺮﺍﻱ ﻛﺸﻒ ﻧﺴﺨﻪ ﻫﺎﻱ ﺍﺳﺘﺨﺮﺍﺟﻲ ﺍﺯ ﺑﻴﺎﻧﻬﺎ‪.‬‬
‫ﻓﺮﻡ ﺑﺮﻧﺎﻣﻪ ﻧﻮﻳﺴﻲ ﭘﻮﻳﺎ ﺑﺮ ﻣﺒﻨﺎﻱ ﺣﻔﻆ ﻛﺮﺩﻥ ﻛﻪ ﺫﺧﻴﺮﻩ ﻣﻲ ﻛﻨﺪ ﺑﻬﺘﺮﻳﻦ ﻃﺮﺡ ﺑﺮﺍﻱ‬
‫ﺯﻳﺮ ﺍﺟﺮﺍﻱ ﺍﻭﻟﻴﻦ ﺯﻣﺎﻥ ﺑﻬﻴﻨﻪ ﺷﺪﻩ ﻭ ﻣﺠﺪﺩ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺷﻮﺩ ﺩﺭ ﻓﺮﺍﺧﻮﺍﻧﻲ ﺗﻜﺮﺍﺭ ﺑﻬﻴﻨﻪ‬
‫ﺳﺎﺯﻱ ﺷﺪﻩ ﺩﺭ ﺯﻳﺮ ﺑﻴﺎﻧﻬﺎﻱ ﺑﺮﺍﺑﺮ‪.‬‬
‫ﺗﻜﻨﻴﻚ ﻫﺎﻱ ﻫﺮﺱ ﻣﺒﺘﻨﻲ ﺑﺮ ﻫﺰﻳﻨﻪ ﻛﻪ ﺟﻠﻮﮔﻲﺭﻱ ﻣﻲ ﻛﻨﺪ ﺍﺯ ﺗﻮﻟﻴﺪ ﻫﻤﻪ ﻃﺮﺡ ﻫﺎ ‪.‬‬

‫ﭘﻴﺸﮕﺎﻣﻲ ﺑﺎ ﭘﺮﻭژﻩ ﻱ ﻭﻟﻜﺎﻧﻮ ﻭ ﭘﻴﺎﺩﻩ ﺳﺎﺯﻱ ﺩﺭﺑﻬﻴﻨﻪ ﺳﺎﺯ ‪. SQL Server‬‬

‫ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﺍﻛﺘﺸﺎﻓﻲ‬

‫ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﻣﺒﺘﻨﻲ ﺑﺮﻫﺰﻳﻨﻪ ﭘﺮﻫﺰﻳﻨﻪ ﺍﺳﺖ ‪،‬ﺣﺘﻲ ﺑﺎ ﺑﺮﻧﺎﻣﻪ ﻧﻮﻳﺴﻲ ﭘﻮﻳﺎ ‪.‬ﺳﻴﺴﺘﻢ ﻫﺎ ﻣﻲ ﺗﻮﺍﻧﻨﺪ ﺍﺳﺘﻔﺎﺩﻩ‬
‫ﻛﻨﻨﺪ ﺍﺯ ﺭﻭﺵ ﺍﻛﺘﺸﺎﻓﻲ ﺑﺮﺍﻱ ﻛﺎﻫﺶ ﺗﻌﺪﺍﺩ ﺍﻧﺘﺨﺎﺑﻬﺎ ﻛﻪ ﺑﺎﻳﺪ ﺳﺎﺧﺘﻪ ﺷﻮﺩ ﺩﺭ ﻣﺪﻝ ﻣﺒﺘﻨﻲ ﺑﺮ ﻫﺰﻳﻨﻪ ‪.‬‬
‫ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﺍﻛﺘﺸﺎﻓﻲ ‪،‬ﺩﺭﺧﺖ ﺟﺴﺘﺠﻮ ﺭﺍ ﺑﺎ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﻗﻮﺍﻧﻴﻦ ﻛﻪ ﻣﻌﻤﻮﻻ ﻣﻮﺟﺐ‬
‫ﺑﻬﺒﻮﺩ ﻧﻤﺎﻳﺶ ﺍﺟﺮﺍ ﻣﻲ ﺷﻮﺩ‪ ،‬ﺗﻐﻴﻴﺮ ﺷﻜﻞ ﻣﻲ ﺩﻫﺪ‪:‬‬

‫‪175‬‬
 ‫ﺍﺟﺮﺍﻱ ﺳﺮﻳﻊ ‪) selection‬ﻛﺎﻫﺶ ﺗﻌﺪﺍﺩ ﭼﻨﺪﺗﺎﻳﻲ ﻫﺎ (‬ ‫‪‬‬
‫ﺍﺟﺮﺍﻱ ﺳﺮﻳﻊ ‪) projection‬ﻛﺎﻫﺶ ﺗﻌﺪﺍﺩ ﺧﺼﻮﺻﻴﺎﺕ (‬ ‫‪‬‬
‫‪ selection‬ﻭ ﺍﺗﺼﺎﻝ ﻋﻤﻠﻜﺮﺩﻫﺎ ‪).‬ﺑﺎ ﻛﻮﭼﻚ‬ ‫ﺍﺟﺮﺍﻱ ﺑﻴﺸﺘﺮﻳﻦ ﻣﺤﺪﻭﺩﻛﻨﻨﺪﻩ ﻫﺎﻱ‬ ‫‪‬‬
‫ﻛﺮﺩﻥ ﺍﻧﺪﺍﺯﻩ ﻱ ﻧﺘﻴﺠﻪ( ﻗﺒﻞ ﺍﺯ ﺑﻘﻴﻪ ﻱ ﻋﻤﻠﻜﺮﺩ ﻫﺎﻱ ﻣﺸﺎﺑﻪ ‪.‬‬
‫ﺑﻌﻀﻲ ﺍﺯ ﺳﻴﺴﺘﻢ ﻫﺎ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺷﻮﺩ ﻓﻘﻂ ﺑﻪ ﺻﻮﺭﺕ ﺍﻛﺘﺸﺎﻓﻲ‪ ،‬ﺑﻘﻴﻪ ﻱ ﺍﻛﺘﺸﺎﻓﻲ ﻫﺎ‬ ‫‪‬‬
‫ﺑﺎ ﺑﺨﺸﻲ ﺍﺯ ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﻣﺒﺘﻨﻲ ﺑﺮ ﻫﺰﻳﻨﻪ ﺗﺮﻛﻴﺐ ﻣﻲ ﺷﻮﻧﺪ ‪.‬‬

‫ﺳﺎﺧﺘﺎﺭ ﺑﻬﻴﻨﻪ ﺳﺎﺯ ﺟﺴﺘﺠﻮ‬

‫ﺑﻴﺸﺘﺮ ﺑﻬﻴﻨﻪ ﺳﺎﺯﻫﺎ ﻓﻘﻂ ﻋﻤﻖ ﭼﭗ ﺍﺗﺼﺎﻝ ﺩﺳﺘﻮﺭﺍﺕ ﺭﺍ ﺑﺮﺭﺳﻲ ﻣﻲ ﻛﻨﺪ ‪ .‬ﺑﻌﻼﻭﻩ ﺭﻭﺵ ﻫﺎﻱ ﺍﻛﺘﺸﺎﻓﻲ‬
‫ﺑﺮﺍﻱ ﻓﺸﺎﺭ ‪selection‬ﻫﺎ ﻭ ‪ projection‬ﻫﺎ ﺩﺭﺧﺖ ﺟﺴﺘﺠﻮ ﺭﺍ ﭘﺎﻳﻴﻦ ﻣﻲ ﺁﻭﺭﺩ‪ .‬ﻛﺎﻫﺶ ﭘﻴﭽﻴﺪﮔﻲ‬
‫ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﻭ ﺗﻮﻟﻴﺪ ﻃﺮﺡ ﻫﺎﻱ ﻗﺎﺑﻞ ﺟﻮﺍﺑﮕﻮﻳﻲ ﺑﻪ ﺧﻂ ﺍﺭﺯﻳﺎﺑﻲ ‪ .‬ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﺍﻛﺘﺸﺎﻓﻲ ﺩﺭ ﺑﺮﺧﻲ ﺍﺯ‬
‫ﻭﺭژﻥ ﻫﺎﻱ ﺍﻭﺭﺍﻛﻞ ﺍﺳﺘﻔﺎﺩﻩ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬

‫** ﭘﻴﭽﻴﺪﮔﻲ ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﺟﺴﺘﺠﻮ ﺩﺭ ‪SQL‬‬

‫ﺳﺎﺧﺘﺎﺭ ﺑﻬﻴﻨﻪ ﺳﺎﺯ ﺟﺴﺘﺠﻮ‬

‫ﺭﺍ‬ ‫ﺑﺮﺧﻲ ﺍﺯ ﺑﻬﻴﻨﻪ ﺳﺎﺯﻫﺎﻱ ﺟﺴﺘﺠﻮ ﺍﻧﺘﺨﺎﺏ ﻫﺎﻱ ﺍﻛﺘﺸﺎﻓﻲ ﺭﺍ ﻭ ﺗﻮﻟﻴﺪ ﻃﺮﺡ ﻫﺎﻱ ﺩﺳﺘﺮﺳﻲ ﻣﺘﻨﺎﻭﺏ‬
‫ﻛﺎﻣﻞ ﻣﻲ ﻛﻨﺪ‪.‬‬

‫‪ o‬ﻣﻌﺒﺮ ﺍﺳﺘﻔﺎﺩﻩ ﺷﺪﻩ ﻱ ﺗﻜﺮﺍﺭ ﺷﻮﻧﺪﻩ‬

‫‪ ‬ﻧﻮﺷﺘﻦ ﺩﻭﺑﺎﺭﻩ ﻱ ﺍﻛﺘﺸﺎﻓﻲ ﺑﻼﻙ ﺳﺎﺧﺘﺎﺭ ﺗﻮﺩﺭﺗﻮ ﻭ ﺍﺟﺘﻤﺎﻉ ‪.‬‬
‫‪ ‬ﭘﻴﺮﻭﻱ ﺩﺳﺘﻮﺭﻫﺎﻱ ﺍﺗﺼﺎﻝ ﻣﺒﺘﻨﻲ ﺑﺮ ﻫﺰﻳﻨﻪ ﺑﺮﺍﻱ ﻫﺮ ﺑﻼﻙ ‪.‬‬
‫‪ o‬ﺑﺮﺧﻲ ﺍﺯ ﺑﻬﻴﻨﻪ ﺳﺎﺯﻫﺎ ﺑﻜﺎﺭ ﻣﻲ ﮔﻴﺮﻧﺪ ﺗﻐﻴﻴﺮ ﺷﻜﻞ ﻫﺎ ﺭﺍ ﺑﺮﺍﻱ ﺟﺴﺘﺠﻮﻱ ﻛﺎﻣﻞ ﻭ ﻋﺪﻡ‬
‫ﻭﺍﺑﺴﺘﮕﻲ ﺑﻪ ﺑﻼﻙ ﺳﺎﺧﺘﺎﺭ‪.‬‬
‫‪ o‬ﺑﻮﺩﺟﻪ ﻱ ﻫﺰﻳﻨﻪ ﻱ ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﺑﺮﺍﻱ ﺗﻮﻗﻒ ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﺳﺮﻳﻊ )ﺍﮔﺮ ﻫﺰﻳﻨﻪ ﻃﺮﺡ ﻛﻨﺘﺮ ﺍﺯ‬
‫ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﺑﺎﺷﺪ‪(.‬‬

‫‪176‬‬
 ‫‪ o‬ﻃﺮﺡ ﻣﺨﺰﻥ ﺑﺮﺍﻱ ﺍﺳﺘﻔﺎﺩﻩ ﻱ ﻣﺠﺪﺩ ﻗﺒﻞ ﺍﺯ ﻣﺤﺎﺳﺒﻪ ﻱ ﻃﺮﺡ ﺍﮔﺮ ﺟﺴﺘﺠﻮ ﻧﺎﺷﻲ ﺍﺯ‬
‫ﭘﺎﻳﺪﺍﺭﻱ ﻫﺎﻱ ﻣﺨﺘﻠﻒ ﺩﺭ ﺟﺴﺘﺠﻮ‪.‬‬

‫ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺍﻛﺘﺸﺎﻓﻲ ﻫﺎ ‪ ،‬ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﺟﺴﺘﺠﻮ ﻣﺒﺘﻨﻲ ﺑﺮ ﻫﺰﻳﻨﻪ ﺭﺍ ﺩﺭ ﻣﺨﺎﺭﺝ ﻛﻠﻲ ﺗﺤﻤﻴﻞ ﻗﺎﺑﻞ‬
‫ﺗﻮﺟﻬﻲ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫‪ o‬ﺍﻣﺎ ﺍﺭﺯﺵ ﺁﻥ ﺑﺮﺍﻱ ﺟﺴﺘﺠﻮﻫﺎ ﭘﺮﻫﺰﻳﻨﻪ ﺍﺳﺖ ‪.‬‬

‫‪ o‬ﺑﻬﻴﻨﻪ ﺳﺎﺯﻫﺎ گﺍﻫﻲ ﺍﺯ ﺍﻛﺘﺸﺎﻓﻲ ﻫﺎﻱ ﺳﺎﺩﻩ ﺑﺮﺍﻱ ﺟﺴﺘﺠﻮﻫﺎﻱ ﺧﻴﻠﻲ ﺍﺭﺯﺍﻥ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ‬
‫ﻛﻨﻨﺪ ﻭ ﺷﻤﺎﺭﺵ ﺟﺎﻣﻊ ﺭﺍ ﺑﺮﺍﻱ ﺟﺴﺘﺠﻮﻫﺎﻱ ﮔﺮﺍﻥ ﺍﺟﺮﺍ ﻣﻲ ﻛﻨﺪ ‪.‬‬

‫ﺁﻣﺎﺭ ﺑﺮﺍﻱ ﺗﺨﻤﻴﻦ ﻫﺰﻳﻨﻪ )ﺍﻃﻼﻋﺎﺕ ﺁﻣﺎﺭﻱ ﺑﺮﺍﻱ ﺗﺨﻤﻴﻦ ﻫﺰﻳﻨﻪ(‬

‫‪ : nr ‬ﺗﻌﺪﺍﺩ ﭼﻨﺪﺗﺎﻳﻲ ﻫﺎ ﺩﺭ ﺭﺍﺑﻄﻪ ﻱ ‪. r‬‬

‫‪ : br ‬ﺗﻌﺪﺍﺩ ﺑﻼﻙ ﻫﺎ ﻛﻪ ﺷﺎﻣﻞ ﭼﻨﺪﺗﺎﻳﻲ ﻫﺎﻱ ‪ r‬ﺍﺳﺖ‪.‬‬
‫‪ : lr ‬ﺍﻧﺪﺍﺯﻩ ﻱ ﭼﻨﺪﺗﺎﻳﻲ ﻫﺎﻱ ‪. r‬‬
‫‪ : fr ‬ﺍﻧﺴﺪﺍﺩ ﻓﺎﻛﺘﻮﺭ ‪r‬‬
‫‪ : V(A, r): ‬ﺗﻌﺪﺍﺩ ﺍﻣﺘﻴﺎﺯﺍﺕ ﻣﻘﺎﺩﻳﺮﺍﺳﺖ ﻛﻪ ﻧﺸﺎﻥ ﺩﻩ ﻥﺩﻩ ‪ r‬ﺑﺮﺍﻱ ﺧﺼﻮﺻﻴﺖ ﺍﺳﺖ ﻭ ‪ A‬ﺑﺮﺍﺑﺮ‬
‫ﺑﺎ ﺍﻧﺪﺍﺯﻩ ﻱ )‪ ∏A(r‬ﺍﺳﺖ‪.‬‬
‫‪ ‬ﺍﮔﺮ ﭼﻨﺪﺗﺎﻳﻲ ﻫﺎﻱ ‪ r‬ﺫﺧﻴﺮﻩ ﺷﺪﻩ ﺑﺼﻮﺭﺕ ﻓﻴﺰﻳﻜﻲ ﺩﺭ ﻳﻚ ﻓﺎﻳﻞ ﺁﻧﮕﺎﻩ ‪:‬‬
‫‪nr ‬‬
‫‪‬‬
‫= ‪br‬‬ ‫‪‬‬
‫‪f r ‬‬
‫‪‬‬
‫‪‬‬

‫ﻧﻤﺎﻳﺶ ﻧﺤﻮﻩ ﻱ ﺍﻧﺘﺸﺎﺭ‬

‫• ﻧﻤﺎﻳﺶ ﻧﺤﻮﻩ ﻱ ﺍﻧﺘﺸﺎﺭ ﺭﻭﺯ ﺧﺼﻮﺻﻴﺖ ﺳﻦ ﺩﺭ ﺭﺍﺑﻄﻪ ﻱ ﺍﻓﺮﺍﺩ ‪.‬‬

‫‪177‬‬
 ‫ﺑﺮﺁﻭﺭﺩ ﺍﻧﺪﺍﺯﻩ ‪selection‬‬
‫• )‪σA=v(r‬‬
‫‪ : nr / V(A,r) o‬ﺗﻌﺪﺍﺩ ﺭﻛﻮﺭﺩﻫﺎ ﻛﻪ ﺭﺿﺎﻳﺖ ‪ selection‬ﺭﺍ ﺍﻳﺠﺎﺩ ﻣﻲ ﻛﻨﺪ‪.‬‬
‫‪ o‬ﺷﺮﺍﻳﻂ ﻣﺘﺴﺎﻭﻱ ﺭﻭﻱ ﻛﻠﻴﺪ ﺧﺼﻮﺻﻴﺖ ‪ :‬ﺍﻧﺪﺍﺯﻩ ﻱ ﺗﺨﻤﻴﻦ=‪.1‬‬
‫• )‪σA≤V(r) (case of σA ≥ V(r) is symmetric‬‬
‫‪ C o‬ﻣﺸﺨﺺ ﻣﻲ ﻛﻨﺪ ﺗﺨﻤﻴﻦ ﺗﻌﺪﺍﺩ ﭼﻨﺪﺗﺎﻳﻲ ﻫﺎ‬
‫‪ o‬ﺍﮔﺮ )‪ min(A,r‬ﻭ )‪ max(A,r‬ﺩﺭ ﻛﺎﺗﺎﻟﻮگ ﺩﺭ ﺩﺳﺘﺮﺳﻨﺪ‪.‬‬

‫• ﺍﮔﺮ ﻧﻤﺎﻳﺶ ﻧﺤﻮﻩ ﻱ ﺍﻧﺘﺸﺎﺭ ﺩﺭ ﺩﺳﺘﺮﺱ ﺑﺎﺷﺪ‪ ،‬ﻣﻲ ﺗﻮﺍﻧﺪ ﺗﺨﻤﻴﻦ ﺑﺎﻻ ﺭﺍ ﺗﺼﺤﻴﺢ ﻛﻨﺪ ‪.‬‬
‫• ﺩﺭ ﻏﻴﺎﺏ ﺍﻃﻼﻋﺎﺕ ﺁﻣﺎﺭﻱ ‪ C‬ﻓﺮﺽ ﺷﺪﻩ ﺍﺳﺖ ﺗﺎ ‪ nr / 2‬ﺑﺎﺷﺪ‪.‬‬

‫‪178‬‬
 ‫ﻫﺎ ﻱ ﺷﺎﺧﻪ ﺍﻱ ﺗﻮ ﺩﺭ ﺗﻮ‪:‬‬ ‫ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ‪sub query‬‬
‫• ﻣﺜﺎﻝ ﻛﻮﺋﺮﻱ ﺷﺎﺧﻪ ﺍﻱ ﺗﻮ ﺩﺭ ﺗﻮ‪:‬‬
‫‪selectname‬‬
‫‪from instructor‬‬
‫* ‪where exists (select‬‬
‫‪from teaches‬‬
‫)‪where instructor.ID = teaches.ID and teaches.year = 2007‬‬
‫‪SQL‬ﻣﻔﻬﻮﻣﻲ ﺍﺯ ﺭﻓﺘﺎﺭ ‪sub query‬ﻫﺎ ﺗﻮ ﺩﺭ ﺗﻮ ﺍﺳﺖ ﺩﺭ ﻛﻪ ﺩﺭ ﺁﻥ ﺗﻮﺍﺑﻊ ﺭﺍ ﻛﻪ ﭘﺎﺭﺍﻣﺘﺮﻫﺎ ﻣﻴﮕﻴﺮﺩ ﻭ‬
‫ﻳﻚ ﺍﺭﺯﺵ ﻭﺍﺣﺪ ﻳﺎ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺍﺭﺯﺵ ﻫﺎ ﺭﺍ ﺑﺎﺯ ﻣﻲ ﮔﺮﺩﺍﻧﺪ‪ .‬ﭘﺎﺭﺍﻣﺘﺮﻫﺎﻱ ﻣﺘﻐﻴﺮ ﺍﺯ ﻛﻮﺋﺮﻱ ﺩﺭ ﺳﻄﺢ‬
‫ﺑﻴﺮﻭﻧﻲ ﻫﺴﺘﻨﺪ ﻛﻪ ﺩﺭ ‪ Sub query‬ﻫﺎﻱ ﺗﻮ ﺩﺭ ﺗﻮ ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺷﻮﺩ‪ ،‬ﺍﻳﻦ ﻣﺘﻐﻴﺮﻫﺎ ﻣﺘﻐﻴﺮﻫﺎﻱ‬
‫ﻫﻤﺒﺴﺘﮕﻲ ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﻧﺪ‪.‬‬
‫ﻣﻔﻬﻮﻡ‪ Sub query ،‬ﺗﻮﺩﺭﺗﻮ ﻳﻚ ﺑﺎﺭ ﺑﺮﺍﻱ ﻫﺮ ﺗﺎﭘﻞ ﺩﺭ ﻣﻘﻄﻊ ﻣﺤﺼﻮﻝ ﺗﻮﻟﻴﺪ ﺷﺪﻩ ﺗﻮﺳﻂ ﺳﻄﺢ‬
‫ﺧﺎﺭﺟﻲ ﺍﺳﺖ‪ .‬ﭼﻨﻴﻦ ﺍﺭﺯﻳﺎﺑﻲ‪ ،‬ﺍﺭﺯﻳﺎﺑﻲ ﺍﺭﺗﺒﺎﻁ ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﻧﺪ‪.‬‬
‫ﺗﻮﺟﻪ‪ :‬ﺩﺭ ﺷﺮﺍﻳﻂ ﺩﻳﮕﺮ ﻛﻪ ﻣﻤﻜﻦ ﺍﺳﺖ ﺑﺮﺍﻱ ﻣﺤﺎﺳﺒﻪ ﭘﻴﻮﺳﺘﻦ ﻗﺒﻞ ﺍﺯ ﺍﺟﺮﺍﻱ ‪ Sub query‬ﻫﺎﻱ ﺗﻮ‬
‫ﺩﺭ ﺗﻮ ﻣﻮﺭﺩ ﺍﺳﺘﻔﺎﺩﻩ ﻗﺮﺍﺭ ﺑﮕﻴﺮﺩ ‪ .‬ﺍﺭﺯﻳﺎﺑﻲ ﻫﺎﻱ ﻫﻤﺒﺴﺘﻪ ﻣﻤﻜﻦ ﺍﺳﺖ ﻛﺎﻣﻼ ﻧﺎﻛﺎﺭﺁﻣﺪ ﺑﺎﺷﻨﺪ ‪ .‬ﺗﻌﺪﺍﺩ‬
‫ﺯﻳﺎﺩﻱ ﺍﺯ ﺗﻤﺎﺱ ﻣﻤﻜﻦ ﺍﺳﺖ ﺍﺯ ‪sub query‬ﻫﺎﻱ ﺗﻮ ﺩﺭ ﺗﻮ ﺳﺎﺧﺘﻪ ﺷﻮﺩ ‪ .‬ﻣﻤﻜﻦ ﺍﺳﺖ ﻳﻚ ﻧﺘﻴﺠﻪ ﻏﻴﺮ‬
‫ﺿﺮﻭﺭﻱ ﺗﺼﺎﺩﻓﻲ ‪ I / O‬ﺑﻪ ﻭﺟﻮﺩ ﺁﻳﺪ ‪ .‬ﺑﻬﻴﻨﻪ ‪ SQL‬ﺗﻼﺵ ﺑﺮﺍﻱ ﺗﺒﺪﻳﻞ ﻭ ﭘﻴﻮﻧﺪ ‪sub query‬ﻫﺎ ﺗﻮ ﺩﺭ‬
‫ﺗﻮ ﻛﻪ ﺩﺭ ﺁﻥ ﻣﻤﻜﻦ ﻗﺎﺩﺭ ﺑﻪ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺗﻜﻨﻴﻚ ﻫﺎﻱ ﻛﺎﺭﺁﻣﺪ ﭘﻴﻮﻧﺪ ﺑﺎﺷﺪ ‪.‬‬
‫ﻣﺜﺎﻝ‪ :‬ﻛﻮﺋﺮﻱ ﺗﻮ ﺩﺭ ﺗﻮ ﺯﻭﺩﺗﺮ ﺭﺍ ﻣﻲ ﺗﻮﺍﻥ ﺑﺎﺯﻧﻮﻳﺴﻲ ﻛﺮﺩ‬
‫‪select name‬‬
‫‪from instructor, teaches‬‬
‫‪where instructor.ID = teaches.ID and teaches.year = 2007‬‬
‫ﺗﻮﺟﻪ‪ :‬ﺍﻳﻦ ﺩﻭﻛﻮﺋﺮﻱ ﻣﺨﺘﻠﻒ ‪ 2‬ﻧﺴﺨﻪ ﺗﻜﺮﺍﺭﻱ ﺗﻮﻟﻴﺪ ﻛﺮﺩﻩ ﺍﻧﺪ )ﭼﺮﺍ؟(‬
‫‪ teaches‬ﻣﻲ ﺗﻮﺍﻧﺪ ﻧﺴﺨﻪ ﺗﻜﺮﺍﺭﻱ ‪ IDs‬ﺭﺍ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ ‪ ،‬ﻣﻲ ﺗﻮﺍﻧﺪ ﻧﺴﺨﻪ ﺗﻜﺮﺍﺭﻱ ﺭﺍ ﺑﻪ ﺩﺭﺳﺘﻲ‬
‫ﻫﻤﺎﻧﻄﻮﺭ ﻛﻪ ﻣﺎ ﺧﻮﺍﻫﻴﻢ ﺩﻳﺪ ﺍﺻﻼﺡ ﻛﻨﺪ‬

‫‪179‬‬
‫• ﻋﻤﻮﻣﺎ‪ ،‬ﻣﻤﻜﻦ ﻧﻴﺴﺖ ﻣﺴﺘﻘﻴﻤﺎ ﺍﺯ ﺳﻄﺢ ﺟﺰ ﺧﺎﺭﺟﻲ ‪query‬ﺑﻪ ﺳﻤﺖ ﺟﺰ ‪Sub query‬ﺗﻮ ﺩﺭ ﺗﻮ‬
‫ﺣﺮﻛﺖ ﻛﻨﻴﻢ‪ .‬ﺭﺍﺑﻄﻪ ﺑﻪ ﻃﻮﺭ ﻣﻮﻗﺖ ﺍﻳﺠﺎﺩ ﺷﺪﻩ‪ ،‬ﻭ ﺩﺭ ﺑﺪﻧﻪ ﺳﻄﺢ ﺑﻴﺮﻭﻧﻲ ‪ query‬ﻣﻮﺭﺩ ﺍﺳﺘﻔﺎﺩﻩ ﻗﺮﺍﺭ‬
‫ﻣﻴﮕﻴﺮﺩ‪.‬‬

‫ﺑﻪ ﻃﻮﺭ ﻛﻠﻲ‪ SQL, query ،‬ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﻪ ﺻﻮﺭﺕ ﺯﻳﺮ ﺑﺎﺯ ﻧﻮﻳﺴﻲ ﺷﻮﺩ‪:‬‬
‫ﺑﺎﺯﻧﻮﻳﺴﻲ‪:‬‬
‫…‪select‬‬
‫‪fromL1‬‬
‫* ‪whereP1and exists (select‬‬
‫‪fromL2‬‬
‫)‪whereP2‬‬
‫ﺑﻪ‪:‬‬
‫‪create tablet1 as‬‬
‫‪select distinct V‬‬
‫‪from L2‬‬
‫‪1‬‬
‫‪where P2‬‬

‫… ‪select‬‬
‫‪from L1,t1‬‬
‫‪2‬‬
‫‪where P1and P2‬‬
‫‪1‬‬
‫‪P2 ‬ﺷﺎﻣﻞ ﮔﺬﺍﺭﻩ ﺩﺭ ‪P2‬ﻛﻪ ﻫﻴﭻ ﻣﺘﻐﻴﺮ ﻫﻤﺒﺴﺘﮕﻲ ﺭﺍ ﺷﺎﻣﻞ ﻧﻴﺴﺖ‪.‬‬
‫‪2‬‬
‫‪ reintroduces P2 ‬ﺷﺎﻣﻞ ﮔﺬﺍﺭﻩ ﻣﺘﻐﻴﺮﻫﺎﻱ ﻫﻤﺒﺴﺘﮕﻲ‪ ،‬ﺑﺎ ﺭﻭﺍﺑﻂ ﻣﻨﺎﺳﺐ ﺗﻐﻴﻴﺮ ﻧﺎﻡ ﺍﺳﺖ ‪.‬‬
‫‪V ‬ﺷﺎﻣﻞ ﺗﻤﺎﻡ ﻭﻳﮋﮔﻲ ﻫﺎﻱ ﻣﻮﺭﺩ ﺍﺳﺘﻔﺎﺩﻩ ﺩﺭ ﮔﺬﺍﺭﻩ ﺑﺎ ﻣﺘﻐﻴﺮ ﻫﻤﺒﺴﺘﮕﻲ ﺍﺳﺖ‪.‬‬
‫‪ ‬ﺩﺭ ﻣﺜﺎﻝ ﻣﺎ‪ query ،‬ﺍﺻﻠﻲ ﺷﺎﺧﻪ ﺍﻱ ﺗﻮ ﺩﺭ ﺗﻮ ﻣﻲ ﺗﻮﺍﻧﺪ ﺗﻐﻴﻴﺮ ﭘﻴﺪﺍ ﻛﻨﺪ‬
‫‪create table t1as‬‬
‫‪select distinct ID‬‬
‫‪180‬‬
 ‫‪from teaches‬‬
‫‪where year = 2007‬‬

‫‪select name‬‬
‫‪from instructor, t1‬‬
‫‪where t1.ID = instructor.ID‬‬
‫ﺭﻭﻧﺪ ﺟﺎﺑﺠﺎﻳﻲ ﻳﻚ ‪query‬ﺗﻮ ﺩﺭ ﺗﻮ ﺑﻮﺍﺳﻄﻪ ﻳﻚ ‪query‬ﺑﺎ ﭘﻴﻮﺳﺘﻦ ﺭﺍ )ﺍﺣﺘﻤﺎﻻ ﺑﺎ ﻳﻚ ﺍﺭﺗﺒﺎﻁ‬
‫ﻣﻮﻗﺖ( ‪ decorrelation‬ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﺩ‪.‬‬
‫• ‪ Decorrelation‬ﭘﻴﭽﻴﺪﻩ ﺗﺮ ﺍﺳﺖ ﺯﻣﺎﻧﻲ ﻛﻪ‪ Sub query :‬ﺍﻱ ﺗﻮ ﺩﺭ ﺗﻮ ﺑﺮﺍﻱ ﺍﺟﺘﻤﺎﻉ ﺍﺳﺘﻔﺎﺩﻩ‬
‫ﺷﻮﺩ‪ ،‬ﻳﺎ ﻫﻨﮕﺎﻣﻲ ﻛﻪ ﻧﺘﻴﺠﻪ ‪sub query‬ﻫﺎﻱ ﺗﻮ ﺩﺭ ﺗﻮﺭ ﺑﺮﺍﻱ ﺗﺴﺖ ﺷﺒﺎﻫﺖ)ﺑﺮﺍﺑﺮﻱ(ﺍﺳﺘﻔﺎﺩﻩ‬
‫ﺷﻮﺩ‪ ،‬ﻳﺎﺯﻣﺎﻧﻲ ﻛﻪ ﺷﺮﺍﻳﻂ ﺍﺭﺗﺒﺎﻁ ‪sub query‬ﻫﺎﻱ ﺗﻮ ﺩﺭ ﺗﻮ ﺑﻪ ‪query‬ﻫﺎﻱ ﺩﻳﮕﺮ ﻭﺟﻮﺩ ﻧﺪﺍﺷﺘﻪ‬
‫ﺑﺎﺷﺪ‪.‬‬

‫‪:‬‬ ‫ﺩﻳﺪﮔﺎﻩ‪materialized‬‬
‫ﺩﻳﺪﮔﺎﻩ‪ ، materialized‬ﺩﻳﺪﮔﺎﻫﻲ ﺍﺳﺖ ﻛﻪ ﻣﺤﺘﻮﻳﺎﺕ ﺁﻥ ﻣﺤﺎﺳﺒﻪ ﻭ ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ ﺍﺳﺖ ‪.‬‬
‫‪create view department_total_salary(dept_name, total_salary)as‬‬
‫)‪select dept_name, sum(salary‬‬
‫‪from instructor‬‬
‫‪group by dept_name‬‬
‫ﺗﺤﻘﻖ ﺩﻳﺪﮔﺎﻩ ﻓﻮﻕ ﺑﺴﻴﺎﺭ ﻣﻔﻴﺪ ﺧﻮﺍﻫﺪ ﺑﻮﺩ ﺍﮔﺮ ‪ total salary‬ﺗﻮﺳﻂ ‪ department‬ﺍﻏﻠﺐ ﻣﻮﺭﺩ‬
‫ﻧﻴﺎﺯ ﺑﺎﺷﺪ ‪ .‬ﺻﺮﻓﻪ ﺟﻮﻳﻲ ﺳﻌﻲ ﺩﺭ ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﺗﺎﭘﻞ ﻫﺎﻱ ﭼﻨﺪﮔﺎﻧﻪ ﻭ ﺍﺿﺎﻓﻪ ﻛﺮﺩﻥ ﻣﻘﺪﺍﺭ ﺑﻪ ﺁﻧﻬﺎ ﻣﻲ‬
‫ﻛﻨﺪ‪.‬‬

‫‪:‬‬ ‫‪Materialized View Maintenance‬‬

‫ﻭﻇﻴﻔﻪ ﻧﮕﻬﺪﺍﺭﻱ ﻳﻚ ﺩﻳﺪﮔﺎﻩ ‪ materialized‬ﺑﺮﻭﺯ ﺭﺳﺎﻧﻲ ﺩﺍﺩﻩ ﺑﺎ ﺩﺍﺩﻩ ﻱ ﺍﺳﺎﺳﻲ ﺍﺳﺖ ﻛﻪ ﺑﻪ‬
‫ﻋﻨﻮﺍﻥ‪ Materialized View Maintenance‬ﺷﻨﺎﺧﺘﻪ ﺷﺪﻩ ﺍﺳﺖ ‪ .‬ﺩﻳﺪﮔﺎﻩ ﻫﺎﻱ‬
‫‪ materialized‬ﻣﻲ ﺗﻮﺍﻥ ﺑﺎ ﻣﺤﺎﺳﺒﻪ ﻣﺠﺪﺩ ﺩﺭ ﻫﺮ ﺑﺮﻭﺯ ﺭﺳﺎﻧﻲ ﻧﮕﻬﺪﺍﺭﻱ ﻛﺮﺩ ‪.‬‬

‫‪181‬‬
‫ﮔﺰﻳﻨﻪ ﺑﻬﺘﺮ ﺍﺳﺖ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ‪incremental view maintenance‬ﺍﺳﺖ‪ .‬ﺗﻐﻴﻴﺮﺍﺕ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ‬
‫ﺑﻪ ﺭﻭﺍﺑﻂ ﺑﺮﺍﻱ ﻣﺤﺎﺳﺒﻪ ﺗﻐﻴﻴﺮﺍﺕ ‪materialized view‬ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺷﻮﺩ ﻛﻪ ﺳﭙﺲ ﺏﺭﻭﺯ ﺭﺳﺎﻧﻲ‬
‫ﻣﻴﺸﻮﻧﺪ‪ View Maintenance .‬ﻣﻲ ﺗﻮﺍﻧﺪ ﺑﺎ ﻣﺤﺮﻙ ﻫﺎﻱ ﺗﻌﺮﻳﻒ ﺩﺳﺘﻲ ﺩﺭ ﺩﺭﺝ‪ ،‬ﺣﺬﻑ‪ ،‬ﻭ ﺑﻪ‬
‫ﺭﻭﺯ ﺭﺳﺎﻧﻲ ﻫﺮ ﺭﺍﺑﻄﻪ ﺩﺭ ‪ view definition‬ﺍﻧﺠﺎﻡ ﺷﻮﺩ ﻭ ﻧﻮﺷﺘﻦ ﺩﺳﺘﻲ ﻛﺪ ﺑﺮﺍﻱ ﺏﺭﻭﺯ ﺭﺳﺎﻧﻲ‬
‫‪view whenever‬ﺭﻭﺍﺑﻂ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﻛﻪ ﺑﻪ ﺭﻭﺯ ﺷﺪﻩ ﺍﺳﺖ‪.‬‬
‫ﺭﻭﺵ ﻫﺎﻱ ﻓﻮﻕ ﺑﻪ ﻃﻮﺭ ﻣﺴﺘﻘﻴﻢ ﺗﻮﺳﻂ ﺑﺴﻴﺎﺭﻱ ﺍﺯ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﭘﺸﺘﻴﺒﺎﻧﻲ ﻣﻴﺸﻮﻧﺪ‪.‬‬

‫ﭘﻴﻮﺳﺘﻦ ﻋﻤﻠﻴﺎﺕ ‪:‬‬

‫‪r‬ﻭ ‪ update‬ﺑﺮﺍﻱ ‪ r‬ﺩﺭ ﻧﻈﺮ ﺑﮕﻴﺮﻳﺪ ‪.‬‬ ‫‪s = materialized view v‬‬ ‫•‬
‫‪new‬‬ ‫‪old‬‬
‫‪r‬ﺗﻔﻜﻴﻚ ﻛﺮﺩﻥ ﻭﺿﻌﻴﺖ ﻫﺎﻱ ‪old‬ﻭ‪ new‬ﺍﺯ ﺭﺍﺑﻄﻪ ‪r‬‬ ‫• ﻭ ‪ Let r‬ﻭ‬
‫• ﺣﺎﻻﺕ ﻳﻚ ﺩﺭﺝ ﺑﺮﺍﻱ ‪ r‬ﺭﺍ ﺩﺭ ﻧﻈﺮ ﺑﮕﻴﺮﻳﺪ‬
‫‪new‬‬ ‫‪old‬‬
‫‪r‬‬ ‫)‪sas (r ∪ir‬‬ ‫• ﻣﺎ ﻣﻲ ﺗﻮﺍﻧﻴﻢ ﺑﻨﻮﻳﺴﻴﻢ ‪s‬‬
‫‪old‬‬
‫‪(r‬‬ ‫• ﻭ ﺑﺎﺯﻧﻮﻳﺴﻲ ﺑﺎﻻ ﺑﺼﻮﺭﺕ)‪s) ∪ (ir s‬‬
‫‪old‬‬
‫‪(r‬ﺳﺎﺩﻩ ﺷﺪﻩ ﺍﺭﺯﺵ ‪old‬ﺍﺯ‪ materialized view‬ﺍﺳﺖ ‪ ،‬ﺑﻨﺎﺑﺮﺍﻳﻦ‬ ‫• ﺍﻣﺎ )‪s‬‬
‫ﺗﻐﻴﻴﺮﺍﺕ‪ incremental‬ﺑﺮﺍﻱ ‪ view‬ﻓﻘﻂ ‪ir s‬ﺍﺳﺖ‪.‬‬
‫‪new‬‬ ‫‪old‬‬
‫‪v‬‬ ‫ﺑﻨﺎﺑﺮﺍﻳﻦ‪ ،‬ﺑﺮﺍﻱ ﺩﺭﺝ ﻫﺎ )‪= v ∪(ir s‬‬ ‫•‬
‫‪new‬‬ ‫‪old‬‬
‫‪v‬‬ ‫• ﺑﻪ ﻃﻮﺭ ﻣﺸﺎﺑﻪ ﺑﻪ ﺣﺬﻑ ﻫﺎ )‪= v –(dr s‬‬

‫ﻋﻤﻠﻴﺎﺕ ‪ Selection‬ﻭ‪: Projection‬‬

‫• ‪Selection‬ﻫﺎ‪ :‬ﺷﺎﻣﻞ ﻳﻚ ‪viewv = σθ(r),‬‬

‫‪new‬‬ ‫‪old‬‬
‫‪v‬‬ ‫• )‪= v ∪σθ(ir‬‬

‫‪182‬‬
 ‫‪new‬‬ ‫‪old‬‬
‫• )‪v = v - σθ(dr‬‬
‫• ‪ projection‬ﻳﻚ ﻋﻤﻠﻴﺎﺕ ﺩﺷﻮﺍﺭ ﺗﺮ ﺍﺳﺖ‬
‫• })‪R = (A,B), and r(R) = { (a,2), (a,3‬‬
‫• )‪∏A(r‬ﻳﻚ )‪ tuple (a‬ﻣﺠﺮﺩ ﺩﺍﺭﺩ‪.‬‬
‫• ﺍﮔﺮ ﻣﺎ )‪ tuple (a,2‬ﺭﺍ ﺍﺯ ‪r‬ﺣﺬﻑ ﻛﻨﻴﻢ‪ ،‬ﻧﺒﺎﻳﺪ )‪ tuple (a‬ﺍﺯ)‪ ∏A(r‬ﺣﺬﻑ ﺷﻮﺩ‬
‫) )‪R‬ﺭﺍ ﺣﺬﻑ ﺍﻣﺎ ﺍﮔﺮ)‪ (a,3‬ﺑﺨﻮﺑﻲ ﺣﺬﻑ ﺷﺪ ﻣﺎ ﺑﺎﻳﺪ ‪ tuple‬ﺭﺍ ﺣﺬﻑ ﻛﻨﻴﻢ ‪.‬‬
‫• ﺑﺮﺍﻱ ﻫﺮ ﺗﺎﭘﻞ ﺩﺭ )‪projection ∏A(r‬ﻣﺎ ﻳﻚ ﺗﻌﺪﺍﺩ ﺍﺯ ﺯﻣﺎﻥ ﻫﺎﻱ ﻣﺸﺘﻖ ﺷﺪﻩ ﺭﺍ ﻧﮕﻪ ﻣﻲ‬
‫ﺩﺍﺭﻳﻢ‪.‬‬
‫• ﺩﺭﺝ ﺩﺭ ﻳﻚ ﺗﺎﭘﻞ ‪ ،r‬ﺍﮔﺮ ﺑﺮﺍﻳﻨﺪﺗﺎﭘﻞ ﺩﺭ )‪ ∏A(r‬ﻭﺟﻮﺩ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ ﻣﺎ ﺍﻓﺰﺍﻳﺶ ﺗﻌﺪﺍﺩ ﺭﺍ‬
‫ﺩﺍﺭﻳﻢ‪ ،‬ﻭﮔﺮﻧﻪ ﻣﺎ ﺍﺿﺎﻓﻪ ﺷﺪﻥ ﻳﻚ ﺗﺎﭘﻞ ﺟﺪﻳﺪ ﺑﺎ‪ count = 1‬ﺭﺍ ﺩﺍﺭﻳﻢ ‪.‬‬
‫• ﺩﺭ ﻣﻮﺭﺩ ﺣﺬﻑ ﻳﻚ ﺗﺎﭘﻞ ﺍﺯ ‪،R‬ﻛﺎﻫﺶ ﻣﺘﻨﺎﻇﺮ ﺗﻌﺪﺍﺩ ﺗﺎﭘﻞ ﺩﺭ )‪ ∏A(r‬ﺩﺍﺭﻳﻢ‪.‬‬
‫• ﺍﮔﺮ ﺗﻌﺪﺍﺩ ‪ 0‬ﺷﻮﺩ‪ ،‬ﻣﺎ ﺗﺎﭘﻞ ﺭﺍ ﺍﺯ)‪ ∏A(r‬ﺣﺬﻑ ﻣﻲ ﻛﻨﻴﻢ‪.‬‬

‫ﻋﻤﻠﻴﺎﺕ ﺍﺩﻏﺎﻡ‪:‬‬

‫)‪(r‬‬
‫• ‪v = Agcount(B) :count‬‬
‫• ﻫﻨﮕﺎﻣﻲ ﻛﻪ ﻳﻚ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺗﺎﭘﻞ ﻫﺎ ‪i‬ﺩﺭﺝ ﻣﻴﺸﻮﻧﺪ‪.‬‬
‫‪r‬‬
‫• ﺑﺮﺍﻱ ﻫﺮ ﺗﺎﭘﻞ ‪r‬ﺩﺭ ‪،i‬ﺍﮔﺮ ﮔﺮﻭﻩ ﻣﺘﻨﺎﻇﺮ ﺣﺎﺿﺮ ﺩﺭ ‪ V‬ﻭﺟﻮﺩ ﺩﺍﺷﺘﻪ ﺑﺎﺷﺪ‪،‬ﻣﺎ‬
‫‪r‬‬
‫ﺍﻓﺰﺍﻳﺶ ‪ count‬ﺭﺍ ﺩﺍﺭﻳﻢ‪ ،‬ﺩﺭ ﻏﻴﺮ ﺍﻳﻨﺼﻮﺭﺕ ﻣﺎ ﻳﻚ ﺗﺎﭘﻞ ﺟﺪﻳﺪ ﺑﺎ ‪count‬‬
‫‪= 1‬ﺍﺿﺎﻓﻪ ﻣﻲ ﻛﻨﻴﻢ‪.‬‬
‫• ﻫﻨﮕﺎﻣﻲ ﻛﻪ ﻳﻚ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺗﺎﭘﻞ ﻫﺎ ‪d‬ﺣﺬﻑ ﻣﻲ ﺷﻮﻧﺪ‪.‬‬
‫‪r‬‬
‫• ﺑﺮﺍﻱ ﻫﺮ ﺗﺎﭘﻞ ‪t‬ﺩﺭ ‪i‬ﻣﺎ ﮔﺮﻭﻩ ‪t.A‬ﺭﺍ ﺩﺭ ‪v‬ﺟﺴﺘﺠﻮ ﻣﻲ ﻛﻨﻴﻢ‪ ،‬ﻭ ‪ 1‬ﺭﺍ ﺍﺯ ‪Count‬‬
‫‪r‬‬
‫ﮔﺮﻭﻩ ﻛﻢ ﻣﻲ ﻛﻨﻴﻢ‪.‬‬
‫• ﺍﮔﺮ ‪ 0 count‬ﺷﻮﺩ‪ ،‬ﺗﺎﭘﻞ ﺑﺮﺍﻱ ﮔﺮﻭﻩ ‪t.A‬ﺍﺯ ‪ v‬ﺭﺍ ﺣﺬﻑ ﻣﻲ‬
‫ﻛﻨﻴﻢ‪.‬‬
‫)‪(r‬‬
‫)‪v = Agsum(B‬‬ ‫• ‪:sum‬‬

‫‪183‬‬
‫• ﻣﺎ ﺑﺮﺍﻱ ﺑﺮﻗﺮﺍﺭﻱ ‪ sum‬ﺩﺭ ﻳﻚ ﺭﻭﺵ ﺳﺎﺩﻩ ‪،count‬ﻣﺎ ﺑﺠﺰ‪ add/substract‬ﺩﺭ‬
‫ﺍﺭﺯﺵ ‪ B‬ﺑﻪ ﺍﺭﺯﺍﻱ‪ adding/substracting‬ﻣﻘﺪﺍﺭ ‪1‬ﺭﺍ ﺑﻪ ‪ count‬ﻣﻲ ﺩﻫﻴﻢ‪.‬‬
‫‪ tupels‬ﺭﺍ ﺣﻔﻆ‬ ‫• ﻋﻼﻭﻩ ﺑﺮ ﺍﻳﻦ ﻣﺎ ‪ count‬ﻣﺮﺗﺒﻪ ﺷﻨﺎﺳﺎﻳﻲ ﮔﺮﻭﻩ ﻫﺎﻱ ﺑﺪﻭﻥ‬
‫ﻣﻴﻜﻨﻴﻢ ‪.‬ﻣﺜﻞ ﮔﺮﻭﻩ ﻫﺎﻱ ‪ deleted‬ﺍﺯ ‪.v‬‬
‫• ﺑﺮﺍﻱ ‪sum‬ﻧﻤﻲ ﺗﻮﺍﻥ ﺑﻪ ﺳﺎﺩﮔﻲ ﺗﺴﺖ ﺍﻧﺠﺎﻡ ﺩﺍﺩ‪) .‬ﭼﺮﺍ؟(‬
‫• ﺑﻜﺎﺭ ﺑﺮﺩﻥ ‪ ،avg‬ﻣﺎ ﻣﻘﺪﺍﺭ ‪sum‬ﻭ ‪ count‬ﺭﺍ ﻧﮕﻪ ﻣﻲ ﺩﺍﺭﻳﻢ‬
‫ﺍﺩﻏﺎﻡ ﺍﺭﺯﺵ ﻫﺎ ﺑﺼﻮﺭﺕ ﺟﺪﺍﮔﺎﻧﻪ ﺍﺳﺖ‪ ،‬ﻭ ﺩﺭ ﭘﺎﻳﺎﻥ ﺗﻘﺴﻴﻢ ﻣﻲ ﻛﻨﻴﻢ‪.‬‬
‫• ‪Max‬ﻭ‪v = Agmin(B) (r) : min‬‬
‫ﺑﻜﺎﺭ ﺑﺮﺩﻥ ‪ inserts‬ﺩﺭ ‪r‬ﺑﺴﻴﺎﺭ ﺳﺎﺩﻩ ﻭ ﺁﺳﺎﻥ ﺍﺳﺖ‪.‬‬ ‫•‬
‫• ﻧﮕﻬﺪﺍﺭﻱ ﻣﺠﻤﻮﻉ ﺍﺭﺯﺵ ﻫﺎﻱ ‪ max‬ﻭ ‪ min‬ﺩﺭ ﺣﺬﻑ ﻣﻤﻜﻦ ﺍﺳﺖ ﭘﺮ ﻫﺰﻳﻨﻪ‬
‫ﺑﺎﺷﺪ‪ .‬ﻣﺎ ﺑﻘﻴﻪ ‪tupel‬ﻫﺎﻱ ‪r‬ﺏ ﺭﺍ ﻛﻪ ﺩﺭ ﮔﺮﻭﻩ ﻣﺸﺎﺑﻪ ﻫﺴﺘﻨﺪ ﺭﺍ ﺟﺴﺘﺠﻮ ﻣﻲ‬
‫ﻛﻨﻴﻢ ﻭ ﻣﻘﺪﺍﺭ ‪ minimum‬ﺟﺪﻳﺪ ﺭﺍ ﭘﻴﺪﺍ ﻣﻲ ﻛﻨﻴﻢ‪.‬‬
‫ﺩﻳﮕﺮ ﻋﻤﻠﻴﺎﺕ ‪:‬‬
‫• ﻣﺠﻤﻮﻋﻪ‪v = r ∩s: intersection‬‬
‫• ﻫﻨﮕﺎﻣﻲ ﻳﻚ ﺗﺎﭘﻞ ﺩﺭ‪ inserted، r‬ﻣﻲ ﺷﻮﺩ ﻛﻪ ﻣﺎ ﭼﻚ ﻣﻲ ﻛﻨﻴﻦ ﺩﺭ ‪ s‬ﺣﺎﺿﺮ‬
‫ﺍﺳﺖ ﺍﮔﺮ ﺣﺎﺿﺮ ﺑﻮﺩ ﺁﻥ ﺭﺍ ﺑﻪ‪ v‬ﺍﺿﺎﻓﻪ ﻣﻲ ﻛﻨﻴﻢ‪.‬‬
‫• ﺍﮔﺮ ﺗﺎﭘﻞ ﺍﺯ ‪ deleted‬ﺣﺬﻑ ﺷﺪ‪ ،‬ﻣﺎ ﺁﻥ ﺭﺍ ﺍﺯ ‪ intersection‬ﺣﺬﻑ ﻛﻨﻴﻢ ﺍﮔﺮ‬
‫ﺁﻥ ﺣﺎﺿﺮ ﺑﻮﺩ‪.‬‬
‫• ‪Update‬ﻫﺎ ﺩﺭ ‪s‬ﻣﺘﻘﺎﺭﻥ ﻫﺴﺘﻨﺪ‪.‬‬
‫• ﺩﻳﮕﺮ ﻣﺠﻤﻮﻋﻪ ﻋﻤﻠﻴﺎﺕ ﻫﺎ‪ union ،‬ﻭ ﻣﺠﻤﻮﻋﻪ ﻭ ‪ defference‬ﻫﺴﺘﻨﺪ ﻛﻪ ﺩﺭ‬
‫ﻳﻚ ﺭﻭﺵ ﻣﺸﺎﺑﻪ ﺑﻜﺎﺭ ﻣﻲ ﺭﻭﻧﺪ ‪.‬‬
‫• ﭘﻴﻮﻧﺪﻫﺎﻱ ﺑﻴﺮﻭﻧﻲ ﺩﺭ ﺭﻭﺵ ﻫﺎﻱ ﻣﺸﺎﺑﻪ ﭘﻴﻮﻧﺪﻫﺎ ﺑﻜﺎﺭ ﻣﻲ ﺭﻭﻧﺪ ﺍﻣﺎ ﺑﺎ ﺗﻌﺪﺍﺩﻱ ﻛﺎﺭ ﺍﺿﺎﻓﻲ ‪.‬‬
‫• ﻣﺎ ﺍﺯﺟﺰﺋﻴﺎﺕ ﺻﺮﻑ ﻧﻈﺮ ﻣﻲ ﻛﻨﻴﻢ‪.‬‬
‫ﻛﺎﺭﺑﺮﺩ ﻋﺒﺎﺭﺍﺕ‪:‬‬

‫ﻫﺮ ﺯﻳﺮ‬ ‫• ﻋﺒﺎﺭﺍﺕ ﺩﺭﺳﺖ ﻭ ﻛﺎﻣﻞ ﺑﻜﺎﺭ ﻣﻲ ﺭﻭﺩ‪ ،‬ﻣﺎ ﺑﺮﺍﻱ ﻣﺤﺎﺳﺒﻪ ﺻﻌﻮﺩﻱ ﻋﺒﺎﺭﺍﺕ ﻣﺸﺘﻖ ﻧﺘﻴﺠﻪ‬
‫ﻋﺒﺎﺭﺍﺕ ﺭﺍ ﺩﺭ ﻧﻈﺮ ﻣﻴﮕﻴﺮﻳﻢ ‪ ،‬ﺍﺯ ﻛﻮﭼﻜﺘﺮﻳﻦ ﺯﻳﺮ ﻋﺒﺎﺭﺕ ﺷﺮﻭﻉ ﻣﻲ ﻛﻨﻴﻢ ‪.‬‬

‫‪184‬‬
‫• ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ‪ E1E2 ،‬ﺭﺍ ﺩﺭ ﻧﻈﺮ ﺑﮕﻴﺮﻳﺪ ﻛﻪ ﺩﺭ ﺁﻥ ﻫﺮ ﻳﻚ ﺍﺯ ‪E1‬ﻭ ‪E2‬ﻣﻤﻜﻦ ﺍﺳﺖ ﻳﻚ ﻋﺒﺎﺭﺕ‬
‫ﭘﻴﭽﻴﺪﻩ ﺑﺎﺷﻨﺪ‪.‬‬
‫• ﻓﺮﺽ ﻛﻨﻴﺪ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺗﺎﭘﻞ ﻫﺎ ‪ inserted‬ﺩﺭ‪ E1‬ﺗﻮﺳﻂ ‪D1‬ﺩﺍﺩﻩ ﻣﻲ ﺷﻮﺩ‪.‬‬
‫• ﻣﺤﺎﺳﺒﻪ ﭘﻴﺸﺘﺮ‪ ،‬ﺍﺯ ﻛﻮﭼﻜﺘﺮﻳﻦ ﺯﻳﺮ ﻋﺒﺎﺭﺍﺕ ﺍﻭﻝ ﺑﻪ ﻛﺎﺭ ﮔﺮﻓﺘﻪ ﺷﺪﻩ‬
‫ﺍﺳﺖ‪.‬‬
‫‪D1‬ﺩﺍﺩﻩ ﻣﻲ ﺷﻮﺩ‬ ‫• ﺳﭙﺲ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺗﺎﭘﻞ ﻫﺎ ‪ inserted‬ﺩﺭ‪E1E2‬ﺗﻮﺳﻂ ‪E2‬‬
‫• ﺍﻳﻦ ﺗﻨﻬﺎ ﻳﻚ ﺭﻭﺵ ﻣﻌﻤﻮﻝ ﻧﮕﻬﺪﺍﺭﻱ ﭘﻴﻮﻧﺪ ﻫﺎ ﺳﺖ ‪.‬‬

‫‪:‬‬ ‫ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ‪query‬ﻭ‪Materialized Views‬‬

‫• ﺑﺎﺯﻧﻮﻳﺴﻲ‪query‬ﻫﺎ ﺑﺮﺍﻱ ﺍﺳﺘﻔﺎﺩﻩ‪: materialized views‬‬
‫‪ v = r‬ﺩﺭ ﺩﺳﺘﺮﺱ ﺍﺳﺖ‪.‬‬ ‫• ﻳﻚ‪s ، materialized views‬‬
‫‪r‬‬ ‫‪s‬‬ ‫• ﻳﻚ ﻛﺎﺭﺑﺮ ‪ submits‬ﻳﻚ ‪t،query‬‬
‫‪v‬‬ ‫• ﻣﺎ ﻣﻲ ﺗﻮﺍﻧﻴﻢ ﺑﺮﺍﻱ ‪ query‬ﺑﺎﺯﻧﻮﻳﺴﻲ ﻛﻨﻴﻦ ﺑﻪ ﻋﻨﻮﺍﻥ‪t‬‬
‫• ﺍﻳﻨﻜﻪ ﺍﻧﺠﺎﻡ ﺍﻳﻦ ﻛﺎﺭ ﺑﺴﺘﮕﻲ ﺑﻪ ﺑﺮﺁﻭﺭﺩ ﻫﺰﻳﻨﻪ ﺩﻭ ﺟﺎﻳﮕﺰﻳﻦ ﺩﺍﺭﺩ ‪.‬‬
‫• ﺟﺎ ﺑﺠﺎﻳﻲ ﻳﻚ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ‪ materialized view‬ﺗﻮﺳﻂ ‪: view definition‬‬
‫‪v = r‬ﺩﺭ ﺩﺳﺘﺮﺱ ﺍﺳﺖ‬ ‫ﻳﻚ‪s ، materialized view‬‬ ‫•‬
‫• ﻛﺎﺭﺑﺮ ‪ submits‬ﻳﻚ )‪queryσA=10(v‬‬
‫• ﻫﻤﭽﻨﻴﻦ ﻓﺮﺽ ﻛﻨﻴﺪ ﻛﻪ ‪ S‬ﻳﻚ ﺷﺎﺧﺺ ﻣﺸﺘﺮﻙ ﺻﻔﺖ ‪b‬ﻭ ‪ r‬ﻳﻚ ﺷﺎﺧﺺ‬
‫ﻣﺸﺘﺮﻙ ﺻﻔﺖ‪ a‬ﺍﺳﺖ‪.‬‬
‫‪،r‬ﻛﻪ ﻣﻲ‬ ‫• ﺑﻬﺘﺮﻳﻦ ﺑﺮﻧﺎﻣﻪ ﺑﺮﺍﻱ ﺍﻳﻦ ‪ query‬ﻣﻤﻜﻦ ﺍﺳﺖ ﺟﺎﺑﺠﺎﻳﻲ ‪V‬ﺗﻮﺳﻂ ‪s‬‬
‫)‪queryσA=10(r‬‬ ‫ﺗﻮﺍﻧﺪ ﻣﻨﺠﺮ ﺑﻪ ﻃﺮﺡ ‪s‬‬
‫• ﺑﻬﻴﻨﻪ ﺳﺎﺯ ‪query‬ﺑﺎﻳﺪ ﺗﻤﺪﻳﺪ ﺷﺪﻩ ﺷﺎﻣﻞ ﺗﻤﺎﻡ ﻣﻮﺍﺭﺩ ﻓﻮﻕ ﻭ ﺑﻬﺘﺮﻳﻦ ﻃﺮﺡ ﻛﻠﻲ ﺭﺍ ﺍﻧﺘﺨﺎﺏ ﻛﻨﺪ ‪.‬‬

‫‪:‬‬ ‫‪Materialized View Selection‬‬

‫‪Materialized Views‬‬ ‫• ‪":Materialized View Selection‬ﺑﻬﺘﺮﻳﻦ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ‬
‫ﭼﻴﺴﺖ؟ "‬

‫‪185‬‬
 ‫• ‪ " :Index selection‬ﺑﻬﺘﺮﻳﻦ ﻣﺠﻤﻮﻋﻪ ﻱ ﺷﺎﺧﺼﻬﺎ ﺑﺮﺍﻱ ﺍﻳﺠﺎﺩ ﭼﻴﺴﺖ؟ "‬
‫• ﻧﺰﺩﻳﻚ ﺗﺮﻳﻦ ﺭﺍﺑﻄﻪ‪materialized view selection ،‬‬
‫• ﺍﻣﺎ ﺳﺎﺩﻩ‬
‫• ‪Materialized view selection‬ﻭ ﺍﻧﺘﺨﺎﺏ ﺷﺎﺧﺺ ﺑﺮ ﺍﺳﺎﺱ ﻧﻮﻉ ﺣﺠﻢ ﻛﺎﺭﺳﻴﺴﺘﻢ )ﻛﻮﺋﺮﻱ ﻫﺎ‬
‫ﻭ ﺑﻪ ﺭﻭﺯ ﺭﺳﺎﻧﻲ ﻫﺎ(‬
‫• ﻫﺪﻑ ﻧﻤﻮﻧﻪ‪ :‬ﺑﻪ ﺣﺪﺍﻗﻞ ﺭﺳﺎﻧﺪﻥ ﺯﻣﺎﻥ ﺣﺠﻢ ﻛﺎﺭ ﺍﺟﺮﺍﻳﻲ‪ ،‬ﻣﻮﺿﻮﻉ ﺑﻪ ﻣﺤﺪﻭﺩﻳﺖ ﺩﺭ‬
‫ﻓﻀﺎ ﻭ ﺯﻣﺎﻥ ﮔﺮﻓﺘﻪ ﺷﺪﻩ ﺑﺮﺍﻱ ﺑﺮﺧﻲ ﺍﺯ ﻛﻮﺋﺮﻱ ﻫﺎﻱ ﺑﺤﺮﺍﻧﻲ ‪ /‬ﺑﻪ ﺭﻭﺯ ﺭﺳﺎﻧﻲ ﻫﺎ‬
‫• ﻳﻜﻲ ﺍﺯ ﻣﺮﺍﺣﻞ ﺩﺭ ﺗﻨﻈﻴﻢ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ‬
‫• ﺑﻴﺸﺘﺮ ﺩﺭ ﻣﻮﺭﺩ ﺗﻨﻈﻴﻢ ﺩﺭ ﻓﺼﻞ ﺑﻌﺪ ﺻﺤﺒﺖ ﻣﻲ ﺷﻮﺩ‬
‫• ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺗﺠﺎﺭﻱ ﺗﻬﻴﻪ ﺍﺑﺰﺍﺭ )ﻧﺎﻣﻴﺪﻩ ﻣﻲ ﺷﻮﺩ "ﺩﺳﺘﻴﺎﺭﺍﻥ ﺗﻨﻈﻴﻢ " ﻭ ﻳﺎ »ﻧﺎﺑﻐﻪ ﻫﺎ "(‬
‫‪Materialized view‬‬ ‫ﺑﺮﺍﻱ ﻛﻤﻚ ﺑﻪ ﻣﺪﻳﺮﻳﺖ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ﺍﻧﺘﺨﺎﺏ ﺷﺪﻩ ﭼﻪ ﺷﺎﺧﺺ ﻫﺎ ﻭ‬
‫ﺍﻳﺠﺎﺩﺷﻮﺩ‪.‬‬

‫‪Top-K Queries:‬‬
‫* ‪select‬‬
‫‪from r, s‬‬
‫‪wherer.B = s.B‬‬
‫‪order byr.Aascending‬‬
‫‪limit 10‬‬
‫ﺟﺎﻳﮕﺰﻳﻦ ‪ :1‬ﺷﺎﺧﺺ ﺣﻠﻘﻪ ﻫﺎﻱ ﺗﻮ ﺩﺭ ﺗﻮ ﭘﻴﻮﻧﺪ ﺑﺎ ‪ r‬ﺑﻪ ﻋﻨﻮﺍﻥ ﺧﺎﺭﺟﻲ‬
‫ﺟﺎﻳﮕﺰﻳﻦ ‪ :2‬ﺑﺮﺁﻭﺭﺩ ﺑﺎﻻﺗﺮﻳﻦ ﺍﺭﺯﺵ ‪r.A‬ﺩﺭ ﻧﺘﻴﺠﻪ ﻭ ﺍﺿﺎﻓﻪ ﻛﺮﺩﻥ )‪selection (and r.A<= H‬ﺩﺭ‬
‫ﺁﻥ ﺑﻨﺪ‪ .‬ﻭ ﺍﮔﺮ ﻧﺘﺎﻳﺞ ﻛﻤﺘﺮ ﺍﺯ‪10‬ﺑﺎﺷﺪ‪ ،‬ﺳﻌﻲ ﻣﺠﺪﺩ ﺑﺎ ‪H‬ﺑﺰﺭﮔﺘﺮ‬

‫ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﺑﺮﻭﺯ ﺭﺳﺎﻧﻲ ﻫﺎ‪:‬‬

‫ﻣﺸﻜﻞ ‪Halloween‬‬

‫‪186‬‬
 ‫‪update R set A = 5 * A‬‬
‫‪where A > 10‬‬
‫ﺍﮔﺮ ﺷﺎﺧﺺ ﺩﺭ ‪ A‬ﺑﺮﺍﻱ ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﺗﺎﭘﻞ ﺭﺿﺎﻳﺖ ‪ A> 10‬ﺍﺳﺘﻔﺎﺩﻩ ﻣﻲ ﺷﻮﺩ‪،‬ﺗﺎﭘﻞ ﻫﺎﻱ ﺑﻼﻓﺎﺻﻠﻪ ﺑﻪ‬
‫ﺭﻭﺯ ﺷﺪﻩ‪ ،‬ﻫﻤﺎﻥ ﺗﺎﭘﻞ ﻣﻤﻜﻦ ﺍﺳﺖ ﭼﻨﺪ ﺑﺎﺭ ﻳﺎﻓﺖ ﺷﻮﺩ ‪.‬‬
‫• ﺭﺍﻩ ﺣﻞ ‪ :1‬ﺑﻪ ﺗﺎﺧﻴﺮ ﺍﻧﺪﺍﺧﺘﻦ ﻫﻤﻴﺸﻪ ﺏﺭﻭﺯ ﺭﺳﺎﻧﻲ ﻫﺎ‬
‫• ﺟﻤﻊ ﺁﻭﺭﻱ ﺏﺭﻭﺯ ﺭﺳﺎﻧﻲ ﻫﺎ )ﺍﺭﺯﺵ ﻫﺎﻱ ﻗﺪﻳﻤﻲ ﻭ ﺟﺪﻳﺪ ﺍﺯ ﺗﺎﭘﻞ ﻫﺎ( ﻭ ﺑﻪ ﺭﻭﺯ ﺭﺳﺎﻧﻲ ﺍﺭﺗﺒﺎﻁ ﻭ‬
‫ﺷﺎﺧﺺ ﺩﺭ ﮔﺬﺭ ﺩﻭﻡ‬
‫• ﺍﺷﻜﺎﻝ‪ :‬ﺳﺮﺑﺎﺭ ﺍﺿﺎﻓﻲ ﺣﺘﻲ ﺍﮔﺮ ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ‪ ،‬ﺑﻪ ﺭﻭﺯ ﺭﺳﺎﻧﻲ ﺗﻨﻬﺎ ﺩﺭ ‪R.B‬ﺍﺳﺖ‪ ،‬ﻧﻪ ﺩﺭ ﻭﻳﮋﮔﻲ ﺩﺭ‬
‫ﺷﺮﺍﻳﻂ ﺍﻧﺘﺨﺎﺏ‪.‬‬
‫• ﺭﺍﻩ ﺣﻞ ‪ :2‬ﺗﺎﺧﻴﺮ ﺗﻨﻬﺎ ﺩﺭ ﺻﻮﺭﺕ ﻟﺰﻭﻡ‬
‫• ﺍﻧﺠﺎﻡ ﺑﻪ ﺭﻭﺯ ﺭﺳﺎﻧﻲ ﻓﻮﺭﻱ ﺍﮔﺮ ﺑﻪ ﺭﻭﺯ ﺭﺳﺎﻧﻲ ﻭﻳﮋﮔﻲ ﻫﺎﻱ ﻛﻪ ﺩﺭ ﺁﻥ ﺑﻨﺪ ﺍﺳﺖ ﺗﺎﺛﻴﺮ ﻧﺪ ﺍﺷﺘﻪ ﺑﺎﺷﺪ‪ ،‬ﻭ‬
‫ﺩﺭﻏﻴﺮ ﺍﻳﻦ ﺻﻮﺭﺕ ﺗﺎﺧﻴﺮ ﺑﻪ ﺭﻭﺯ ﺭﺳﺎﻧﻲ ﻫﺎ ‪.‬‬

‫ﺑﻪ ﺣﺪﺍﻗﻞ ﺭﺳﺎﻧﺪﻥ ﭘﻴﻮﻧﺪ‪:‬‬

‫• ‪Join minimization‬‬
‫‪selectr.A, r.B‬‬
‫‪from r, s‬‬
‫‪wherer.B = s.B‬‬
‫• ﺑﺮﺭﺳﻲ ﻛﻨﻴﺪ ﺍﮔﺮ ﭘﻴﻮﺳﺘﻦ ﺑﻪ ‪ s‬ﺍﺳﺖ‪ ،‬ﺁﻥ ﺣﺬﻑ ﺷﻮﺩ‪.‬‬
‫ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ‪ ،‬ﺷﺮﻁ ﭘﻴﻮﻧﺪ ﻛﻠﻴﺪ ﺧﺎﺭﺟﻲ ﺍﺯ ‪r‬ﺑﻪ‪s‬ﺍﺳﺖ ‪r.B ،‬ﺑﻪ ﻋﻨﻮﺍﻥ ﺗﻬﻲ ﻧﻴﺴﺖ ﺍﻋﻼﻡ ﻛﺮﺩ‪ ،‬ﻭ ﻧﻪ‬
‫ﺍﻧﺘﺨﺎﺏ ﺍﺯ‪.s‬‬
‫ﺳﺎﻳﺮ ﺷﺮﺍﻳﻂ ﻛﺎﻓﻲ ﻣﻤﻜﻦ ﺍﺳﺖ‪:‬‬
‫‪selectr.A, s2.B‬‬
‫‪from r, s as s1, s as s2‬‬
‫‪wherer.B=s1.B andr.B = s2.B and s1.A < 20 and s2.A < 10‬‬
‫‪187‬‬
 ‫ﭘﻴﻮﺳﺘﻦ ﺑﻪ ‪ S1‬ﺯﻳﺎﺩﻱ ﺍﺳﺖ ﻭ ﻣﻲ ﺗﻮﺍﻧﺪ ﺩﻭﺭ ﺍﻧﺪﺍﺧﺘﻪ ﺷﻮﺩ )ﻫﻤﺮﺍﻩ ﺑﺎ ﺍﻧﺘﺨﺎﺏ ﺍﺯ ‪.(S1‬‬
‫• ﺑﺴﻴﺎﺭﻱ ﺍﺯ ﭘﮋﻭﻫﺶ ﻫﺎ ﺩﺭ ﺍﻳﻦ ﻣﻨﻄﻘﻪ ‪ 70s/80s‬ﺍﺳﺖ‪.‬‬

‫‪:‬‬ ‫ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ‪multi query‬‬

‫ﻣﺜﺎﻝ‬
‫‪Q1: select * from (r natural join t) natural join s‬‬
‫‪Q2: select * from (r natural join u) natural join s‬‬
‫ﻫﺮ ﺩﻭ ﻧﻤﺎﻳﺶ ﻛﻮﺋﺰﻱ ﺑﻴﻦ ﺯﻳﺮ ﻋﺒﺎﺭﺍﺕ ﻣﺸﺘﺮﻙ ﺑﻪ ﺍﺷﺘﺮﺍﻙ ﮔﺬﺍﺷﺘﻪ ﻣﻲ ﺷﻮﺩ )‪ r‬ﭘﻴﻮﻧﺪ ﻃﺒﻴﻌﻲﺑﻪ ‪(s‬‬
‫ﻣﻤﻜﻦ ﺍﺳﺖ ﻣﻔﻴﺪ ﺑﺎﺷﺪ ﺑﺮﺍﻱ ﻣﺤﺎﺳﺒﻪ )‪ r‬ﭘﻴﻮﻧﺪ ﻃﺒﻴﻌﻲﺑﻪ ‪ (s‬ﻳﻚ ﺑﺎﺭ ﻭ ﺍﺳﺘﻔﺎﺩﻩ ﺍﺯ ﺁﻥ ﺩﺭ ﻫﺮ ﺩﻭ‬
‫ﻛﻮﺋﺮﻱ‪‌ ،‬ﺍﻣﺎ ﺍﻳﻦ ﻣﻤﻜﻦ ﺍﺳﺖ ﺩﺭ ﺑﺮﺧﻲ ﺷﺮﺍﻳﻂ ﭘﺮ ﻫﺰﻳﻨﻪ ﺷﻮ ﺩ‪ ،‬ﺏﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ‪ r) ،‬ﭘﻴﻮﻧﺪ ﻃﺒﻴﻌﻲﺑﻪ ‪(s‬‬
‫ﻣﻤﻜﻦ ﺍﺳﺖ ﭘﺮ ﻫﺰﻳﻨﻪ ﺷﻮﺩ‪ ،‬ﻗﺼﺪ ﺩﺍﺭﺩ ﻫﻤﺎﻧﻄﻮﺭ ﻛﻪ ﺑﺮﻧﺎﻣﻪ ﻫﺎ ﻧﺸﺎﻥ ﻣﻲ ﺩﻫﺪ ﻣﻤﻜﻦ ﺍﺳﺖ ﺩﺭ ﻛﻮﺋﺮﻱ‬
‫ﻫﺎ ﺍﺭﺯﺍﻧﺘﺮ ﺷﻮﺩ‪.‬‬
‫ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ‪ multi query:‬ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﺑﻬﺘﺮﻳﻦ ﻃﺮﺡ ﻛﻠﻲ ﺑﺮﺍﻱ ﻳﻚ ﻣﺠﻤﻮﻋﻪ ﺍﺯ ﻛﻮﺋﺮﻱ ﻫﺎ‪ ،‬ﺑﻪ‬
‫ﺍﺷﺘﺮﺍﻙ ﮔﺬﺍﺭﻱ ﺯﻳﺮ ﻋﺒﺎﺭﺕ ﻫﺎﻱ ﻣﺸﺘﺮﻙ ﺑﻴﻦ ﻛﻮﺋﺮﻳﻬﺎﻳﻲ ﻛﻪ ﺩﺭ ﺁﻥ ﻣﻔﻴﺪ ﺍﺳﺖ ‪ .‬ﺷﻨﺎﺳﺎﻳﻲ ﻭ ﺑﻬﺮﻩ‬
‫ﺑﺮﺩﺍﺭﻱ ﺍﺯ ‪sub expressions‬ﻫﺎﻱ ﻣﺸﺘﺮﻙ ﺩﺭ ﻃﺮﺡ ﻫﺎﻱ ﻣﻨﺤﺼﺮ ﺑﻪ ﻓﺮﺩ ﻛﻮﺋﺮﻱ ﺑﻬﻴﻨﻪ‪ ،‬ﻛﻪ ﻣﻤﻜﻦ‬
‫ﺍﺳﺖ ﻫﻤﻴﺸﻪ ﺑﻬﺘﺮﻳﻦ ﻃﺮﺡ ﻧﺒﺎﺵﺩ‪ ،‬ﺍﻣﺎ ﭘﻴﺎﺩﻩ ﺳﺎﺯﻱ ﺁﻥ ﺍﺭﺯﺍﻥ ﺍﺳﺖ ‪.‬‬
‫ﺑﻪ ﺍﺷﺘﺮﺍﻙ ﮔﺬﺍﺷﺘﻦ ﺍﺳﻜﻦ ‪ :‬ﻣﻮﺭﺩ ﺧﺎﺹ ﺍﺯ ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ‪ multi query‬ﺍﺳﺖ ﺏﻃﻮﺭ ﮔﺴﺘﺮﺩﻩ ﻣﻮﺭﺩ‬
‫ﺍﺳﺘﻔﺎﺩﻩ ﻗﺮﺍﺭ ﻣﻲ ﮔﻴﺮﺩ‪ .‬ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﻛﻮﺋﺮﻱ ﻫﺎ ﻧﻤﻴﺘﻮﺍﻧﺪ ‪sub expressions‬ﻣﺸﺘﺮﻙ ﺑﻪ ﺍﺷﺘﺮﺍﻙ‬
‫ﻋﻨﻮﺍﻥ ﻳﻚ ﻧﺘﻴﺠﻪ‪ View Maintenance ،‬ﻧﻤﻴﺘﻮﺍﻧﺪ ‪ sub expressions‬ﺭﺍ ﺑﻪ ﺍﺷﺘﺮﺍﻙ‬ ‫ﺑﮕﺬﺍﺭﺩ ﺏ‬
‫ﺑﮕﺬﺍﺭﺩ‪ .‬ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ‪Multi query‬ﻣﻲ ﺗﻮﺍﻧﺪ ﺩﺭ ﭼﻨﻴﻦ ﺷﺮﺍﻳﻄﻲ ﻣﻔﻴﺪ ﺑﺎﺷﺪ‬

‫‪:‬‬ ‫ﭘﺎﺭﺍﻣﺘﺮ ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ‪query‬‬

‫ﺑﻪ ﻋﻨﻮﺍﻥ ﻣﺜﺎﻝ‬

‫‪188‬‬
 ‫* ‪select‬‬
‫‪from r natural join s‬‬
‫‪.a< $1 where‬‬
‫ﻣﻘﺪﺍﺭ ﭘﺎﺭﺍﻣﺘﺮ ‪ $1‬ﺩﺭ ﺯﻣﺎﻥ ﻛﺎﻣﭙﺎﻳﻞ ﻣﺸﺨﺺ ﺷﺪﻩ ﻧﻴﺴﺖ‬
‫• ﺗﻨﻬﺎ ﺩﺭ ﺯﻣﺎﻥ ﺍﺟﺮﺍ ﻣﺸﺨﺺ ﻣﻲ ﺷﻮﺩ‬
‫• ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﻣﺨﺘﻠﻒ ﻣﻤﻜﻦ ﺍﺳﺖ ﺑﺮﺍﻱ ﻣﻘﺎﺩﻳﺮ ﻣﺨﺘﻠﻒ ﺍﺯ ‪ $1‬ﻣﻄﻠﻮﺏ ﺑﺎﺷﺪ‬
‫• ﺭﺍﻩ ﺣﻞ ‪ :1‬ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﺩﺭ ﺯﻣﺎﻥ ﺍﺟﺮﺍ‪ ،‬ﻫﺮ ﻛﻮﺋﺮﻱ ﺯﻣﺎﻥ ﺭﺍ ﻣﺸﺎﻫﺪﻩ ﻛﻨﻴﺪ‬
‫ﻣﻲ ﺗﻮﺍﻧﺪ ﭘﺮ ﻫﺰﻳﻨﻪ ﺑﺎﺷﺪ‬ ‫•‬
‫• ﺭﺍﻩ ﺣﻞ ‪ :2‬ﭘﺎﺭﺍﻣﺘﺮﻱ ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﻛﻮﺋﺮﻱ‪:‬‬
‫• ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ ﺗﻮﻟﻴﺪ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺑﺮﻧﺎﻣﻪ ﻫﺎ‪ ،‬ﻣﻨﺎﺳﺐ ﺑﺮﺍﻱ ﻣﻘﺎﺩﻳﺮ ﻣﺨﺘﻠﻒ ﺍﺯ ‪$1‬‬
‫• ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﺑﻬﻴﻨﻪ ﻣﻌﻤﻮﻻ ﺑﺮﺍﻱ ‪ 1‬ﺗﺎ ‪ 3‬ﭘﺎﺭﺍﻣﺘﺮ ﻛﻮﭼﻚ‬
‫• ﻣﻮﺿﻮﻉ ﻛﻠﻴﺪﻱ‪ :‬ﭼﮕﻮﻧﻪ ﭘﻴﺪﺍ ﻛﺮﺩﻥ ﻣﺠﻤﻮﻋﻪ ﺍﻱ ﺍﺯ ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﺑﻬﻴﻨﻪ ﻣﻮﺛﺮ‬
‫• ﻳﻜﻲ ﺍﺯ ﺑﻬﺘﺮﻳﻦ ﺍﻳﻦ ﻣﺠﻤﻮﻋﻪ ﻫﺎ ﺍﻳﻦ ﺍﺳﺖ ﻛﻪ‪ $1‬ﻣﺸﺨﺺ ﺩﺭ ﺯﻣﺎﻥ ﺍﺟﺮﺍ ﺍﻧﺘﺨﺎﺏ ﺷﻮﺩ‬
‫• ﺭﺍﻩ ﺣﻞ ‪ :3‬ﺫﺧﻴﺮﻩ ﺳﺎﺯﻱ ﻃﺮﺡ ﻛﻮﺋﺮﻱ‬
‫• ﺍﮔﺮ ﺑﻬﻴﻨﻪ ﺳﺎﺯ ﺗﺼﻤﻴﻢ ﻣﻲ ﮔﻴﺮﺩ ﻛﻪ ﻃﺮﺡ ﻫﻤﺎﻥ ﺑﺎﺷﺪ ﺑﻪ ﺍﺣﺘﻤﺎﻝ ﺯﻳﺎﺩ ﺑﻪ ﺑﻬﻴﻨﻪ ﻛﺮﺩﻥ ﺑﺮﺍﻱ‬
‫ﺗﻤﺎﻡ ﻣﻘﺎﺩﻳﺮ ﭘﺎﺭﺍﻣﺘﺮ ﻣﻲ ﭘﺮﺩﺍﺯﺩ‪ ،‬ﺁﻥ ﻃﺮﺡ ﺭﺍ ﺫﺧﻴﺮﻩ ﻭ ﺩﻭﺑﺎﺭﻩ ﺑﺎﺯﻳﺎﺑﻲ ﻣﻴﻜﻨﺪ‪،‬ﻭﮔﺮﻧﻪ ﺩﻭﺑﺎﺭﻩ ﺑﻬﻴﻨﻪ‬
‫ﺳﺎﺯﻱ ﻫﺮ ﺯﻣﺎﻥ ﺍﻧﺠﺎﻡ ﻣﻴﺸﻮﺩ‪.‬‬
‫• ﺍﺟﺮﺍ ﺩﺭ ﺑﺴﻴﺎﺭﻱ ﺍﺯ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ‬

‫ﺛﺒﺎﺕ ﺑﺮﻧﺎﻣﻪ ﺩﺭ ﺳﺮﺍﺳﺮ ﺗﻐﻴﻴﺮﺍﺕ ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ‪:‬‬

‫• ﭼﻪ ﻣﻲ ﺷﻮﺩ ﺍﮔﺮ ‪ ٪95‬ﺍﺯ ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﺳﺮﻳﻊ ﺩﺭ ﭘﺎﻳﮕﺎﻩ ﺩﺍﺩﻩ ‪ /‬ﺑﻬﻴﻨﻪ ﺳﺎﺯ ﻧﺴﺨﻪ ‪ N +1‬ﻧﺴﺒﺖ ﺑﻪ‬
‫‪N‬ﺑﺎﺷﻨﺪ‪،‬ﺍﻣﺎ ‪ 5‬ﺩﺭﺻﺪ ﻛﻨﺪﺗﺮ ﻫﺴﺘﻨﺪ؟‬
‫• ﭼﺮﺍ ﺑﺎﻳﺪ ﺑﺮﻧﺎﻣﻪ ﻫﺎﻱ ﻛﻨﺪﺗﺮ ﺑﻬﻴﻨﻪ ﺳﺎﺯ ﺟﺪﻳﺪ ﻭ ﺑﻬﺒﻮﺩ ﻳﺎﺑﻨﺪ؟‬
‫• ﺟﻮﺍﺏ‪ :‬ﺩﻭ ﺍﺷﺘﺒﺎﻫﺎﺕ ﻣﻲ ﺗﻮﺍﻧﺪ ﻳﻚ ﺩﺭﺳﺖ ﺭﺍ ﺑﺴﺎﺯﺩ‪،‬ﺩﺭﺳﺖ ﻛﺮﺩﻥ ﻳﻚ ﺍﺷﺘﺒﺎﻩ ﻣﻲ‬
‫ﺗﻮﺍﻧﺪ ﺍﻭﺿﺎﻉ ﺭﺍ ﺑﺪﺗﺮ!‬
‫• ﺭﻭﺵ ﻫﺎ‪:‬‬
‫‪189‬‬
 ‫• ﻧﻜﺎﺕ ﺑﻪ ﺗﻨﻈﻴﻢ ﻛﻮﺋﺮﻱ ﻫﺎ ﺍﺟﺎﺯﻩ ﻣﻴﺪﻫﺪ‬
‫• ﻣﻬﺎﺟﺮﺕ ﺑﺮﺍﻱ ﺳﻴﺴﺘﻢ ﻫﺎﻱ ﺑﺰﺭگ ﺑﺪﻭﻥ ﺩﺳﺘﺮﺳﻲ ﺑﻪ ﻛﺪ ﻣﻨﺒﻊ ﻛﺎﺭﺑﺮﺩﻱ ﻧﻴﺴﺖ ‪.‬‬
‫• ﺗﻨﻈﻴﻢ ﺳﻄﺢ ﺑﻬﻴﻨﻪ ﺳﺎﺯﻱ‪ ،‬ﺑﻪ ﻃﻮﺭ ﭘﻴﺶ ﻓﺮﺽ ﻧﺴﺨﻪ ‪) N‬ﺍﻭﺭﺍﻛﻞ(‬
‫• ﻭ ﻣﻬﺎﺟﺮﺕ ﻳﻚ ﻛﻮﺋﺮﻱ ﺩﺭ ﺯﻣﺎﻥ ﭘﺲ ﺍﺯ ﺁﺯﻣﺎﻳﺶ ﻫﺮ ﺩﻭ ﻃﺮﺡ ﺩﺭ ﺑﻬﻴﻨﻪ ﺳﺎﺯ ﺟﺪﻳﺪ‬
‫• ﺻﺮﻓﻪ ﺟﻮﻳﻲ ﺩﺭ ﻃﺮﺡ ﻧﺴﺨﻪ ‪N‬ﻭ ﺩﺍﺩﻥ ﻧﺴﺨﻪ ﺑﻬﻴﻨﻪ ‪N +1‬‬
‫• ‪XML , Sybase‬ﻧﻤﺎﻳﻨﺪﻩ ﻃﺮﺡ ﻫﺎ ﻫﺴﺘﻨﺪ)‪SQL‬ﺳﺮﻭﺭ(‬

‫‪190‬‬
191
192
193