0% found this document useful (0 votes)
57 views36 pagesFeb 22, Doc 1
Uploaded by
Ankush SharmaCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
57 views36 pagesFeb 22, Doc 1
Uploaded by
Ankush SharmaCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 36
Normalization is part of the Successful database design, without normalization, database
systems may be inaccurate, slow and inefficient and they might not produce the data we
expect.
When we normalize the database, we have four goals :
1. Arranging data into lo,
the whole.
2. Minimizing the amount of duplicate data,
3. Organizing the data such that, when you
one place.
sical groupings such that each group describes a small part of
called redundancy, stored in a database,
modify it, you make the changes only in
4. Building a database in which you can access and manipulate the data quickly and
efficiently without compromising the integrity of the data in storage.
155 Database Management S}
f reducing data redundancy in a relational da
rocets Of Fiaty, normalization is a refinement process inte ee
ee fter identifying the data objects of the proposed Matabontsbase
‘and columns within each table. The finden
data should not be stored in multiple place: egal
Sipbliniet op tabled mehctilly (tcReaaman
ines the tables required
‘alization is “The same
in the process, however,
ges of normalization are +
from certain add, edit and delete anomalies.
restructure the database as new kinds of data are introduced
tive to users, including different users making
The main advantat
« It frees the database
«+ Itreduces the need to
» Itmakes the database more informal
different queries.
« jt avoids biasing the database design in favour of certain queries at the expense of
others,
+ It allows the operations needed
otherwise have to be.
Designing a normalized database structure is the first step when building a database.
‘Taking time to properly design the database is the most important step in developing
database-oriented solutions. Normalization is @ simple, common sense process that leads
cotenible, efficient and maintainable database structures:
eee
5.2, FUNCTIONAL DEPENDENCY (UPTU, 2006) (Imp,)
'd by XY, between two sets of) attributes Xand Y that are
toy the possible tuples that can from a relation, state r
two tuples t, and f, in that have
for data access to be simpler than they would
‘A functional dependency, denotes
subsets of R specifies a constraint
or R. The constraint is that, for any
1X] = 4141.
We must also have
4111 = of¥)-
‘this means that the values of the’ Y component of a tuple in r depends on, or are
values of the X component
determined, by the values of the. _X component; or alternatively, the ve
of a tuple uniquely (or func! ‘ne Y component. We also say
tionally) determine the value of t]
that there is a functional dependency from X to ¥ oF that Yis functionally dependent on x,
The abbreviation for Yis functionally dependent is FD or f.d. The set of attributes X is
called the left-hand side of the FD, and Yis called the right-hand side.
‘An attribute in a relational model table is said to be functionally dependent on another
attribute in the table if it can take one value for @ given value of the attribute upon which
it is functionally dpendent.
play a key role in
Functional dependencies
bad database designs. A functional dependency i
of the notion of key-
Itis denoted by >; A> B
differentiating good database designs from
sa type of constraint that is a generalization
For example: Consider the relation schema
—————————___——— __ EMP_PROJ
[ssn | PNUMBER [HOURS] ENAME | PNAME [_PLocarion
FD, ae
FD,
FD, ;
From the semantics of the attributes, we know that the following functional dependencies
should hold :
(a) SSN > ENAME
(6) PNUMBER ~ (PNAME, PLOCATION}
(o {SSN, PNUMBER} > HOURS,
specify that (a) the value of an employee's social security
‘These functional dependencies
number (SSN), uniquely determines the employeename (ENAME), (b) the value of a projects's
vember (PNUMBER) uniquely determines the project name (PNAME) and location
.d PNUMBER values uniquely determines
{PLOCATION) and (¢) a combination of SSN an
cre number of hours the employee works on the Project Pet Week (HOURS). Alternatively,
we say that ENAME is functionally determined by (or functionally dependent on) SSN.
'A FD. is a property of the relation schema (intension) R, not of a particular legal
relation state (Extension) r of R, Hence, and FD, cannot be inferred automatically from a
given relation extension r but must be defined explicitly by someone who knows the
ccmantics of the attributes of R, The above mentioned figure introduces a diagramatic
Sefation for displaying FDs: Each FD is displayed as a horizontal line, The left hand-side
attributes of the FD are connected by vertical lines to the line representing the FD, while
the right hand side attributes are connected by arrows pointing towards the attributes.
5.2.1. Closure of a Set of Functional Dependencies (uPrU, 2003)
“The schema designer specifies the F.D. that the semantically obvious, usually, however,
jumerous other F-D. holdin al legal relation instances that satisfy the dependencies in Fr
whose other dependencies can be inferred or deduced from the FDs in F. For feab-life
ceenple, it is practically impossible to specify all possible functional dependencies that
Grey hold. The set of all such dependencies is called the closure of Fand is denoted by /*.
‘ror example : Suppose that we specify the following F of obvious FD on the relation
schema.
F= {SSN > (ENAME, BDATE, ADDRESS, DNUMBER),
DNUMBER + DNAME, DMGRSSN}
We can infer the following additional F.D. from F:
SSN = (DNAME, DMGRSSN},
SSN = SSN,
DNUMBER = DNAME
An FD X-> Yis inferred from a set of dependencies F specified on Rif X> Yholds im
that is legal extension of R ; ie., whenever r satisfies all the
Yalso holds in r, The closure F* of Fis the set of all F.D. that can
IR®) are well-known as Inference
every relation state 1,
dependencies in F, X>
be inferred from F. The following six Inference Rules (IR, to I
rules for functional dependencies : then & >, holds,
(PTY 2099
aset of attributes, then yq,
> by
sa set of attributes and poe®
IR, : Reflexivity Rule * If ai
js holds and 718
Ifa Pr
holds, then @ > ¥ holds.
IR,: Augmentation Rule
holds. -
IR, : Transitivity Rule + Ifa B holds ans
IR, : Union Rule : If >? holds and & “hy
IR, : Decomposition Rule: Ifa>B hol A
IR, : Pseudotransitivity Rule: Ifa 7B ho -
«pat Axio!
‘The rules IR;, IR, and IR, are ale RA.
honour of the person who first propose? 1 ; a oe
‘The Reflexivity rule (IR,) states that 4 ae a att endencied beeen OF any of
; Ve ems a s
its subsets, which is obvious. BH, Sepa Vie trivel XE othe Ue,
iL g the same set of attributes * 7 itig
oth
such dependencies are called trivi Se aadic
ner valid dependency. Aceon
nontrivial. The augmentation Rule (IR,) Svs
nee fa dependency results in anothe “ i
es are transitive. The union rule (IR,) aj ing
i lows
the left and right handsides 0 oe
to Transitivity rule (IR,); functional dependene™ > ;
1p combine alset of dependencies > A137" Ay eX aon nt single
{A,, Ag) «11» Ap}: The decomposition rule (IR) says that we cae ributes from
id : applying this rule repeatedly we can decompose the p®
jes (XK Ay, X79 Ay XA; le RD
right-hand side of a dependency
X-— {A,, Ay, «. 4,} into the set of dependenci
ed from the definition of 5
proved iré F.D. either by
Each of preceding inference rules can be
direct proof or by contradiction.
poo
holds,then & > fy holds.
then a 3B holds, and «+ holds,
jas and vB -* 5 holds,then 1 8 hojg,
ms” or Armstrong's Axiom,
In
EXAMPLE 5.1. 1 Given a set of F.D-
A> B, ABCD Eand EF > G
Is ACDF > G, implied by the set of given FD’s ?
A> B, ABCD>E
ACD = E (by Pseudotransitivity Rule)
ACDF > EF (by Augmentation Rule)
ACDF 4 G (by Transitivity Rule)
Hence ACDF > G
First three rules are known as Armstron; i ivi
: axions. The Reflexivity Rul
a a set of attributes always determines itself or any of its subsets, which eae ag 4
such type of obvious dependencies is called as trivial dependencies. S
dependency X—> Yis trivial, if Xc Y; ee
Otherwise it is non-trivial.
Solution: Given
Implies
Again
EXAMPLE 5.2. Given two sets F, and F, of FDs for a relation.
F,:A> B AB>C,D>AC DOE
F,:A— BC, D> AE
Are the two sets equivalent ?
Solution: From F, , D> AC
a 7
DA (by Decomposition Rule)
Doc
Also given that DE
So DAE (by Union Rule)
which is a FD of set Fy.
Again A> Band AB C
Imply A C (by Pseudotransitivity Rule)
A BC (by Union Rule)
which is FD of set Fy.
Hence F, and F, are equivalent.
5.2.2. To Find Closure
A BG, E> CF, B> Eand C> EF
EXAMPLE 5.3.
Compute the closure X* of the set of attributes (4, 3 under the given set of Fs,
Initialize X= {A,B
xt) = {A, B,Q
Xx) = {A, B, C, B
X08) = {A, B,C EA
x) ={A, B,C, BP
x0) = x)
¢ of the given F.D. over (4, 5) is
{AB ‘= (4, B,C BA
Hence closur’
A> B, BC>D,C>E, CD>F.
EXAMPLE 5.4.
Find closure of (A, Q*
X) = (4, O
XO) = (4, BO
B68
{A,B,C E, Dt
(A, B, C, D, E, F}
={(4,B,6D,E7
x8 = xO)
Hence (A, 0° = (4B, CDE
EXAMPLE 5.5. A B, BC DE, AEF > G.
Find closure of (A, O)
xX) = {AG
Xi) = {A, B, O
th
ed ais}
S Database Management System
X@) = (A, B, CF, B
X0) = {A, B,C, D, B
Xx) = xe)
Hence (A, Q* = {A, B, CD, B
5.2.3. To Find Super Key, Candidate Key and Irreducible
EXAMPLE 5.6. ABCD E Super Key = {ABCDG)}
AB~> G
BoF
Cu Super Key = (ABCDO}
aol
GoH
= ABCDGJ (:. C>)
= ABCDG
=ABCD (-. AB? Q)
Because neither of (ABCD) is functionally determine by any F.D. So (ABCD) is an
irreducible set.
Hence candidate key = {A, B, C, Di
EXAMPLE 5.7. AB Super Key = (ABCEF,
BC>D
BC>E
AEF G = ABCEF (-. C> E)
B3G =ABCF
=ACF . (--A>B)
Candidate key = {4,6
EXAMPLE 5.8. AB
BCD Super key = {ABCDE}
CoE
Doc
E— F=ABCDE (:.C> 4)
Since it replace only one attributes.
=ABCDE (-.B— CD)
=(4B,B (--A>B)
= {A, B
= (AB) = Candidate key or irreducible set.
EXAMPLE 5.9. AB>C
CoA Normaiizat,
a
roe
Ace 161
28), CSui
Per key = (ABCD Ey
BE DEF
PACA AECD (cop sp
CEs p ~E,D3A
CEA
CFB = (ABQ |.
CFoap (ABQ (-. BC+ Dy
DAE AD in (e
Desir (. ABs 9
+ Candidate key (4,
EXAMPLE 5.10. A BC,
BOE
CD EF
= ABCD (-. A BQ)
=AD
Super key = (ABCD,
= Candidate key or irreducible set.
ABS C
A DE
Bor
F> GH
DoW
=(ABDFR)
ABD (-.B>
ABD
EXAMPLE 5.11.
‘Super key = {ABDF)
= Candidate key or irreducible set.
Canonical Cover
Consider the following set F of functional dependencies on shemea (A, B, Q:
ABC, B>C, A+B, AB>C
Let us compute the canonical cover for F.
+ There are two F.D. with the same set of attributes on the left side of the arrow :
A+BC,A3B
We combine these F.D. into A > BC.
+ Ais extraneous in AB C, because F logically implies (F={AB — C}) U {B > C}. This
assertion is true because B > C is already in our set of F-D.
* Cis extraneous in A BC, since A > BC is logically implied by A> Band B> C.
Thus, our canonical cover is
A >B
BOC a
Database Management System
IPTU, 2006)
5.3. LOSSLESS/LOSSY JOIN peEcomposiTioN P ca pas 2 a
i ncie: .
Let Rbe a relation schema, and let Fbe a set of functions! See ath ae ae ae
R, form a decomposition of R. Thus decomposition ee i on ion o
at least one of the following functional dependencie
. RRR
e RORFR
. RAR OR R, a f
: se. Bi ropeiata SODETRCY of either R, oF Ro the composition of R oe
in other words, if Ry 7 Rp fo of decomposition of arelation into multiple
ele re complicated.
ition. For the gen
for lossless join dec
position i
f all cons
omposition is m0
is clearly a SU)
traints are func’
lossless-join decompos'
parts of once, the test
While the test for binary decom)
join, it is a necessary condition only i
sider scheme
S=(VW%YD
Z>V
woy
XY 9Z
vo Wx
S,=(VW,%)
5-7
S39 S-)
V—> Wx
Since (V, W, X) are present in S;-
Hence it is lossless-join decomposition.
ficient condition for lossless
tional dependencies.
For example: Con
(a S, = (V, WX)
(KF ;
8,95 - (4)
Since there is no f.d.
Hence it is lossy-join decomposition.
Same example can be done by matrix method.
Example : Consider the scheme
Sa (VW, Xy%Q
Suppose, the following F.D holds
Suis wh ZoV,WY, XY> Z, U> WX
oustity et bade oe following decomposition of scheme Sis lossless join decompositio;
(S,=(V, WX) 4
(i) S,=(V, ¥, 2) ul
Normalization
163
Matrix S will be
v w x ¥ Zz
Ss a1 12 13 oh Oleh.
isl meta y baa 2) 94 a
After applying functional dependencies, we get matrix Sas
v Ww x v6 Zz
Sy hy So, Bs Pry Dis
Sy || 03 a a, %, %s
Since second row filled wi
decomposition.
(9S, =(V, W, x
(i) S = (X, ¥,2)
Matrix S will be
ith all a’s, so we stop. Hence decompositio;
lossless join
v w x ¥ Z
Si an 12 413 bia bis
Ss by by 5 44 5
After applying functional itependencies, we get matrix Sas
4 w x 3 Zz
oH a S12 a3 big a5,
S bay bay 3 O54 ie
There is no row filled with all a’s and Bs,
Hence the decomposition is lossy join decomposition,
5.4. NORMAL FORMS
SS NORMAL FORMS —
together, should possess. These would include two propertice
1. The lossless join or non-additive join property, which guarantees that the spurious
Spe Beneratonproblems does not occur with respect to the relation schemas
created after decomposition,
2. The dependency presentation, which ensure that
each FD is represented in some ~
individuals relations resulting after decomposition
n. agement System
js extremely critical and must be achieved at
a fal erty ‘i i ;
di tive foreseen property, although eal is sometimes sao
whereas the ve soning tbe fi of higher normal form rethon® as a base Tela, 4
ae ae name normal fo: is known as ‘Deno: i *
which is in sore normal forms, which he called first, second an
Initially, Codd ProP™ nition of 3 NF, called Boycee-Codd Normal Form (By
LA aT and Codd. All these normal forms are based on the f 2 a
by Boyer attributes of 2 relation, Later, a fourth normal form (ANF | STUDENT_ID | STUDENT_NAME
SUBJECT_NAME] GRADE
[ STUDENTID
SUBJECT Table—> [STUDENT_ID
Inessence, INF calls for the atomicity of data (one fact
i D per field, bs
smaller parts without losing meaning), and the elimination of ate ane
of data
through the creation of separate tables of related data. 5.4.2. Second Normal Form
To be in 2NF, a table must be in First Normal form and no attributes of the table
should be functionally dependent on only one part of a concatenated primary key. Thus,
2NF can be violated only when a key is a composite key, or in the other words, one that
consists of a more than one attributes.
2NF is based on the concept of full F.D. A. F.D. X > Yis a full F.D. if removal of any
attribute (A) from X mean that the dependency does not hold any more.
That is, for any attribute AI X,. X - {A} does not functionally determine Y.A. FD Xe Y
is a partial dependency if some attribute A « Xcan be removed from X and the dependency
still holds, ie., for some Ae X, (X-{A)) > Y.
The test for 2NF involves testing for FD whose left-hand side attributes are part of the
primary key. If the primary key contains a single attribute, the test need not be applied to
all.,A relation schema Ris in 2 NF if every nonprime attribute A in Ris fully functionally
dependent on the primary key of R.,
Second normal form depends on the concepts of the primary key and functional
dependency. A database is in 2NF if it is in INF and every attribute is fully functionally
dependent on the primary key. Thus, the relation is in INF with no repeating groups and
all non-key attributes must depend on the whole key, not just some part of it. Functionally
dependent means that given a primary key, the value of any attribute can be retrieved For
example, in the above mentioned example, if the STUDENT_ID is given the value of
STUDENT_NAME can be obtained. Hence, STUDENT_NAME is functionally dependent on
STUDENT_ID.
EXAMPLE 5.12.
EMP_PROJ
SSN PNUMBER | HOURS NAME | PNAME PLOCATION
FD, j
FD,
FD,
Relation Not in 2NF
EP, EP, EP,
ssw | enumpen [Hours | ssn | enave ] _ [Pwumpen [pnaMe] pLocarion |
FD, [a J FD, FD, f j
Relation in 2™4 Normal Form o-
Domain
Database Management System
value from {
Domain usd
tested easily EXAMPLE 5.13.
Domain ar
{iis pe jerNo. | Title Ss
ae “a Computer Networks 250
ee . Graphi _
Neale 1. ee ; ae '
ee 2. Multimedia 300
ind a
Thus, if we a Data structure 55d
name and ‘ DBMS 295)
Soe, 3. Multimedia 300
‘aa 3, ‘Computer Networks 250
compariso}
permits de ‘Relation Not in 2NF ‘ 3
| ihe a eore nie cence nce
The ‘The above table lists the various orders of compu a :
satisfied t wee eet poole and tho unit prion of eRe Hos nat eee tat
the primary key eince bath orton or feat title can repeat in the table. This table isin am
aes normal form as every ‘attribute is atomic, but ‘tle notin second BOT ea because
Fait price ie vot functionally Gependent on order NO, part of the concatent, ea primary
Key, but only on the ttle emonent. Quantity on the no Pe hand is functionally dependent
on connected primary Key: ,
Tecan be seen that the prices of some books are repeated because 2 title can be orderea
in multiple orders. For ‘example, the title SDIMS' exists in two orders order_no Land order no
3. Thus if the prices of this book is to be updated then we have to update it in all
ce rresponding entries in the table. Moreovere if we want to add a book to our library, then
cols of this book can Pe Nes Mere until onl ordes JS)Plner hia bole tectities
suppose, order_no 2 is « ancelled, so we have to delete this order from the database, which
‘he information about that book including the price of the book,
we have to decompose it in {Wo tables as follows :
qnty.) and
will loose all t
For this table,
ORDER MASTER
to be in 2NF,
ORDER MASTER
(order_no, title,
until_price).
BOOK MASTER (Title,
Computer Network
Graphics
DBMS
Multimedia —
Data Structure
DBMS,
Multimedia
Computer Networks
Sere et Cormatizaton )
BOOK_MASTER 67
Title 5
UNIT _PRIC!
Computer Networks Baul =
Graphics
DBMS Ae
Multimedia ae
Data structure ee
Relation in 2NF p
‘ll the problem, of deletion, insert
5.4.3. Third Normal Form
A relation is in 3NF, ifit is in 2NF and no non-pri
? ‘Prime
on other non-prime attributes. © attributes functionally dependent
It is also define as a relation is in 3NF, if it is in ONF a:
: eT
should be transitively functionally dependent on the primary key, utes of the tabla
3NF is based concept of transitive dependency.
A functional dependency X--» Yin a relation schema R, is a transitive dependency if
there is a set of attributes Z that is neither a candidate key nor a subset of any key oR
and both X > Z and Z> Yhold. i
A table is said to be in Third Normal Forms if all the non-key fields are independent,
ie, no two non-key fields of the table depend on each other. The 3NF goes one step further
in removing redundant data by removing field that are not wholly dependent on the primary
key. In essence, 3NF seeks to eliminate all attributes from a table that are not direcvtly
dependent upon the primary key. Hence, a table is said to be in 3NF if and only ifit is in
2NF and every non-key field is non-transistively dependent on the primary key.
Consider the following PROJECTS Table.
PROJECTED ID MANAGER ADDRESS
P, Susan Park Street Q
P, Janet High Street
P, Susan Park Street me
P, Susan Park Street
PROJECTS TABLE
In 3NF, we are looking for data in our tables that is not fully dependent on the primary
key, but dependent on another value in the table. In the PROJECTS table above the
ADDRESS field is dependent on the MANAGERS and not the PROJECT_ID. To bring this
table into compliance with 3NF, a separate table base on MANAGERS is needed. Hence, Domain
value from
Domair
tested casi]
Domair
It is pe
customer-r
names. Ho
Itis perhar
Normally ;
“Find €
Thus, if wi
name and
So, De
compariso
permits dc
The ch
satisfied E
Query
the PROJE!
pher for all
and anotl
pROJECTS TABLE
tabase ‘Management System
gplit into two, one with the transitive q
e
Pendeney f
table needs to be n
CT al of the remaining attributes
ns
PREOJECT IP MANAGER
Py Susan
a Janet
Susan
Ps
Susan
MANAGERS TABLE
MANAGER ADDRESS
Susan Park Street
Janet High street
Tables in 3NF
The determinant attribute becomes the primary key in the new table, i
to the MANAGERS table. The original key he MANAGER
utes—in this case, PROJECTED ID ha
ID remaing
the
‘anager’
field becomes
ining non
to the remal
key to the ne’
address, it car
between the
are now in
For exam,
EMP_DEPT, because
DMGRSSN hold and D
EXAMPLE 5.14.
EMP_DEPT
ENAME ssn | BDATE /ADDRESS| DNUMBER
ED,
Ww
n be looked up in tl
in the PROJECTS
‘Third normal form.
the primary key
-transitive attruibi
e address is stored only once.
he MANAGERS table. The MAGE Se
eis the ling
PROJECTS table. Now the
table, it is now a forei,
ign key. The:
Se tables
two tables, and i
dependency SSN > DMGRSSN is transitive th
dependencies. SSN > ENUM BERIEEAES a
nor a subset of the key of Sra A
PT.
both the
iple, the
NUMBER is neither a key itself
DMGRSSN
ED,
DNAME | DMGRSSN
Ppoonessfonwoe] fi DNUMBER
|!
t
[ ENAME | ssw [aoare
valu fo arf ol
Relation in 3 NF
—————— ~—S—“‘i‘i rT DD
EXAMPLE 5.15.
ie
Course Name Head-Deptt. Room No. Room-Capacity
B.Tech (CS) Prof. Gupta 102 60
B.Tech (IT) Prof. Smith 107 50.
B.Tech (EC) Dr. Sundaram 105 60
B. Tech (Al) Dr. Srinivasan 103 100
MCA Mr. Sharma 11 40
Relation not in 2NF
In the above relation, room-capacity is functionally dependent on room_no and room_no
is functionally dependent on course_name. So a transitive functional dependency exists
here. Room-capacity is transitive functioanlly dependent on course-name.
COURSE_NAME—> ROOM_NO—> ROOM_CAPACITY
‘The Presence of this transitive functional dependency will cause the problems. The
capacity of a room cannot be entered in the database un. less a course is scheduled in that
room; and the capacity of a room in which only one course is scheduled will be deleted if
the only course scheduled in that room is deleted.
So this relation is not in 3NF. For this relation to be in 3NF, we have to decompose it
into two tables COURSE and ROOM as follows :
COURSE
COURSE_NAME HEAD_DEPT ROOM_NO
B.Tech (CS) Prof. Gupta 102
B.Tech (IT) Prof. Smith 107
B.Tech (EC) Dr. Sundaram 105
B.Tech (Al) Dr. Srinicasan 103
MCA Mr. Sharma lll
ROOM
ROOM_NO ROOM _CAPACITY
102 60
107 50:
105, 60
103 100
111 40.
Relation in 3NF
Consider one more table EMPLOYEE having the attributes (emp_no, ename, sal,
dept_name, dept_location). In this relation, emp_no is the primary key attribute. Dept name
is the determinant of dept_location and being prime key attribute, emp_no is the determinant
of dept_name. So, a transitive functionally dependency exists in between dept_location
and emp_no. Database Management System
TION
|
Domair
value from rT "au
t ied cail EMP_NO—> DEPT_NAME—* DEPT_
es
Donal EXAMPLE 5.16. ae aeration
eS ¢ Sal Dept! see 5
It is pe | emp no. E-Name Ea ‘
customer-t 1001 Vishal 7500 hes vat
names. Ho ee Ait 5000 4 as
It is perhar a4 ea rian ee a a
Normally ( ue cae bead oi 104
“Find € ae
Thus, ite 1005 Sumit 6500
name and /
So, Da Not in 3 NF. assay Nama Dept Location 4
compariso d remove the transitive dependency
vermits de To make it in 3 NF we decompose a”
The cl Emp No — Dept-Location Petre |
atisfied b
Query
Vishal
Amit
Anuj
| Vikas
d 2
aie ! Sumit 6500
ae Department
shee’ Dept-Name Dept-location
Accounts 102
Sales 104
Store 106
ue Relation in 3 NF
LOTS ANF
LOTS1 LOTS2 2NF
LOTS2 SNF
LOTSIA LOTS3B.
Summary of normalization of lots
(UPTU, 2002, 2006)
5.4.4, Boyce-Codd Normal Form (BCNF)
BCNF is an extended form of 3 NF. If a relation is BCNF then it must be in 3NF. In
BCNF we extend our concept up to all the candidate keys of the relation, which are
concatenated and two or more of the candidate share a common attribute.
eT i
BCNF, a table must only have candidate key as determinants”,
i *
dd normal form (BCNF) was proposed as a sim,
tricter form of 3 NF, because every relation i
Rin of BCNF differs slightly from the definition
if whenever a nontrival functional dependency
pler form of 3 NF, but it was
n BCNF is also in 3 NF. The
of 3 NF. A relation schema R
X— A holds in R, then Xisa
functional dependency if, all
ictional dependencies in F* of the form o 8 where
© Rand fc R, at least one of the following holds.
> fis a trivial functional dependency (Le, Bea)
@ is a super key for schema R.
ple : Consider an example that is not in BCNF.
bor_loan = (cust_id, loan_no, amount)
olds on bar_loan, but loan_no is not
msortium of many customers).
- The functional de
pendency loan_no > amount
a superkey (becau:
se a loan may be made to a
Example of BCNF :
Stud_ID S_Name Subject Grade
1001 Axay Physics A
1001 Axay Chemistry &
1001 Axay Maths iG
1002 Sparsh Physics, A
1002 Sparsh Chemistry A
1002 le Sparsh Maths B
Relation is not in BCNF
In this relation following FDs exist :
S_NAME, SUBJECT —, GRADE
STUD_ID, SUBJECT —, GRADE
SNAME —5 STUD_ID
STUD_ID —, S_NAME
ition. The rule for
a relation R, which is not in BONE, ig as follows :
Let XR, A be the single attribute in Rand X—> Abe an FD that causes a violation
of BCNF. ve
1.4.4,
Bc?
CNF ¥
neate
i
XA.
in BCNF,
‘SUBJECT
Physics
Chemisty
Maths
Physics
chemistry
Maths
Decompose Rinto RAand
jation to be
the decomposition will be as follo
So, for above rel
__ GRADE
w>p>aa>
NAME
7001 “Axay
1002 Sparsh
pep sal
‘Relation in BCNF
55. MULTI-VALUED GEPENDENGY AND FOURTH NORMAL FORM
dent multi-valued facts about th
neraily, it is denoted by 'e same attribuy
iti valued dependency of Yor X* and m,
and
more indepen’
More ge!
“there is a mul
occurs when two or
nthe same zelation-
readers
MVD
occur withit
x 5 > ¥ and will be
determines a
eyby denoted by >.» epecttieala i
Fp specifies the following ee schem,
such that (X) = (09, then two vo
rties, where we use Z to one ‘e
Multivalued dependency (
where X and Y are poth subsets of
ror R: If two tuples 4 ‘and ¢, exist in Ts
should also exist in rwith the following prope!
1, (39 = GIA) = A) = 209
#4, (9 = 40) and 40 = 99
+4 = HQ and 44 = 44
An MVD X— Yin Ris called
‘An MVD that satifies neither (a) nor (Bis
a trival MVD if (a) Xis a subset of X
or () XU ve
5 called a nontrival MVD. (0 XO
For example :
Relation with MVD
Faculty ‘Subject Committee
DBMS _ Placement
Ne ear Placement
eas Placement
ae » Scholarship
pete Scholarship
Scholarship @
Multi-valued dependency (MVD) occurs when two or more independent multi-valued
facts about the same attribute occur within the same relation. More generally, MVD can
be defined as follows :
Let R be arelational scheme and let Xand Ybe the subsets of attributes of R. Intutively
the MVD X > > Yis said to hold over R if, in every legal instance rof R, each X value is
associated with a set of Y values and this set is independent of the values in the other
attributes. Intuitively, a raulti-valued dependency is denoted by X > + Yand will be read
as “there is a multi_valued dependency of Yon X” or “X multi-determines Y”.
Fourth Normal Form
A table is in 4NF, if it is in BCNF and it contains no multi-valued dependencies.
Arelation schema R is in 4 NF, with respect to a set of dependencies F (that includes
FD and multivalued dependencies) if, for every nontrival multivalued dependency X—>» Y
in F*, Xis s superkey for R,
Consider again the same faculty relation with MVD. Clearly a faculty has multiple
courses to teach and he is leading several committees. This relation is in BCNF, since all
the three attributes cancatenated together consitutes its key, yet it is clearly wrong and
requires decomposition, The rule for decomposition is to decompose the offending table
into two, with the multi-determinant attribute or attributes as part of the key of both, In
this case to put the relation in 4NF, two separate relations are formed as follows :
FACULITY_COURSE (FACULITY, COURSE)
FACULITY_COMMITTEE (FACULITY, COMMITTEE)
For example;
Faculty Course Faculty-Committee
John ‘Subject John Placement _
John Networking John Scholarship
John MIS
Relation is in 4.NF
5.6. JOIN DEPENDENCY AND FIFTH NORMAL FORM (UPTU, 2005)
‘A join dependency (JD), denoted by (Rj, Ry, R;). Specified on relation scheme R, sepecties
a constraints on the states rof R. The constraint states that every legal state rof R should
have a lossless join decompositon into R,, Ry, --» Ry
‘That is, for every such r, we have
* (Tal), pale)» --» Healt) = 1
Notice that an MVD is a special case of a JD where n= 2, ie. a JD denoted as JD
(R,, R,) implies an MVD (Rj 0 Ry) > (R, = R,). A join dependency JD (Rj, Ro, ..- Ry),
Specified on relation schema R, is a trivial JD if one of the relation schemas R; in JD
(R;, Roy « R,) is equal to R. Such a dependency is called trivial because it has the losslessjoin
property for any relation state r of R and hence does not specify any constraint on R. Pfs.
TION PR
ONCEPT
TS Tnside this Chapter
6.4, Transaction System; 6.2. Properties of Transaction; 6.3. Transaction State, 6.4. Transaction |
processing System; 6.5. Recovery from Transaction Failures; 6.6. Distributed Database;
6.7. The Major Advantages of DDBMS; 6.8. Transaction Processing in DDBMS: 6.9. Data
Fragmentation.
core tte ieee apsinetaiah
&.1._ TRANSACTION SYSTEM iEPTU, 2006)
“Collections of operations that form a single logical unit of work are called transaction.”
‘Atransaction is a unit of program execution that accesses and possibly updates various
data items. It is also define as a logical unit of database processing that includes one or
more database access operations.
‘Transactions access data using two operations :
1, Read (X) : Which transfers the data item (X) from the database to a local buffer
belonging to the transaction that executed the read operation.
2. Write (X): Which transfers the data item (X) from the local buffer of the transaction
that executed the write back to the database. "
6.2. PROPERTIES OF TRANSACTION (UPTU, 2003, 2006) (Imp.)
Transactions have four basic properties, which are called ACID properties. These properties
are closely related to each other.
A Atomicity
C > Consistency
I Isolation
D > Durability
183 , jetely. OF Not
6.2.1. Atomicity oP)
_— com? “rion wil
‘A transaction is an atomic unit of processing: It ig either perfor’ pansaclc, ee
performed atall. At the need of a transaction the updates made bY
accessible to other transaction.
|
|
5 in a consistent
6.2.2. Consistency wa!
.s that if databaractions the database
‘The consistency property of a transaction implies
state before the initiation of a transaction, then at the end
will also be in a consistent state.
6.2.3. Isolation steps OF operations performeg
ti
‘The isolation property of a transaction indicates that he ution of 2 transaction,
Zated concurrently.
bya transaction should be isolated from other trans:
should not be interfered with by any other transactio!
6.2.4. Durability ae ansaction must persist in the |
‘The changes applied to the database b: }
database. These changes must not be lost bece!
ts are there and @ set of transactions |
C-I-D Properties Example
‘Example : In a Banking system there are 2% nts ewe assume that the datal
Me ig temporarity residing in main,
|
that access, and updates, those accounts.
permanently resides on disk, but that som
memory.
i database :
‘The following two opearations are accompklished to acces the/database
+ Fead (X), which transfers the data item X from the database to. local buffer belonging
to the transaction that execuited the read operation.
+ Write (x), which transfer the data item X from the local buffer of the transaction,
that executed the write pack to the database:
In real Database system, the write operation does not necessarily result in the imm eda
update of the data on disk, the write operation may be temporarily stored in memory a
executed on the disk later. However We shall assume that the write operation updates the
database immediately.
T, be a transaction that transfers $5!
7, : read (A);
A:=A-50;
write (A);
read (B);
=B+50;
Write (B).
portion o
0 from Account A to acount B. This is defined as
Applying Acid Properties
1. Consistency : Here consistency req. is that the sum of A and B be un
the execution of transaction. wy Transaction Processing Concepts
ut)
2. Atomicity : Let A= $1000
B= $2000
If failure oceures after write (A) operation was executed but buffer writ
avas executed, In, this. case, Me valucobaccouatA and Birellected in hs (B) operation
$1950 and $ 2000, Here we destroy $ 50 as the result of failure and ig detahase are
sum A +'B not
Preserred.
‘To ensure atomicity, the database system keeps back (on disk) of th
hhich a transaction performs a write, and, if the transaction di le old value of any
ceceution, the old values are restored to make it appear as though th loes n
sXeouted. Atomatic ensures by transaction management comp, Rene we
3, Durability : The durability property guarantes not, once a trans,
action complet
es
all updates that carried out on the database persist, even i
he transaction completes execution. To ensure durable thereis a system
4 '
ot Complete its
data on wi
sactions never-
successfully,
failure after tl
1, The updates carried out by the transaction have been
transaction completes. written to disk before the
2, Information about the updates carried out by the transaction
sufficient to enable the database to reconstruct the updates SHEAR
restarted after failure. Durability Ensures by Recovery Mtaiagsinehe de system is.
en
4. 18olation : If consistency and atomicity properties are ensured for cele
‘eral transactions are executed concurrently them operations Se
ble way resulting in an inconsistent state. may interleave some
Two transactions executing concenrrentl i
ly: may interleay ‘
B property.is not obeyed Ta solve it concurrently executing transaniene
ictions is
if'se
undesiral
Example :
value and A +
execute serially.
i
@.3._TRANSACTION STATE
‘pctive! It is the initial state; the transaction stays in this state while itis
executing,
partially committed : After the final statement has been executed,
Fig. 6.1, State diagram of a transaction stored
s been te
‘Aborted : After the transaction has been rolled back and the database D&
to its state prior to the start of the transaction. pally 38 said to be
Committed: A transaction that completes 1t8 execution succes?
committed. eed
no longer PM
Failed : After the discovery that no
ESSING SYSTEM
rmal execution Cc!
targe siz database and
‘s aiway reservation
6.4,_ TRANSACTION PROC!
Transaction processing systems are the systems, which include pias
19 use same database concurrently- Hoes mem iS ‘called transaction
other i
allows many users t
stem oF
system, Banking, Credit card processing SY"
6.5. RECOVE
A computer, like any oth
failure, including power
information pertaining to di
‘A recovery process is an integral Part ©
detection of failures and recovery of database.
‘There are two types of failures *
L. Loss of Volatile Storage : Itcan Dé r
based recovery, Buffer managements
! . Non-Volatile Storage Loss : The most PoP
the entire database to a stable storage perio
database to one or more magn:
js used in restoring the database.
chniques of recovery.
er device, is subject to
failure, disk crash, 5
atabase system is lost:
|
‘There are various te’
ack
6.5.1. Cascading Ro
‘The phenomenon, in which a single transaction
processing system.
RY FROM TRANSACTION FAILURES
failure. There may B
oftware e170!
fa database
covered U!
‘checkpoints
ular
ically For &*
etic tapes. Ifa
(UPTU, 2006,07)
e several cause of such
mY | n,cach of these ceca
system, which is responsible for
various methods. example,
a shadow paging techniqact
nomical scheme is to dum,
ample, we may dump Pp
the most recent dump
sing
an
and €0O!
failure occurs
ip
&
failure leads to a Series of tr
‘ANSAction
rollback, is known as “Cascading Rollback.”
| For example !
1, I y |
Read (X)
Read (Y)
X+Y
write (2) 7
Read (2) s
write (Z) 7
Read (Z)
Transaction T,,
writes a value of Zis read by transaction T.
Pp
writes a value of Z, that is read by transaction T,
2
Transaction >
h suppose that, at this point, 7; fails, 7, must be rolledback, §
ack. Since T, is de
ty must be rolled back. Since T, is dependent on Ty, Ty must also be rollen eet
4.5.2. Recoverable Schedules so be rolled back.
a h
Read (X) |); alee
write (x)
Read (x)
Read (Y)
in this schedule 7, transaction performs only one transaction
suppose that the system allows 7; to commit immediate cage
pstruction. It means 7, commits before 7,.
Now suppose that 7; fails before it commits, Since 7, has alread
(99 written by 7}, So itis necessary to abort 7, to ensure transaction mens the data item
However, 7, has already committed and it cannot be aborted. new
a situation where it is not possible to recover correctly from the faihine we
eo
Hence transaction processing system must ensure that it is pos ‘
the failure of any active transaction Possible to recover from
6.5.3 Log Based Recovery ‘ (UPTU, 2004, 2007, 200
To keep track of the database transactions, the DBMS maintains special fler ai 8)
files or journals that contain information about all updates. The log fle contarne eae o8
tike transaction identifier, type of the log record etc, ‘ains information
since the information contained in the log files are critical for the
two separate copies are maintained. In the past the log files were Hoda aaapeene
fapes. But nowadays, DBMS are expected to recover from minor failures very quick wea
fo the log files are stored online on a fast direct access storage devices (DASD), ae
Since the online log files are also prone to failures, the log files are periodically archi
and are stored in off-line storage, These log files are called “archive ig ai ly archived
ly after excluding the read (x)
arrived in.
Log
‘The most widely used structure for recording database modifications is the log, Th
js a sequence of log records, recording all the update activities in the database, Tete
several types of log records. An update log describes a single database write, It has these
fields:
Transaction identifier is the unique iden
write operation.
bata item identifier is the unique identifier of the data item written, Typically, iti
the location on disk of the data item. .
#old value is the value of the data item prior to the write.
he value that the data item will have after the writer.
records exist to record significant events during transaction processing,
we denote the
ier of the transaction that performed the
#New value is t
Other special log :
nas that start ofa transaction and the commit or abort ofa transaction,
sucl
various type of log records as: Database Management Syste™
T, d valu
* <7, start.> Transaction T; has started. gata item Kp Ee
+ <1LX, Vy, V_>. ‘transaction 7; has performed @ W'S 0
V, before the write and will have V, after the write:
* <7, commit> Transaction 7; has committed
* < Tabort> Transaction 7; has aborted.
It is of two types : (i) Deferred D.M. ae
(i), Immediate D.M- +> aden technique &
(i) Deferred Database Modification : The deterred ein jog, ut defer ng
transactio tomi Jing all database modi faction Pa ;
ransaction atomicity by recording all dav i until the tare’ the final actions oh
eta nique that
execution of all write operations of a alte
Recall that a transaction is said to:be partially.come - odinication Hes
Re eaiaction has been executed. The version of the defer™ ted serially.
Trae tion hie ecction assumes that transactions ATS SS" ace 7, followed by
‘Suppose that these transactions are executed serially, in {PF ook place were $1
na thot the Jalilea of accounts A, Band C vcore thc <1 e, tng the relevene informe
$2000, and $700, respectively. The portion of the le8 OO", orders in which the ac
eae cec tol rar soeione appearsin Agure: there 80 7 saaateiresulUor ene cae
cutputs can take place to both the database system Mm
of Ty and T;.
<7) A,950>
be
(ii) Immediate Database Modification: The immediate modification technique
databhee modifications to be output to the database while the transaction is still
active state. Data modifications written by active transactions are called uncon
failure, the system must u
se
modifications. In the event of a crash or a transaction i
value field of the log records described to restore the modified data items to the val
had prior to the start of the transaction. The undo operation described next, accon
this restoration.
Log Database
< To start >
< Tp B, 2000, 2050 > A= 950
< Ty commit > B=2050
< Tp commit > as)
__ gyre shows one possible order in which the actual outputs t i
mies ‘and the log as a result of the execution of T, Ee ATR ne
gota! te
ints
checkpoin (uPTy,
\ ? 1 i, 200:
658 checkpoint is @ po of synchronization between the database and the transact “
_iffers are force-written to secondary storage at the cl is ction
all heckpoint. Checkpoints
jog file: ied syncpoints or save points.
c
ae pel nedules at predetermined i i
ints are schedules at predetermined interval and involves o
cco gngnelmemnnno messed eer ee erations like writing
all ae transactions are executed serially, when a failure occurs, we check
d the transaction that started before the last checkpoint, ck the log file to
fin
6.5.5 Backup Mechanism
“ base can be used to recov
backup copy of the datal Be
patie database oF incremental copy: An incremental backup consist only me
entire date tne last complete or incremental backup, ee
6.5.6- Shadow Pa! i 9. " (UPTU, 2008)
‘the shadow paging scheme does not require the use of a log file in a singl
est hough a log is needed in multi-user environment for concirenty SRE
ee nen geonsiers the database to bemade up ofa numberof ixed-size disk
or disk blocks for recovery Purposes. es
‘the shadow paging technique maintain two directories (two for each database page)
quring the life of a transaction—a current directory and a shadow directory. When the
transaction starts, the two directories are the same. The shadow directory is saved to the
tip and the current directory is used by the transaction. The shadow directory is never
Ghanged and is used to restore the database in case of a failure.
During the transaction execution, the shadow directory is never modified.
Backup Database
Media Failure
Check points.
Recovered
Restored
Database
Database
Restoring and recovering a database, Z 4 ee
: ABILITY AND-
K-HANDEING
Ya, Inside this Chapter
jalizability of les; 7.
rial 'y of Schedules; 7.2. Types of Schedules; 7.3. festing of Seriali
3, erializability;
7.4. Se
7.4, Deadlock Handling; 7-5. Deadlock Detection; 7.
. ; 7.6. Recovery from Dea
idlock.
7.4. SERIALIZABILITY OF SCHEDULES (UPTU,
1, 2004, 2006) (Imp.)
fwhen several transactions are executing concurre
Re ee deta ntly then the order of execution of various
‘The concept of serializability of schedules is used to identify which schedul
when transaction executions have interleaving of their operations in fe ached’ aaa
schedules.
eee
7.2. TYPES OF SCHEDULES —_
1. Serial schedules.
"9. Non-serial schedules.
3. Conflict schedules.
4, View schedules.
197 e
read-ite™ (X) =
Serial Schedules
1 ID ;
Xext Mi
read:item (X) ,
item (X)
XiKNG write-item (X)
write-item (x)
read-item (Y) 5
YiY+N:
‘waite-item (Y)
‘ \6
item (X) &
ey. aN f waiteitem (Y) ?
peter schedule B
Schedule A
erations of each transact
.dule if the oP*
.d operations from
.ed in serial orde
the other transactic
called serial sche
+ T, and then T, or
‘two schedules Aand Bare :
are executed consecutively, without any interleave’
Ina serial schedule, entire transaction ar¢ performs
and then T,
Non-serial Schedules ‘ ;
‘two schedules C and D are called non-serial schedule if the operations cee
trans ectien are executed non-consecutively with interleaved operations from the ott
transaction.
1
readitem (X) ;
X2XNi
i)
ut
ue
read-item (xX) 7
XiX+Mi
Write-item (X):
read-item (¥) +
write-item (X) :
YiY4N:
waite-item (Y) :
Conflict Schedules : i i
ee les: The graph which contains no cycle is called conflict serializa
If the pre
ae eee graph for Shas a cycle, then schedule S is not conflict serializal
ph contains no cycle, then the schedule Sis conflict serializable.” Serializability and Deadlock Handling
1, For each
the transaction
hn data item %, if transaction 7) executes,
eradtemoXsaftranse
Lee amet chiet ce
must, lle S, also read the iniual valug cognate Si»
, For eae
For each roused by wansaction 7 fan a (A) in sched
z T (if any), then trans: lens
zat ene HAG tno Fa
also rea
hh data item X, the transaction (if
n (if any) th
of view equivalence leads to th PRrratan a
te concept of view equivalence 1 concept of schedule 5
a achedule Sis ew cevilizability fis vew equivalence rag gerbe We say th
ey confct-srializabity schedule 19 also view bari estas ‘
cctisable schedule that are not conflict serializable. wuaable,
> wie z
7a,_DEADLOCK HANDLING aon
(UPTU, 2006, 67,08)
ip aaid to be in deadlock state when in a set oft
i fpother transaction to finish its operations, sive
action 7,
Performs the fi
3, For ea!
final write (x)
coperatior
but there are view
Asystem :
ms, every transaction is
waiting £0)
‘Adeadloc!
‘each waiting fOr
‘There are two Pt
1, Deadlock prevention protocol
ection and recovery process.
{kj a state of statement that may result whe
ig Hild By the other ts be velegacd: 94 Hate
rincipal methods for dealing with deadlock problems»
lore transactions are
9, Deadlock de
_ peadlock Prevention Protocol : This protocol ensures that the
Ziatoce state There ar diferent methods that canbe use for deailoce oe
ig he simplest acheme fequires that each transaction locks allt data roa
starts execution. fata item before it
1p Another method forthe prevention of deadlocks toimpose '
ter mentor vt toe ae ee ne
order
specified by partial order.
‘Another approach for the prevention of deadlock is to use pre-empti /
iption and
transaction rollbacks.
n and Recovery : Ifa system does not imply some protocol that
a
2. Deadlock Detectior
then a detection and recovery scheme must be used.
ensures deadlock freedom,
____—_—__—___—_=EcTION
JETECTION
(UPTU, 2007)
7.5. DEADLOCK DI
Deadlock can be described precisely in terms of a directed graph
This graph consist of a pair G=(V, B), (=)\—€s)
called a wait for graph.
f vertices and Bis a set of edges. Each element
where Vis a set 0}
iges is an ordered pair T;—> 7, If 7, Tis in E, @)
n the set Eof edi
hen there is a directed edge from transaction T; + 7p implying
hat transaction 7; is waiting for transaction T; to release a data (ia)
‘em that it needs. m ifand
‘A deadlock exists in the syster
transaction involved in the cycle is said t
‘This wait for graph contains a eycle: Hence Tao
7.6. RECOVERY FROM DEADLOCK
When a detection algorithm determine:
from the deadlock. The most common
break the deadlock: There actions nee
1. Selection of a Victim
to roll back to break the dea‘
incur the minimum cost.
sol
dt
2. Rollback : Once we have deci
back, we must determine how far
solution is a “total rollback". Abo!
is more effective to roll back the
deadlock. Such partial rollbac
information about the state of all
3. Starvation : In a system where sel 4
tin cost factor it may happen that the some transaction, 12
———
SOLVED EXAMPLES
oman ree 2,
‘Database Manag\
0 be deadlocked.
g that a deadlock exists,
: In this we determine
dock. We should F
ided that a particull
this transaction SH0"
only if the wait for graph con’
Tyqs Taq are all dea
lution is to rollback one or mi
jo be taken : ee
wentin transaction (or ens Cr
stout tions that will
oliback those transact
on-must be rolled
sijd be rolled back. The simplest
ion on and then restart.it. However, it
iy ae far as necessary to break the
Wy Gystem to maintain, additional
tions.
for rollback, is based on
always picked up,
jar transactic
rt the trans
transaction on!
¢ requires th
|| the running transact
lection of transactions,
dule is a conflict serializable ? For each,
EXAMPLE 7.1. Which of the following sche
tested aevlutigable achedule, determine its equivalent serial schedule:
Solution : (a) r,(X); 75(X); Wy; r21%)3 Wal)
q, T, Ty
(X)
1X)
WX)
1(X) c
wx)
Fig. 7.9: NN
or Concurrency Control;
8.5. Graph Based Protocol;
S,omas Write Rule; 8.8,
Modes:
Intension
|
|| e.1. concurrency
| g.3. Mode of Locking; 8.4. es
ing Protocol for Cone pee
orate Sfuitiversion Schemes;
|| 8.6, Time Stamp’
| Protocol;
|| .11. The Multiple-Granularity:
|] 8.13. Multiversion ‘pimeStamp Pro!
@:1._CONCURRENCY CONTROL.
he mi ent of concurrent transaction execution. DBMg
Ht ofp ensure cerialien DI and isolation
and integrity of the database: of
the DBMS, with various methods such
based techniques and optimistic meth oil
jmultaneous operations (querj 3
them interfere with one ancl
hy
currency control is th
implement concurrency control
imp gaction in order to guard the consistency
Concurrency control can be performed by
locking methods, timestamp methods, validation
_ Concurrency control is the process of managing §
updates, inserts, deletes) 07 the database without having,
‘The objective of concurrency control is similar to the objecitve of multi-user c
eystem where many users can performs Sjferent operations at the same tims omputer
systesm the different users are cilowed to perform different actions ninulancoen : in these
sly due to
concept called multi-programming.
@.2,_LOGKING TECHNIQUES FOR CONCURRENCY CONTROL LAG
(UPTU, 2007)
ccess to data. When one tran
ionis
used to control concurrent a
a lock n
lock may be deny access to other transactions to pre|
vent incorrect
Cont
Lockingis a procedure
accessing the database,
results. j
208 Concurrency Control Techniques
he most widely used
jgone ofthe most widely used mechanism t
sas follows : A transa 0 ensure se,
isa A transaction must obtain a reaq 2
ying
ie of loa crform a read or writ ; s
priciPle van Pen rite operation. add cart
ag jes for locking are :
ity. The working
clock on a dats
erul
ihe basi
he basi action has aread lock on a data item, i
it can read th
€ item but not
Update
action has a read lock on a data it
item, other trar
Nnsactio,
a trans
2. ead rapion the data item, but not write locks, nS Can obj
: tain a
ay ata transaetion has a write lock on‘a data item. It can both
ata iter Fead and update th
tion has a write lock on a data i 7
ans k ata item,
4, Ifa wanetea read lock or a write lock on the data tienes cana cannat
tem. oe
vamonly used technique to control the concur
saost emash look on a data lem species erurency is based on the conceptot
jock operations like oe id anawrite that can be Applet Sibsite eee a ae
opersrem, a lock must be associated with that item. Ifa ‘a transaction Twants to
iy Gy accesses the same data item, 7, has to wait unt 7 ges up theiecn GH
7, locks the data item. ? gives up the lock. Otherwise
{ODE OF LOCKING a
‘two modes in which a data item may be locked :
(UPTU, 2004)
rere are mainly
1, shared lock
2, Exclusive lock.
1.9.1. Shared Lock
i, onsection T,has obtained 4 shared znbde lodicon ier
item Q, then 7, can read
" but
annot write Q. Itis denoted by (S)-
on
1.3.2. Exclusive Lock
*y, ransgction Tjhas obtained én ex-tusive-mode leceon
‘write 0. It is denoted by (%)- tem O, then jean both read
nd
THE TWO-PHAS! PI es
iE E LOCKING PROTOCOL —— = (wrrv, 2004, 2006, 2008)fmp,)
4.
ing protocol ensures serializability. This protocol requires that each
ne two-phase lockis
he faction issue lock and unlock requests in éwo phase:
1, Growing phase
2, Shrinking phase
1, Growing Phase : In this phase a transaction may obtain locks but may not release
sy lock. Here the no. of locks increases from 2¢r9 to the maximum for the transaction.
2. shrinking Phase: In this phase a transaction may release locks but may not obtain
wy new locks. Here the no. of locks decreases from maximum to zero. Database Management S
7 There are three types of two-phase Jocking protocol
0) Conservative Two-phase Locking-
(p) Strict Two-phase Locking.
(9 Rigorous ‘Two-Phase Locking.
8.4.1. Conservative two-Phase Locking
It requires @ transaction to predeclare all the locks it requires before its execution.
Ifany of predared items cannot be locked the transaction waits until all item are lockeg,
free but it is difficult to use in practice. i
the locking protocol is deadlock
8.4.2. Strict Two-Phase Locking
In this protocol a transaction does not release any ofits write locke untill it commits
aborts. or
Rigorous two-Phase Locking
ck and read lock untip j
it
tocol, transaction does not release 2nY of write lo
8.4.3.
In this prot
commits or aborts.
cor
@.5._GRAPH-BASED PROT — PROTOCOL
rderingon the set D= {d,, d.
tap Gan 14} Of aly
8.5._GRAPH-BA!
owledge, we impose a partial 01
tien accessing both d, and d,must
MUST ACCESS defor,
To acquire prior kné
transact
|
data item. If d, > 4 then any
5 that the set D may now be viewed as a dir
ected
Acyclic
accessing d;
in this section, for the sake the simplicity, we wi
» we wil
ll restricy
‘The partial ordering implie
graph called a database graph.
o rattention to only those graphs that are rooted trees.
‘Tree protocl is restricted to employ only exclusive locks.
only lock instruction allowed is lock - X. Each
4 transact
ion 7
In the tree protocol, the
can lock a data item at most once, and must observe the following rules :
1, The first lock by T;may be on any data item.
2, subsequently, a data item Q can be locked by T;only if the parent of Qi
it is
currently
locked by 7;-
3, Data items may be unlocked at any time.
4. Adata item that has been locked and unlocked
by 7,
‘The tree-locking protocol has an advantas
: ; 2 wes over the two-phi i
that, unlike two-phase locking, it is deadlock-free so no Behackoe seo Protocol in|
pe te oe protocol has another advantage over the two-phas ns ie
unlocking may occur earlier. Earlier unlocking may lead to ceate ae Cae
iting tim
8, and
to an increase in concurrency.
However, i
. the protocol has the disadvantage that in some cases, trans:
, transaction ma
ty have:
by T,cannot subsequent _
is tly be relockeg
to lock data items that it does not access. ‘Concurrency Control Techniques en
76_TME STAMPING PROTOCOLS FOR CONCURRENCY CONTROL —~
sjampingis-« concurrency control protocolin which the Sarna
peccgtcienent ce ant ote aner
then. areas oF apie onesation Je allowed, only if the last updata on iubdae rae
ihe by an oer ansacton. Otherwise the ansacon,requentin ee
amperes irom continuously aborting and restating See eer
sus method sequence of transaction is selected in advance by using the
wrote ses tn Sr ce
1 system clock value, ie :
neat mestamp value, We can also use a logical counter that i incremented after aiog
toe gia transaction 7 has ben atagneda timestamp 9, and 9 sc eanbae
ater ro he sytem witha time stamp value 8, then §,« Sand ayte il allow tha
1enfon 7, torn fst followed by the transaction 7, i sia aenaae aaa
ifany cont
une conflicting
‘there are two
4, wetime stamp (Q)
2, ReTime stamp (Q)
4, wetime stamp (Q}: Denotes the largest time stamp of any transaction
arite (Q) successfully transaction that executed
ig. Rtime stamp (Q}: Denotes the largest time stamp of any transact
a ea yy transaction that executed
> AS WRITE RULE
THOMAS WRITE RULE
B.7-
“pe modification tothe time stamp-ordering protocol, called ‘Thomas write rule
‘Thomas write rae provide a modification to the timestamp ordering protocl
reater potential concurtency, thatallows
uppose that transaction 7, issties write (Q)
Re a ae a ‘of O that Tis producing wet Teen
eded and it had been assumed that the value would never be oud
thc aystem rejects the write operation and rools 7, back, Hence,
a Was (t) < Welime stamp (Q ten Ts attempting o write an absolute value of
Hence, this write operation can be ignored. .
4, otherwise, the system executes the write operation and sets W-timestamps (9) to
TS(T}
ction is found between transaction using ti
transactions is rolled back. 2 Snestanp badeslgyster/ ond at
types of timestamps. Not
EB
child nodes.
before,
belttesent in more than one file.
‘each node in the tree can be locked individual
ly.
propriate granularity
In multiple granularity locking, each transa ey avon moseert
locking, each iction.
2 mode of operations. For this sc y bane y:
heme we may consider adatabace eet hic
1¢ hierarchy of
‘Database Management Syst
@.6._ VALIDATION-BASED PROTOCOLS ae cee’
‘The two concurrency control techt control, another c
‘control techniques, One is locking based concurrency access
rae cone ing baced concurrency technique. Butin both the echnid ey eel See ee ene ee IaeaLinieeie
levels of drawbacks. In locking based techniques, the item being accessed must PT a Fre nmmber of 10K 1 be maintained would Increase causinghigher overhead: Méreoes
‘Aiso, here it performs the operations in two steps: In first tery it locks @) At hea more strat space is required as per example forread and, ae
Which are required in transaction, and if all the items 9% Tocked success)" Dane peqoey Buasautblei cece ce Lae aga neincrease grteumeatampfoceachidsta
aarcy phase, the data tems are modified. n timestamp bases {eennIaUS cee hence ‘
's assigned a timestamp value and this transaction “aimeatamp 18 cheCKES TT stow ined
sai tmestamps of the and io ee hes ease =a
jese overheads.
fata item directly to 4 DB—Datab:
+ lan Hon wae eBncIGAe 82 HOE EPIRIS 2 tog gat Sa er
datahase, rather It keeps local copies ofall the UPdor dari ty of tion ang Ree
a ereeineaction reaches its end, a validation POS) checks te te Io ey ote call Long agate
items due to the “updates, and ifit is consistent #heP ‘all the da Beg, faerntas® Record
to the database. tion any d :
¢ transac ata item
Reading is done when required and if at te ead of the caine
found to be inconsistent then the transaction {rolled back an cg
optimistic) based coneurTen cy control technique, "2
: tree consist of four levels of nodes.
sno tree conn 1s of nodes. The highest level re}
pelow itare m type area; the database consists of exactly these mean Bachar
turn nas nodes of Kisabitachiltentcatweea tte me ati
(file is in fea contains “thar apeite
sie etcn marta ea Ry ehcp
les, and no record can
three phase for validation
‘committed values from the dat
eee vd and reads the last’
and purihesevalacs 8 to aden ta oly nese el este
0.8.2. Validation Phase
chine Pree at
wn is activat
focal variabl
values during the first phase
8ainy
eified database should
reat
forit
the various modes.
fa node is locked in an intension m
ae ce:
tf , explicit locking i
ilenin ckennn odie ae eee
entire tree to determine whethe explicity. Thus, a transaction d on all the ancestors of a
ffek a node say (Q) must ee ee eeatal At ae
werse a path in the tree from the zoo eas wishing to
© (Q). Whil g
mode. i raveraiogs
. for the read ¥
In this phase che tems are
the current values. Also it
the database in a consist
tent mog,
consistent.
8.8.3. Write Phase .
«passed second phase leaving eo is
ated othe database permanently Iin
f serializability, so if discards, elle
ithe transaction bal nd
ln raeade bythe eansacton ar
ten up age sin connie Wal
.d transaction is ‘restarted.
Janiques all the checks or validations are the tree, the transaction locks th
pa | eee Te ie yadtts naner Rane
10, INTENSION MODES
all the updates an‘
rrency control tec
‘ead phase, transac
In optimistic concu
"Before that in
e transactions that access
oo 160856 the databy|_ There are three types of intension modes:
1. Intention-Shared (IS) mode.
once in validation phas
overhead.
‘So, this technique is very much suitable foF
for andi only operations and modifies disjoints sets of data item.
(Orr a 2, Intention-Exclusive (IX) mode.
3, Shared and Intention-Exclusive (SIX) mod:
le:
Ipention-Shared (1S) Mode
eS eS
MULTIPLE GRANULARITY
data item being locked. Granularity is consi
it can effect the performance of con wel 9.40.4. I
Ifa node is locked in intention-+
ntion-shared (IS) mode, explicit locking is being d
. ig done at a lower
a9.
By granularity we mean is the size of the
factor concurrency control because i
data item is very large then overhead
@
evel of the tree, but with only shared, mode I
: lor:
as the major:
na recovery. Because, if the granularity of a
locking is very low but it will effect the concurrency. nvion-Exctusive (1X) Mi
wee mtention exctuave FX) Teds a explicit ocking is being done
atte or ahared-mode Jos q
jusive (SIXI Mode
cclusive
explicit locking
oe, Ite
oa, shared and intent
anette exci sine mete SSS
tne de remo ok
er GATE CRANURI OEE TTS TIPLE SEAMUDATT LOOMIS ST
cen regimen lai Se Toby te fllwine hese ral
a 1 Temust observe the lock-comt
2 mpc ne ot ote
2 nese ye Gns =e ay
ss teste
anode 91218 Se
so ix ede
is associated. .
‘contains three data fit
{me stamp before the tra
ynsaction
£oh version Or
i content
2} ime star (On)
2 nesta (On)
RECOVERY WITH CONCURRENT TRANSAGT
ION
5
ra conden ee
saved netion with Concurrency Coenen the various
. ends greatly on the concur In this sct arava
Ee sy cnttshee Ont
tion Rebck in bis chene wera shee ii eS ras
Tranter saa te of backward or rey (eahsacton yy an
dow restores the da te 87 log record found be eee
4, Dueck points: this scheme we used check nthe log the
ect Fate system must scan when It recovers froma veg
.saction-processing system, we ascovers from al
we require that the el ‘rast
We alte oreecte eee
(UPTV, 2008
the number
nce
og record be othe
ictions active at the time of the
ana epomnt E>, where
Srcck pint
che point sa checkpoint where tea
Ae butler blocks are being written out.
A ene Recovery When the system recover
The unde ee redone, The aystem constructs the fee
tranaaet uh emply. The system scans the log backward,
ina it Binds the first record, °& backward,
lowed to perform updates
iteonatrt
sol
thereat cama
st flows Ia,
mining each record,
pxampLe 8.1, Show that the two-phase locking prot
ett ni tae a
ons suppose phe ing Ss oo ‘crdlibiity, 1a
sae sacle Ty Tyo Ty Which obey 2FL and w Then there ei
at enelanbe schedule ples a len 2 precedence py 1on-eliabe
a eee prace such yes, Witham greta ee eae a
snr iene precedence graph Ty» 7, — v7 ality assume the phos
Fr re eck poh The Bo Re a ten
‘Then for the cycle we have en for all transactions. aan ]
Oy
