Database I

Methodology

Physical Design

1
Physical Database Design
 Throughout the processes of conceptual and
logical database designs and the
normalization, the primary objective has been
the storage efficiency and the consistency of
the database
 In the physical database design, however,
the focus shifts from storage efficiency to the
efficiency in execution

2
Physical Database Design
(Cont.)
 The physical DB design involves:
 Transforms logical DB design into technical
specifications for storing and retrieving data
 Does not include practically implementing the
design however tool specific decisions are
involved
 The Physical design requires the following
input
 Normalized relations
 Definitions of each attribute (means the purpose 3
or objective of the attributes)
Physical Database Design
(Cont.)
 Descriptions of data usage (how and by whom
data will be used)
 Requirements for response time, data security,
backup etc.
 Tool to be used
 Decisions that are made during this process
are:
 Choosing data types
 Deciding file organizations
 Selecting structures
 Preparing strategies for efficient access 4
De-normalization
 De-normalization is a technique to move from higher
to lower normal forms of database modeling in order
to speed up database access
 De-normalization process is applied for deriving a
physical data model from a logical design
 In logical design we group things logically related
through same primary key
 In physical database design fields are grouped, as
they are stored physically and accessed by DBMS

5
De-normalization (Cont.)
 We should be aware that each new RDBMS
release usually bring enhanced performance
and improved access options that may
reduce the need for De-normalization
 A fully normalized database schema can fail
to provide adequate system response time
due to excessive table join operations

6
De-normalization (Cont.)
 De-normalization Situation 1:
 Merge two Entity types into one with one to one
relationship
 Even if one of the entity type is optional, so joining
can lead to wastage of storage, however if two
accessed together very frequently their merging
might be a wise decision
 So those two relations must be merged for better
performance, which have one to one relationship

7
De-normalization (Cont.)
 De-normalization Situation 2:
 Many to many binary relationships mapped to three
relations
 Queries needing data from two participating relations
need joining of three relations that is expensive
 Join is an expensive operation from execution point of
view
 Consider the many to many relationship b/w EMP,
PROJ and WORK
EMP (empID, eName,pjId,Sal)
PROJ (pjId,pjName)
8
WORK (empId.pjId,dtHired,Sal)
De-normalization (Cont.)
 So now if we by de-normalizing these relations
and merge the WORK relation with PROJ relation
 But in this case it is violating 2NF and anomalies
of 2NF would be there
 But there would be only one join operation
involved by joining two tables, which increases
the efficiency
EMP (empID, eName,pjId,Sal)
PROJ (pjId,pjName, empId,dtHired,Sal)

9
De-normalization (Cont.)
 De-normalization Situation 3:
 In 1:M situation when the ET on side does not
participate in any other relationship, then many side
ET is appended with reference data rather than the
foreign key
 In this case the reference table should be merged with
the main table
 Consider STUDENT and HOBBY relations
 One student can have one hobby and one hobby can
be adopted by many students
 Here hobby can be merged with the student relation
 Thus redundancy of data would be there, but there
would not be any joining of two relations, which will
have a better performance 10
Partitioning
 Partitioning splits same relation into two
 Aims of data partitioning in database are to
 Reduce workload (e.g. data access,
communication costs, search space)
 Balance workload
 Speed up the rate of useful work (e.g. frequently
accessed objects in main memory)
 There are two types of partitioning:
 Horizontal Partitioning
 Vertical Partitioning 11
Partitioning (Cont.)
 Horizontal Partitioning
 Table is split on the basis of rows, which means a
larger table is split into smaller tables
 The advantage of this is that time in accessing the
records of a larger table is much more than a
smaller table
 Range Partitioning
 In this type of partitioning range is imposed on any
particular attribute
 For Example for those students whose ID is from 1-
1000 are in partition 1 and so on 12
Partitioning (Cont.)
 Hash Partitioning
 A particular algorithm is applied and DBMS knows
that algorithm
 So hash partitioning reduces the chances of
unbalanced partitions to a large extent
 List Partitioning
 In this type of partitioning the values are specified for
every partition
 So there is a specified list for all the partitions

13
Partitioning (Cont.)
 Vertical Partitioning
 Vertical partitioning is done on the basis of
attributes
 Same table is split into different physical records
depending on the nature of accesses
 Primary key is repeated in all vertical partitions of
a table to get the original table
 Consider the Student relation
STD (stId, sName, sAdr, sPhone, cgpa, prName,
school, mtMrks, mtSubs, clgName,
14
intMarks, intSubs, dClg, bMarks, bSubs)
Partitioning (Cont.)
 We can partition this relation vertically as
under
STD (stId, sName, sAdr, sPhone, cgpa,
prName)
STDACD (sId, school, mtMrks, mtSubs,
clgName, intMarks, intSubs,
dClg, bMarks,bSubs)

15
Data Storage Concepts
 Physical Storage Media Storage media are
classified according to following characteristics:
 Speed of access
 Cost per unit of data
 Reliability
 RAID – Redundant Array of Inexpensive Disks
 Many disk that look as a single disk to OS but have better
performance and better reliability
 RAID disk drives are used frequently on servers
 RAID have the property that the data are distributed over
the drives to allow parallel operations
16
Data Storage Concepts (Cont.)
 Fundamental to RAID is "striping", a method
of concatenating multiple drives into one
logical storage unit
 Striping involves partitioning each drive's
storage space into stripes which may be as
small as one sector (512 bytes) or as large as
several megabytes
 The type of application environment, I/O or
data intensive, determines whether large or
small stripes should be used
17
Data Storage Concepts (Cont.)
 RAID-0
 Simple Striping
 Virtual single disk is divided up into strips of k
sectors each
 Since no redundant information is stored,
performance is very good, but the failure of
any disk in the array results in data loss

18
Data Storage Concepts (Cont.)

1 2 3 4
5 6 7 8
9 10 11 12

Note: This example is a basic virtual drive where

20
Data Storage Concepts (Cont.)

1 1’ 1’’
2 2’ 2’’
3 3’ 3’’

21
Data Storage Concepts (Cont.)
 RAID-2,3
 For reliability simple parity check code is used
 Parity bit is stored on separate disk
 RAID-4
 RAID Level 4 stripes data at a block level across
several drives, with parity stored on one drive
 The performance of a level 4 array is very good
for reads (the same as level 0)
 Writes, however, require that parity data be
updated each time
22
Data Storage Concepts (Cont.)
 RAID-5
 RAID Level 5 is similar to level 4, but distributes parity
among the drives
 This can speed small writes in multiprocessing systems,
since the parity disk does not become a bottleneck
 RAID-0 is the fastest and most efficient array type
but offers no fault-tolerance
 RAID-1 is the array of choice for performance-
critical, fault-tolerant environments
 RAID-2 is seldom used today since ECC is
embedded in almost all modern disk drives
23
Data Storage Concepts (Cont.)
 RAID-3 can be used in data intensive or single-user
environments which access long sequential records
to speed up data transfer. However, RAID-3 does
not allow multiple I/O operations to be overlapped
 RAID-4 offers no advantages over RAID-5 and does
not support multiple simultaneous write operations
 RAID-5 is the best choices in multi-use
environments which are not write performance
sensitive. However, at least three and more typically
five drives are required for RAID-5 arrays
24