Unit 5
Unit 5
UNIT V
IMPLEMENTATION TECHNIQUES
RAID
Databases are stored in file formats, which contain records. At physical level, the actual data is s
● Primary Storage
The memory storage that is directly accessible to the CPU comes under this category
● Secondary Storage
Secondary storage devices are used to store data for future use or as backup. Seco
● Tertiary Storage
Tertiary storage is used to store huge volumes of data. Since such sto
● RAID 1
RAID 1 uses mirroring techniques. When data is sent to a RAID controller,
it sends a copy of data to all the disks in the array. RAID level 1 is also
called mirroring and provides 100% redundancy in case of a failure.
● RAID 2
RAID 2 records Error Correction Code using Hamming distance for its data,
striped on different disks. Like level 0, each data bit in a word is recorded on
a separate disk and ECC codes of the data words are stored on a different set
disks. Due to its complex structure and high cost, RAID 2 is not
commercially available.
● RAID 3
RAID 3 stripes the data onto multiple disks. The parity bit generated for data
word is stored on a different disk. This technique makes it to overcome
single disk failures.
● RAID 4
In this level, an entire block of data is written onto data disks and then the
parity is generated and stored on a different disk. Note that level 3 uses byte-
level striping, whereas level 4 uses block-level striping. Both level 3 and
level 4 require at least three disks to implement RAID.
● RAID 5
RAID 5 writes whole data blocks onto different disks, but the parity bits
generated for data block stripe are distributed among all the data disks rather
than storing them on a different dedicated disk.
● RAID 6
RAID 6 is an extension of level 5. In this level, two independent parities are
generated and stored in distributed fashion among multiple disks. Two
parities provide additional fault tolerance. This level requires at least four
disk drives to implement RAID.
ChoiceofRAIDLevel
Factors
inchoosingRAIDlevel
Monetarycost
⮚ Performance: Number ofI/O operations per second, and bandwidth
duringnormal operation
⮚ Performance duringfailure
⮚ Level3isnotusedanymoresincebit-
stripingforcessingleblockreadstoaccessalldisks, wastingdisk arm
movement, which block striping (level 5) avoids.
oLevel 6 is rarelyusedsincelevels 1 and 5 offeradequate safetyfor almost
allapplications o So competition is between 1 and 5 only
Level 1 provides muchbetter write performance than level 5
oLevel 5 requires at least 2 block readsand 2 block writes to writea single block,
whereasLevel 1 onlyrequires 2 block writes.
oLevel 1 preferred for high update environmentssuch as logdisks
Level 1 had higher storage cost than level 5
odiskdrivecapacitiesincreasingrapidly(50%/year)whereasdiskaccesstimeshavede
creased much less (x3 in 10years)
oI/Orequirements haveincreasedgreatly,E.g. for Web servers
OWhenenoughdiskshavebeenboughttosatisfyrequiredrateofI/O,theyoftenhavespa
re storage capacity
sthere is often no extra monetary cost for Level 1!
o
Level 5 is preferred for applications with low update rate, and large
amounts of data Level 1 is preferred for all other applications.
File Organization
● The File is a collection of records. Using the primary key, we can access the
records. The type and frequency of access can be determined by the type of
file organization which was used for a given set of records.
● File organization is a logical relationship among various records. This
method defines how file records are mapped onto disk blocks.
● File organization is used to describe the way in which the records are stored
in terms of blocks, and the blocks are placed on the storage medium.
● The first approach to map the database to the file is to use the several files
and store only one fixed length record in any given file. An alternative
approach is to structure our files so that we can contain multiple lengths for
records.
● Files of fixed length records are easier to implement than the files of variable
length records.
File organization contains various methods. These particular methods have pros
and cons on the basis of access or selection. In the file organization, the
programmer decides the best-suited file organization method according to his
requirement.
Types of file organization are as follows:
Storing the files in certain order is called file organization. The main objective of
file organization is
There are various methods of file organizations. These methods may be efficient
for certain types of access/selection meanwhile it will turn inefficient for other
selections. Hence it is up to the programmer to decide the best suited file
organization method depending on his requirement.
This method is the easiest method for file organization. In this method, files are
stored sequentially. This method can be implemented in two ways:
1. Pile File Method:
● In this method, the new record is always inserted at the file's end, and then it
will sort the sequence in ascending or descending order. Sorting of records is
based on any primary key or any other key.
● In the case of modification of any record, it will update the record and then
sort the file, and lastly, the updated record is placed in the right place.
● It contains a fast and efficient method for the huge amount of data.
● In this method, files can be easily stored in cheaper storage mechanism like
magnetic tapes.
● It is simple in design. It requires no much effort to store the data.
● This method is used when most of the records have to be accessed like grade
calculation of a student, generating the salary slip, etc.
● This method is used for report generation or statistical calculations.
● It is the simplest and most basic type of organization. It works with data
blocks. In heap file organization, the records are inserted at the file's end.
When the records are inserted, it doesn't require the sorting and ordering of
records.
● When the data block is full, the new record is stored in some other block.
This new data block need not to be the very next data block, but it can select
any data block in the memory to store new records. The heap file is also
known as an unordered file.
● In the file, every record has a unique id, and every page in a file is of the
same size. It is the DBMS responsibility to store and manage the new
records.
Insertion of a new record
Suppose we have five records R1, R3, R6, R4 and R5 in a heap and suppose we
want to insert a new record R2 in a heap. If the data block 3 is full then it will be
inserted in any of the database selected by the DBMS, let's say data block 1.
If we want to search, update or delete the data in heap file organization, then we
need to traverse the data from staring of the file till we get the requested record.
If the database is very large then searching, updating or deleting of record will be
time-consuming because there is no sorting or ordering of records. In the heap file
organization, we need to check all the data until we get the requested record.
● This method is inefficient for the large database because it takes time to
search or modify the record.
● This method is inefficient for large databases.
When a record has to be received using the hash key columns, then the address is
generated, and the whole record is retrieved using that address. In the same way,
when a new record has to be inserted, then the address is generated using the hash
key and record is directly inserted. The same process is applied in the case of
delete and update.
In this method, there is no effort for searching and sorting the entire file. In this
method, each record will be stored randomly in the memory.
B+ File Organization
● In this method, searching any record can be traversed through the single path
and accessed easily.
● In this method, searching becomes very easy as all the records are stored
only in the leaf nodes and sorted the sequential linked list.
● Traversing through the tree structure is easier and faster.
● The size of the B+ tree has no restrictions, so the number of records can
increase or decrease and the B+ tree structure can also grow or shrink.
● It is a balanced tree structure, and any insert/update/delete does not affect
the performance of tree.
If any record has to be retrieved based on its index value, then the address of the
data block is fetched and the record is retrieved from the memory.
Pros of ISAM:
● In this method, each record has the address of its data block, searching a
record in a huge database is quick and easy.
● This method supports range retrieval and partial retrieval of records. Since
the index is based on the primary key values, we can retrieve the data for the
given range of value. In the same way, the partial value can also be easily
searched, i.e., the student name starting with 'JA' can be easily searched.
Cons of ISAM
● This method requires extra space in the disk to store the index value.
● When the new records are inserted, then these files have to be reconstructed
to maintain the sequence.
● When the record is deleted, then the space used by it needs to be released.
Otherwise, the performance of the database will slow down.
● When the two or more records are stored in the same file, it is known as
clusters. These files will have two or more tables in the same data block, and
key attributes which are used to map these tables together are stored only
once.
● This method reduces the cost of searching for various records in different
files.
● The cluster file organization is used when there is a frequent need for joining
the tables with the same condition. These joins will give only a few records
from both tables. In the given example, we are retrieving the record for only
particular departments. This method can't be used to retrieve the record for
the entire department.
In this method, we can directly insert, update or delete any record. Data is sorted
based on the key with which searching is done. Cluster key is a type of key with
which joining of the table is performed.
In indexed cluster, records are grouped based on the cluster key and stored
together. The above EMPLOYEE and DEPARTMENT relationship is an example
of an indexed cluster. Here, all the records are grouped based on the cluster key-
DEP_ID and all the records are grouped.
2. Hash Clusters:
It is similar to the indexed cluster. In hash cluster, instead of storing the records
based on the cluster key, we generate the value of the hash key for the cluster key
and store the records with the same hash key value.
● The cluster file organization is used when there is a frequent request for
joining the tables with same joining condition.
● It provides the efficient result when there is a 1:M mapping between the
tables.
● This method has the low performance for the very large database.
● If there is any change in joining condition, then this method cannot use. If
we change the condition of joining then traversing the file takes a lot of
time.
● This method is not suitable for a table with a 1:1 condition.
Indexing in DBMS
● The first column of the database is the search key that contains a copy of the
primary key or candidate key of the table. The values of the primary key are
stored in sorted order so that the corresponding data can be accessed easily.
● The second column of the database is the data reference. It contains a set of
pointers holding the address of the disk block where the value of the
particular key can be found.
Indexing Methods
Indexing Methods
Seco
Orde Clust ndary
Primar ering Index
red
y Index
Indices Index
Ordered indices
The indices are usually sorted to make searching faster. The indices which are
sorted are known as ordered indices.
Example: Suppose we have an employee table with thousands of record and each
of which is 10 bytes long. If their IDs start with 1, 2, 3 and so on and we have to
search student with ID-543.
● In the case of a database with no index, we have to search the disk block
from starting till it reaches 543. The DBMS will read the record after
reading 543*10=5430 bytes.
● In the case of an index, we will search using indexes and the DBMS will
read the record after reading 542*2= 1084 bytes which are very less
compared to the previous case.
Primary Index
● If the index is created on the basis of the primary key of the table, then it is
known as primary indexing. These primary keys are unique to each record
and contain 1:1 relation between the records.
● As primary keys are stored in sorted order, the performance of the searching
operation is quite efficient.
● The primary index can be classified into two types: Dense index and Sparse
index.
Dense index
● The dense index contains an index record for every search key value in the
data file. It makes searching faster.
● In this, the number of records in the index table is same as the number of
records in the main table.
● It needs more space to store index record itself. The index records have the
search key and a pointer to the actual record on the disk.
Sparse index
● In the data file, index record appears only for a few items. Each item points
to a block.
● In this, instead of pointing to each record in the main table, the index points
to the records in the main table in a gap.
Clustering Index
In the sparse indexing, as the size of the table grows, the size of mapping also
grows. These mappings are usually kept in the primary memory so that address
fetch should be faster. Then the secondary memory searches the actual data based
on the address got from mapping. If the mapping size grows then fetching the
address itself becomes slower. In this case, the sparse index will not be efficient.
To overcome this problem, secondary indexing is introduced.
For example:
● If you want to find the record of roll 111 in the diagram, then it will search
the highest entry which is smaller than or equal to 111 in the first level
index. It will get 100 at this level.
● Then in the second index level, again it does max (111) <= 111 and gets
110. Now using the address 110, it goes to the data block and starts
searching each record till it gets 111.
● This is how a search is performed in this method. Inserting, updating or
deleting is also done in the same manner.
B++ TREE
B+ Tree
Structure of B+ Tree
● In the B+ tree, every leaf node is at equal distance from the root node. The
B+ tree is of the order n where n is fixed for every B+ tree.
● It contains an internal node and leaf node.
Internal node
● An internal node of the B+ tree can contain at least n/2 record pointers
except the root node.
● At most, an internal node of the tree contains n pointers.
Leaf node
● The leaf node of the B+ tree can contain at least n/2 record pointers and n/2
key values.
● At most, a leaf node contains n record pointer and n key values.
● Every leaf node of the B+ tree contains one block pointer P to point to next
leaf node.
Searching a record in B+ Tree
Suppose we have to search 55 in the below B+ tree structure. First, we will fetch
for the intermediary node which will direct to the leaf node that can contain a
record for 55.
So, in the intermediary node, we will find a branch between 50 and 75 nodes. Then
at the end, we will be redirected to the third leaf node. Here DBMS will perform a
sequential search to find 55.
B+ Tree Insertion
Suppose we want to insert a record 60 in the below structure. It will go to the 3rd
leaf node after 55. It is a balanced tree, and a leaf node of this tree is already full,
so we cannot insert 60 there.
In this case, we have to split the leaf node, so that it can be inserted into tree
without affecting the fill factor, balance and order.
The 3rd leaf node has the values (50, 55, 60, 65, 70) and its current root node is 50.
We will split the leaf node of the tree in the middle so that its balance is not
altered. So we can group (50, 55) and (60, 65, 70) into 2 leaf nodes.
If these two has to be leaf nodes, the intermediate node cannot branch from 50. It
should have 60 added to it, and then we can have pointers to a new leaf node.
This is how we can insert an entry when there is overflow. In a normal scenario, it
is very easy to find the node where it fits and then place it in that leaf node.
B+ Tree Deletion
Suppose we want to delete 60 from the above example. In this case, we have to
remove 60 from the intermediate node as well as from the 4th leaf node too. If we
remove it from the intermediate node, then the tree will not satisfy the rule of the
B+ tree. So we need to modify it to have a balanced tree.
After deleting node 60 from above B+ tree and re-arranging the nodes, it will show
as following B+ index files
Hashing
In a huge database structure, it is very inefficient to search all the index values and
reach the desired data. Hashing technique is used to calculate the direct location of
a data record on the disk without using index structure.
In this technique, data is stored at the data blocks whose address is generated by
using the hashing function. The memory location where these records are stored is
known as data bucket or data blocks.
In this, a hash function can choose any of the column value to generate the address.
Most of the time, the hash function uses the primary key to generate the address of
the data block. A hash function is a simple mathematical function to any complex
mathematical function. We can even consider the primary key itself as the address
of the data block. That means each row whose address will be the same as a
primary key stored in the data block.
The above diagram shows data block addresses same as primary key value. This
hash function can also be a simple mathematical function like exponential, mod,
cos, sin, etc. Suppose we have mod (5) hash function to determine the address of
the data block. In this case, it applies mod (5) hash function on the primary keys
and generates 3, 3, 1, 4 and 2 respectively, and records are stored in those data
block addresses.
Static Hashing
In static hashing, the resultant data bucket address will always be the same. That
means if we generate an address for EMP_ID =103 using the hash function mod
(5) then it will always result in same bucket address 3. Here, there will be no
change in the bucket address.
Hence in this static hashing, the number of data buckets in memory remains
constant throughout. In this example, we will have five data buckets in the memory
used to store the data.
● Searching a record
When a record needs to be searched, then the same hash function retrieves the
address of the bucket where the data is stored.
● Insert a Record
When a new record is inserted into the table, then we will generate an address for a
new record based on the hash key and record is stored in that location.
● Delete a Record
To delete a record, we will first fetch the record which is supposed to be deleted.
Then we will delete the records for that address in memory.
● Update a Record
To update a record, we will first search it using a hash function, and then the data
record is updated.
If we want to insert some new record into the file but the address of a data bucket
generated by the hash function is not empty, or data already exists in that address.
This situation in the static hashing is known as bucket overflow. This is a critical
situation in this method.
To overcome this situation, there are various methods. Some commonly used
methods are as follows:
1. Open Hashing
When a hash function generates an address at which data is already stored, then the
next bucket will be allocated to it. This mechanism is called as Linear Probing.
For example: suppose R3 is a new address which needs to be inserted, the hash
function generates address as 112 for R3. But the generated address is already full.
So the system searches next available data bucket, 113 and assigns R3 to it.
2. Close Hashing
When buckets are full, then a new data bucket is allocated for the same hash result
and is linked after the previous one. This mechanism is known as Overflow
chaining.
For example: Suppose R3 is a new address which needs to be inserted into the
table, the hash function generates address as 110 for it. But this bucket is full to
store the new data. In this case, a new bucket is inserted at the end of 110 buckets
and is linked to it.
Dynamic Hashing
Firstly, you have to follow the same procedure for retrieval, ending up in
some bucket.
If there is still space in that bucket, then place the record in it.
If the bucket is full, then we will split the bucket and redistribute the records.
For example:
Consider the following grouping of keys into buckets, depending on the prefix of
their hash address:
The last two bits of 2 and 4 are 00. So it will go into bucket B0. The last two bits
of 5 and 6 are 01, so it will go into bucket B1. The last two bits of 1 and 3 are 10,
so it will go into bucket B2. The last two bits of 7 are 11, so it will go into B3.
Insert key 9 with hash address 10001 into the above structure:
Since key 9 has hash address 10001, it must go into the first bucket. But
bucket B1 is full, so it will get split.
The splitting will separate 5, 9 from 6 since last three bits of 5, 9 are 001, so
it will go into bucket B1, and the last three bits of 6 are 101, so it will go
into bucket B5.
Keys 2 and 4 are still in B0. The record in B0 pointed by the 000 and 100
entry because last two bits of both the entry are 00.
Keys 1 and 3 are still in B2. The record in B2 pointed by the 010 and 110
entry because last two bits of both the entry are 10.
Key 7 are still in B3. The record in B3 pointed by the 111 and 011 entry
because last two bits of both the entry are 11.
Advantages of dynamic hashing
In this method, the performance does not decrease as the data grows in the
system. It simply increases the size of memory to accommodate the data.
In this method, memory is well utilized as it grows and shrinks with the data.
There will not be any unused memory lying.
This method is good for the dynamic database where data grows and shrinks
frequently.
In this method, if the data size increases then the bucket size is also
increased. These addresses of data will be maintained in the bucket address
table. This is because the data address will keep changing as buckets grow
and shrink. If there is a huge increase in data, maintaining the bucket address
table becomes tedious.
In this case, the bucket overflow situation will also occur. But it might take
little time to reach this situation than static hashing.
Query Processing Overview
WHERE DNO=5);
S4. Using a primary index to retrieve multiple records: If the comparison condition
is >;_; <;_ on a key field with a primary index { for example, DNUMBER >5 in
OP2- use the index to find the record satisfying the equality condition(DNUMBER
= 5), then retrieve all subsequent (preceding) records in the (or-dered) file.
S8. Conjunctive selection using a composite index: If two or more attributes are
involved in equality conditions and a composite index exists on the combined
fields, we can use the index directly. For example, OP5 can use this method if
there is a composite index (ESSN; PNO) for WORKS ON file.
S9. Conjunctive selection by intersection of record pointers: If secondary indexes
are available on more than one of the fields in the conjunctive condition, and if the
indexes include record pointers (rather than block pointers). The intersection of
these sets of record pointers would satisfy the conjunctive condition, which are
then used to retrieve those records directly. If only some of the simple conditions
have secondary indexes, each retrieved record is further tested for the remaining
conditions.
Implementing the JOIN Operation
Two operations for demonstration:
(OP6): EMPLOY EE. / DNO=DNUMBER DEPARTMENT
(OP7): DEPARTMENT. / MGRSSN=SSN EMPLOY EE
Methods for implementing JOINs:
Query Plans: A query plan (or query execution plan) is an ordered set of steps
used to access data in a SQL relational database management system.
Query Optimization is the process of selecting the most efficient Query evaluation
plan from among many.
Initially, the SQL query is scanned. Then it is parsed to look for syntactical errors
and correctness of data types. If the query passes this step, the query is
decomposed into smaller query blocks. Each block is then translated to equivalent
relational algebra expression.
Query optimization involves three steps, namely query tree generation, plan
generation, and query plan code generation.
During execution, an internal node is executed whenever its operand tables are
available. The node is then replaced by the result table. This process continues for
all internal nodes until the root node is executed and replaced by the result table.
For example, let us consider the following schemas EMPLOYEE
DEPARTMENT
D DNa Locatio
No me n
Example 1
πEmpID(σEName="ArunKumar"(EMPLOYEE))
Example 2
πEName,Salary(σDName="Marketing"(DEPARTMENT))⋈DNo=DeptNo(EMPLO YEE)
After the query tree is generated, a query plan is made. A query plan is an extended
query tree that includes access paths for all operations in the query tree. Access
paths specify how the relational operations in the tree should be performed. For
example, a selection operation can have an access path that gives details about the
use of B+ tree index for selection.
Besides, a query plan also states how the intermediate tables should be passed from
one operator to the next, how temporary tables should be used and how operations
should be pipelined/combined.
Code generation is the final step in query optimization. It is the executable form of
the query, whose form depends upon the type of the underlying operating system.
Once the query code is generated, the Execution Manager runs it and produces the
results.
Terminologies:
QUERY OPTIMIZATION
Query optimization is the process of selecting the most efficient query-evaluation
plan from among the many strategies usually possible for processing a given query.
As there are many equivalent transformations of same high-level query, aim of
Query Optimization is to choose one that minimizes resource usage. Generally,
Query optimization aims to reduce total execution time of query. Problem
computationally intractable with large number of relations, so strategy adopted
is reduced to finding near optimum solution. There are two query optimizations
are available such as,
1. Heuristic or Rule based Optimizations
2. Systematic or Cost based Optimizations
1. Heuristic Query Optimization Oracle and IBM DB2 use Rule Based
optimization.
A query can be represented as a tree data structure. Operations are at the interior
nodes and data items (tables, columns) are at the leaves. The query is evaluated in
a depth-first pattern.
Example: SELECT PNUMBER, DNUM, LNAME FROM PROJECT,
DEPARTMENT, EMPLOYEE WHERE DNUM=DNUMBER and
MGREMPID=EMPID and PLOCATION = 'Stafford';
Normal Execution Plan:
Optimized Query Execution: These transformations can be used in various
combinations to optimize queries. Some general steps follow:
[1] Using rule 1, break up conjunctive selection conditions and chain them together.
[2] Using the commutatively rules, move the selection operations as far down
the tree as possible.
[3] Using the associativity rules, rearrange the leaf nodes so that the most
restrictive selection conditions are executed first.
[4] Combine Cartesian product operations with associated selection conditions
to form a single Join operation.
[5] Using the Commutatively of projection rules, move the projection
operations down the tree to reduce the sizes of intermediate result sets.
[6] Finally, identify sub trees that can be executed using a single efficient
access method.
2. Systematic or Cost based Optimizations Cost Estimation for RA Operations is
done and the best execution plan which incurs lowest cost is chosen.
The cost of query evaluation can be measured in terms of a number of different
resources, including disk accesses, CPU time to execute a query, and, in a
distributed or parallel database system, the cost of communication The number of
block transfers from disk and the number of disk seeks to estimate the cost of a
query-evaluation plan It uses formulae that estimate costs for a number of
options, and select one with lowest cost. It also considers only cost of disk access,
which is usually dominant cost in Query Processing.
The response time for a query-evaluation plan is very hard to estimate without
actually executing the plan, for the following reasons:
a. The response time depends on the contents of the buffer when the query begins
execution; this information is not available when the query is optimized, and is
hard to account for even if it were available.
b. In a system with multiple disks, the response time depends on how accesses are
distributed among disks, which is hard to estimate without detailed knowledge of
data layout on disk. A cost-based optimizer explores the space of all query-
evaluation plans that are equivalent to the given query, and chooses the one with the
least estimated cost.
The term “memorization” is a process of storing the optimal query evaluation plan
for a subexpression when it is optimized for the first time; subsequent requests to
optimize the same subexpression are handled by returning the already
memoized plan. The optimizer in PostgreSQL is cost based query
optimizer.
Several Cost components to consider like
● Fully replicated
● Partially replicated
● Fragmented
● Mixed
Fully Replicated
In this design alternative, at each site, one copy of all the database tables is
stored. Since, each site has its own copy of the entire database, queries are very
fast requiring negligible communication cost. On the contrary, the massive
redundancy in data requires huge cost during update operations. Hence, this is
suitable for systems where a large number of queries is required to be handled
whereas the number of database updates is low.
Partially Replicated
Copies of tables or portions of tables are stored at different sites. The
distribution of the tables is done in accordance to the frequency of access. This
takes into consideration the fact that the frequency of accessing the tables vary
considerably from site to site. The number of copies of the tables (or portions)
depends on how frequently the access queries execute and the site which
generate the access queries.
Fragmented
In this design, a table is divided into two or more pieces referred to as fragments
or partitions, and each fragment can be stored at different sites. This considers
the fact that it seldom happens that all data stored in a table is required at a
given site. Moreover, fragmentation increases parallelism and provides better
disaster recovery. Here, there is only one copy of each fragment in the system,
i.e. no redundant data.
The three fragmentation techniques are
● Vertical fragmentation
● Horizontal fragmentation
● Hybrid fragmentation
Mixed Distribution
This is a combination of fragmentation and partial replications. Here, the tables
are initially fragmented in any form (horizontal or vertical), and then these
fragments are partially replicated across the different sites according to the
frequency of accessing the fragments.