See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/265350441

Horizontal Fragmentation Technique in Distributed Database

Article

CITATIONS READS
28 2,659

3 authors, including:

Avinash Gawande Amar Deshmukh

Sipna College of Engineering and Technology Amravati RMD Sinhgad College Of Engineering
23 PUBLICATIONS 112 CITATIONS 4 PUBLICATIONS 42 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Avinash Gawande on 10 December 2015.

The user has requested enhancement of the downloaded file.

International Journal of Scientific and Research Publications, Volume 2, Issue 5, May 2012 1
ISSN 2250-3153

Horizontal Fragmentation Technique in Distributed

Database
Ms. P. R. Bhuyar
M.E. I st Year (CSE)
Sipna College of Engineering & Technology, Amravati, India

Dr.A.D.Gawande
HOD (CMPS)
Department of computer science & Engineering
Sipna College of Engineering & Technology, Amravati, India

Prof. A.B.Deshmukh
Professor (CMPS)
Department of computer science & Engineering
Sipna College of Engineering & Technology, Amravati, India

Abstract- Distributed database technology is expected to have a of a computer network. It is not necessary that database system
significant impact on data processing in the upcoming years. have to be geographically distributed. The sites of the distributed
Today’s business environment has an increasing need for database can have the same network address and may be in the
distributed database and Client/server applications as the desire same room but the communication between them is done over a
for consistent, scalable, reliable and accessible information is network instead of shared memory. As communication
steadily growing. Distributed processing is an effective way to technology, hardware, software protocols advances rapidly and
improve reliability and performance of a database system. prices of network equipments falls every day, developing
Distribution of data is a collection of fragmentation, allocation distributed database systems become more and more feasible.
and replication processes. Previous research works provided Design of efficient distributed database is one of the major
fragmentation solution based on empirical data about the type research problems in database & information technology areas.
and frequency of the queries submitted to a centralized system.
These solutions are not suitable at the initial stage of a database A distributed database management system (DDBMS) is then
design for a distributed system. The purpose of this work is to defined as the software system that permits the management of
present an introduction to Distributed Databases which are the DDB and makes the distribution transparent to the users.
becoming very popular now days with the description of Distributed database system (DDBS) is the integration of DDB
distributed database environment, fragmentation and horizontal and DDBMS. This integration is achieved through the merging
fragmentation technique. Horizontal fragmentation has an the database and networking technologies together. Or it can be
important impact in improving the applications performance that described as, a system that runs on a collection of machines that
is strongly affected by distributed databases design phase. In this do not have shared memory, yet looks to the user like a single
report, we have presented a fragmentation technique that can be machine. Assumptions regarding the system that underlie these
applied at the initial stage as well as in later stages of a definitions are:
distributed database system for partitioning the relations.
Allocation of fragments is done simultaneously in the algorithm. 1. Data is stored at a number of sites. Each site is assumed
Result shows that proposed technique can solve initial to logically consist of a single processor. Even if some
fragmentation problem of relational databases for distributed sites are multiprocessor machines, the distributed
systems properly. DBMS is not concerned with the storage and
management of data on this parallel machine.
Index Terms- Distributed database, Fragmentation, Horizontal
Fragmentation, Allocation.

I. INTRODUCTION

1.1 Distributed Database System

A distributed database (DDB) is a collection of data that

logically belongs to the same system but is spread over the sites

www.ijsrp.org
International Journal of Scientific and Research Publications, Volume 2, Issue 5, May 2012 2
ISSN 2250-3153
information that is used to query at the site and the information at
the site should be fragmented so the queries of the site run faster.
Horizontal fragmentation is defined as selection
operation, σ _p(R).
For example, the following relation
EMPLOYEE (eid, fname, lname, site ,pos, salary)

Eid Fname Lname Site Pos Salary

Fragment1

Fragment2

Fragment3
Figure 1: A Distributed Database Environment Figure 2: Horizontal fragmentation

2. The processors at these sites are interconnected by a 1.2.2. Vertical Fragmentation

computer network rather than a multiprocessor Vertical fragmentation (VF) allows a relation or class to be
configuration. partitioned into disjoint sets of columns or attributes except the
3. To form a DDB, distributed data should be logically primary key. Each partition must include the primary key
related, where the relationship is defined according to attribute(s) of the table. This arrangement can make sense when
some structural formalism, and access to data should be different sites are responsible for processing different functions
at a high level via a common interface. The typical involving an entity.
formalism that is used for establishing the logical Objective of vertical fragmentation is to partition a relation
relationship is the relational model. into a set of smaller relations so that many of the applications
4. The system has the full functionality of a DBMS. will run on only one fragment.
a. Vertical fragmentation of a relation R produces
Distributed processing on database management fragments R1,R2, . . . , each of which contains a subset
systems (DBMS) is an efficient way of improving of R’s attributes.
performance of applications that manipulate large volumes b. Vertical fragmentation is defined using the projection
of data. This may be accomplished by removing irrelevant operation of the relational algebra:
data accessed during the execution of queries and by П _A1,A2,...,An(R)
reducing the data exchange among sites, which are the two
main goals of the design of distributed databases. Primary Eid Fname Lname Site Eid Pos Salary
concern of distributed database system design is to making
fragmentation of the relations in case of relational database
or classes in case of object oriented databases, allocation and
replication of the fragments in different sites of the Fragment1 Fragment2
distributed system, and local optimization in each site.

1.2 Fragmentation
Primary concern of distributed database system design Figure 3: Vertical fragmentation
is to making fragmentation of the relations in case of relational
database or classes in case of object oriented databases, 1.2.3. Hybrid fragmentation
allocation and replication of the fragments in different sites of the Combination of horizontal and vertical fragmentations
distributed system, and local optimization in each site. is mixed or hybrid fragmentations (MF). In this type of
Fragmentation is a design technique to divide a single relation or fragmentation scheme, the table is divided into arbitrary blocks,
class of a database into two or more partitions such that the based on the needed requirements. Each fragmentation can be
combination of the partitions provides the original database allocated on to a specific site. This type of fragmentation is the
without any loss of information. This reduces the amount of most complex one, which needs more management, in most
irrelevant data accessed by the applications of the database, thus cases simple horizontal or vertical fragmentation of DB
reducing the number of disk accesses. Fragmentation can be applications.
horizontal, vertical or mixed/hybrid. Mixed fragmentation (hybrid fragmentation) Consists of
a horizontal fragment followed schema will not be sufficient to
1.2.1. Horizontal fragmentation satisfy the requirements of the by a vertical fragmentation, or a
Horizontal fragmentation (HF) allows a relation or class vertical fragmentation followed by a horizontal fragmentation.
to be partitioned into disjoint tuples or instances. Intuition behind Mixed Fragmentation is defined using the selection and
horizontal fragmentation is that Every site should hold all projection operations of relational algebra:
П_p(_A1,...,An(R))
www.ijsrp.org
International Journal of Scientific and Research Publications, Volume 2, Issue 5, May 2012 3
ISSN 2250-3153
П _A1,...,An(_p(R)) affinity graph thus defines horizontal class fragments [4].
Navathe et al. used attribute usage matrix (AUM) and Bond
energy algorithm to produce vertical fragments [6]. Shin and
Salary Eid Fname Lname Site Pos
Irani proposed knowledge based approach in which user
reference clusters are derived from the user queries to the
Fragment1 Fragment2 database and the knowledge about the data [7]. Ra presented a
graph based algorithm for HF in which predicates are clustered
based on the predicate affinities [8]. Cheng et al. presented a
genetic algorithm based fragmentation approach that treats
Fragment3 Fragment4 horizontal fragmentation as a travelling salesman problem [9].
Ma et al. Used an attribute uses frequency matrix (AUFM) and a
cost model for VF [10]. Alfares et al. used AAM to generate
Figure 4: Hybrid fragmentation groups based on affinity values [11]. Marwa et al. uses the
instance request matrix to horizontally fragment object oriented
The main reasons of fragmentation of the relations are database [12]. Abuelyaman proposed a static algorithm StatPart
to: increase locality of reference of the queries submitted to for VF [13]. Mahboubi H. and Darmont J. used predicate
database, improve reliability and availability of data and affinity for HF in data warehouse [14].
performance of the system, balance storage capacities and To the best of our knowledge, only Abuelyaman [13]
minimize communication costs among sites. provided a solution for initial fragmentation of relations of a
Previous techniques of HF, VF or MF have the following distribution database. A randomly generated reflexivity matrix, a
problems in common: symmetry matrix and a transitivity module has been used to
 They use frequency of queries, midterm predicates’ produce vertical fragments of the relations and no algorithm for
affinity or attribute affinity matrix (AAM) as a basis of horizontal fragmentation. But he could not justify his hypothesis
fragmentation. These require sufficient empirical data that why it will produce good fragments.
that are not available in most cases at the initial stage.
 Most of them concentrate only fragmentation problem
and overlooked allocation problem to reduce complexity III. RELATED WORK
To solve the problem of taking proper fragmentation
Allocation is the process of assigning the fragments of a decision at the initial stage of a distributed database, we have
database on the sites of a distributed network. When data are provided a new technique of fragmentation. That is to fragment a
allocated, it may either be replicated or maintained as a single relation horizontally according to locality of precedence of its
copy. The replication of fragments improves reliability and attributes. Attribute locality precedence (ALP) can be defined as
efficiency of read-only queries but increase update cost the value of importance of an attribute with respect to sites of
distributed database. ALP table will be constructed by database
In this report, we have presented a new technique for designer for each relation of a DDBMS at the time of designing
horizontal fragmentation of the relations of a distributed the database with the help of modified CRUD (Create, Read,
database. This technique is capable of taking proper Update, and Delete) matrix and cost functions. A block diagram
fragmentation decision at the initial stage by using the of our system is depicted in Figure 5.
knowledge gathered during requirement analysis phase without
the help of empirical data about query execution. It can also Relation CRUD Matrix MCRUD Matrix ALP Table
allocate the fragments properly among the sites of DDBMS.

Allocation Fragmented Sub-Relations Predicate Set

II. LITERATURE REVIEW
Distributed databases are not new, nor are they a Figure 5: Block diagram of the system
consideration unique to client/server architectures or relational
databases. Data distribution needs, no doubt, arose immediately A relation in a database contains different types of
after the first database management systems appeared 30 years attributes those describe properties of the relation. But the
ago, and various solutions to the distribution problem have been important thing is that the attributes of a relation do not have
implemented over the years on mainframe and minicomputer same importance with respect to data distribution in different
platforms using a wide variety of database management software. sites. According to above importance we can calculate locality
HF using min-term predicate is first proposed by Ceri et precedence of each attribute for each relation and construct ALP
al.[5]. Ozsu and Valduriez proposed an iterative algorithm table for the relations.
COMMIN to generate a complete and minimal set of predicates
from a given set of simple predicates [1]. Navathe et al. proposed 3.1 CRUD Matrix
a MF technique. The input of the procedure comprises a A CRUD ( data-to-location ) matrix is a table of which
predicate affinity table and an attribute affinity table [3]. Bai˜oo rows indicate attributes of the entities of a relation and columns
et al. inputted predicate affinity matrix to build a predicate indicate different locations of the applications (processes that
affect those attributes). If a particular process uses a particular
www.ijsrp.org
International Journal of Scientific and Research Publications, Volume 2, Issue 5, May 2012 4
ISSN 2250-3153
entity attribute, the appropriate cell is filled in with the letters C, is that at the design time of a distributed database, the designer
R, U, or D.A "C" in the cell of a CRUD matrix indicates that the will not know the actual frequencies of read, delete, create and
process sometimes creates new instances of the corresponding update of a particular attribute from different applications of a
entity type. An "R" in the cell indicates that the process site and generally update incurs more cost than create and delete,
sometimes reads existing instances of the entity type. A "U" in and reading from database always incurs least cost.
the cell indicates that the process sometimes updates instances of
the corresponding entity type. A "D" in the cell indicates that the After construction of ALP table for a relation, predicate
process sometimes deletes instances of the corresponding entity set P will be generated for the attribute with highest precedence
type. value in the ALP table. Finally each relation will be fragmented
A process does not necessarily use an entity every time horizontally using the predicates of P as selection predicate. The
it occurs. This does not mean that the interaction should not be procedures can be clearly understood from the following
shown on the CRUD matrix. If the process ever uses the entity, algorithm and pseudo code of Fig 6 and 7.
the interaction must be documented in the CRUD matrix. A
CRUD matrix is used by the system analysts and designers in the Input: Total number of sites: S = {S1, S2,… ,Sn}
requirement analysis phase of system development life cycle for Relation to be fragmented: R
making decision of data mapping to different locations. Modified CRUD matrix: MCRUD[R]
MCRUD Matrix - We have modified the existing CRUD matrix Output: Fragments F = {F1, F2, F3,…, Fn}
according to our requirement of HF and name it Modified Create, Step 1: Construct ALP[R] from MCRUD[R] based on
Read, Update, and Delete (MCRUD) matrix. It is a table Cost functions
constructed by placing predicates of attributes of a relation as the Step 2: For the highest valued attribute of ALP table
rows and applications of the sites of a DDBMS as the columns. a. Generate predicate set P={ P1, P2, … ,Pm }
We have used MCRUD to generate ALP table for each relation. b. Rearrange P so that #P = #S
We treated cost as the effort of access and modification c. Fragment R using P as selection predicate
of a particular attribute of a relation by an application from a
particular site. For calculating precedence of an attribute of a p l σp
relation we take the MCRUD matrix of the relation as an input d. Allocate F to S
and use the following cost functions:
Figure 6: Fragmentation Allocation algorithm
Ci,j, k, r = fCC + fRR + fUU + fDD (1)
Ai,j,k Input: MCRUD of a relation that to be fragmented
Si,j,k= ∑Ci, j, k, r (2) Output: ALP table for that relation
r=1
Si,j,m =Max(Si,j,k) (3) for ( i =1; i <= TotalAttributes; i++)
Ai,j,k {
for ( j =1; j <= TotalPredicates[i]; j++)
ALPi,j= Si,j,m -∑Si,j,k (4)
{
k≠m
MAX[i][j] = 0;
l for ( k =1; k <= TotalSites; k++)
ALPi= ∑ALPi,j (5) {
j=1 for ( r =1; r <= TotalApplications[k]; r++) /* Calculating
sum of all applications” cost of predicate j of attribute i at site k */
where fC = frequency of create operation {
fR = frequency of read operation C[i][j][k][r] = fc*C + fr*R + fu*U + fd*D
fU= frequency of update operation S[i][j][k] + = C[i][j][k][r]
fD = frequency of delete operation }
C = weight of create operation If S[i][j][k] > MAX[i][j] /*Find out at which site cost of
R = weight of read operation predicate j is maximum*/
U = weight of update operation {
D = weight of delete operation MAX[i][j] = S[i][j][k]
Ci,j, k, r= cost of predicate j of attribute i accessed by POS[i][j] = k
application r at site k }
Si, j, k = sum of all applications’ cost of predicate j of SumOther = 0
attribute i at site k for ( r =1; r <= A[i][j][k][k]; r++)
Si, j, m = maximum cost among the sites for predicate j {
If (r!=k)
of attribute i
SumOther + = S[i][j][r]
ALPi j = actual cost for predicate j of attribute i
}
ALPi = total cost of attribute i (locality precedence) }
ALPsingle[i][j] = S[i][j][POS[i][j]] – SumOther /* actual
For simplicity we have assumed that fC, fR, fU and fD=1 cost for predicate j of attribute i */
and C=2, R=1, U=3 and D=2. The justification of the assumption }

www.ijsrp.org
 International Journal of Scientific and Research Publications, Volume 2, Issue 5, May 2012 5
ISSN 2250-3153
ALP[i] = 0 cost functions of equation (1)-(5). Calculating the locality
for ( j =1; j <= TotalPredicates[i]; j++) /*calculating total cost precedence of the attribute Branch is shown in Figure 9.
for attribute i (locality precedence)*/ According to the cost functions, value of the predicate
{ Branch=Pune is (8+4+8) - (1+1) = 18, Branch=Nagpur is
ALP[i] + = ALPsingle[i][j] (8+8+1) – (1+1) = 15 and Branch=Mumbai is (8+3+6) – 0 = 17.
} So ALP of Branch = 18+15+17 = 50.
}
Site
Figure 7: ALP-table-construction Pseudo-code Application Site1 Site2 Site3

To justify our technique, we have implemented a

distributed banking database system. One of the relations of the Ap Ap Ap Ap Ap Ap Ap Ap
Entity,Attribute.Predica Ap3
1 2 3 1 2 3 1 2
database is Accounts shown in Table 1. Initially number of sites te
Accounts.AccNo<1000
of the distributed system is three. 0
C RU R
Accounts.AccNo>=100
R
Table 1: Accounts Relation 00
CR RU
Accounts.Type=Saving RU R
D D
AccNo Type CustId OpenDate Balance Branch Accounts.Type=Curren CR
01 Saving 101 05/01/12 30000 Pune RU R RU R
t UD
02 Current 102 18/01/12 48000 Pune .
.
03 Current 103 10/02/12 15900 Nagpur .
04 Saving 104 06/03/12 37750 Mumbai Accounts.Balance<500
R R
CR
R
00 UD
05 Current 105 12/03/12 50000 Pune Accounts.Balance>=50
CR
06 Saving 106 25/03/12 25000 Nagpur 000
CR CR
07 Current 107 28/03/12 45000 Mumbai Account.Branch=Pune
UD
RU
UD
R R
Account.Branch=Nagp CR CR
R R R
3.2. Construction of MCRUD Matrix ur UD UD
Account.Branch=Mum CR
We have constructed the MCRUD matrix for the bai UD
RD CRU
Accounts relation in the requirement analysis phase. Part of Figure 9: ALP cost for Branch=Pune
MCRUD matrix is shown in Figure 8.

Site
3.4. Construction of ALP Table
Application
Site1 Site2 Site3
ALP values of all the attributes of the Accounts relation
was computed from its MCRUD matrix. The attribute with
highest precedence value will be treated as most important
Ap Ap Ap Ap Ap Ap Ap Ap Ap
Entity,Attribute.Predica
1 2 3 1 2 3 1 2 3
attribute for fragmentation. Table 2 shows the ALP table for
te
Accounts relation.
Accounts.AccNo<1000
C RU R
0
Accounts.AccNo>=100
R Table 2: ALP table for Accounts relation
00
Accounts.Type=Saving
CR
RU
RU
R Attribute Name Precedence
D D
Accounts.Type=Curren CR AccNo 6
RU R RU R
t UD Type 22
.
.
CustId 6
. OpenDate 7
Accounts.Balance<500 CR
00
R R
UD
R Balance 10
Accounts.Balance>=50
CR
Branch 50
000
CR CR
Account.Branch=Pune RU R R 3.5. Generation of Predicate Set
UD UD
Account.Branch=Nagp
R
CR CR
R R Predicate set was generated for Branch, the attribute
ur UD UD
Account.Branch=Mum CR CR
with highest locality precedence of Accounts relation.
RD
bai UD U P= {p1: Branch=Pune, p2: Branch=Nagpur, p3:
Branch= Mumbai}
Figure 8: MCRUD matrix of Accounts
3.6. Fragmentation of Relation
According to the predicate set P, Account relation was
fragmented and allocated to 3 sites (figure 10) shown in table 3-
3.3 Calculation of ALP 5.
We have calculated locality precedence of each attribute
from the MCRUD matrix of Accounts relation according to the

www.ijsrp.org
International Journal of Scientific and Research Publications, Volume 2, Issue 5, May 2012 6
ISSN 2250-3153
researchers using empirical knowledge of data access and query
frequencies. But proper fragmentation and allocation at the initial
stage of a distributed database has not yet been addressed. In this
report, we have presented a fragmentation technique to partition
Pune Site1 Site1 Mumbai relations of a distributed database properly at the initial stage
when no data access statistics and query execution frequencies
are available. Using our technique, no additional complexity is
Site1 added for allocating the fragments to the sites of a distributed
database as fragmentation is synchronized with allocation. So
Nagpur performance of a DDBMS can be improved significantly by
avoiding frequent remote access and high data transfer among
Figure 10: Distributed banking database system the sites. This work can be extended to support fragmentation in
distributed object oriented databases as well.
Table 3: Part of Accounts relation allocated to site 1
REFERENCES
Acc
Type CustId OpenDate Balance Branch [1] M. T. Ozsu and P. Valduriez, Principles of Distributed Database Systems,
No
2nd ed., New Jersey: Prentice-Hall, 1999.
01 Saving 101 05/01/12 30000 Pune [2] S. Ceri and G. Pelagatti, Distributed Databases Principles and System, 1st
02 Current 102 18/01/12 48000 Pune ed., New York: McGraw-Hill, 1984.
05 Current 105 12/03/12 50000 Pune [3] S. Navathe, K. Karlapalem, and M. Ra, “A mixed fragmentation
methodology for initial distributed database design,” Journal of Computer
and Software Engineering Vol. 3, No. 4 pp 395–426, 1995.
Table 4: Part of Accounts relation allocated to site 2 [4] F. Bai˜ao, M. Mattoso, and G. Zaverucha, “A distribution design
methodology for object DBMS,” Distributed and Parallel Databases,
Acc Springer, Vol. 16, No. 1, pp. 45–90, 2004.
Type CustId OpenDate Balance Branch
No [5] S. Ceri, M. Negri, and G. Pelagatti, “Horizontal data partitioning in
04 Saving 104 06/03/12 37750 Mumbai database design,” in Proc. ACM SIGMOD, 1982, pp. 128–136.
[6] S. B. Navathe, S. Ceri, G. Wiederhold, and J. Dour, “Vertical partitioning
07 Current 107 28/03/12 45000 Mumbai algorithms for database design,” ACM Transactions on Database Systems
(TODS), Vol. 9, No. 4, pp. 680–710, 1984.
[7] D. G. Shin, and K. B. Irani, “Fragmenting relations horizontally using a
Table 5: Part of Accounts relation allocated to site 3 knowledge based approach,” IEEE Transactions on Software Engineering
(TSE), Vol. 17, No. 9, pp. 872–883, 1991.
Acc [8] M. Ra, “Horizontal partitioning for distributed database design,” In
Type CustId OpenDate Balance Branch
No Advances in Database Research, World Scientific Publishing, pp. 101–120,
03 Current 103 10/02/12 15900 Nagpur 1993.
[9] C. H. Cheng, W. K. Lee, and K. F. Wong, “A genetic algorithm-based
06 Saving 106 25/03/12 25000 Nagpur clustering approach for database partitioning,” IEEE Transactions on
Systems, Man, and Cybernetics, Vol. 32, No. 3, pp. 215–230, 2002.
From the above result, we can see that our technique has [10] H. Ma, K. D. Schewe, and M. Kirchberg, “A heuristic approach to vertical
successfully fragmented the Accounts relation and allocated the fragmentation incorporating query information,” in Proc. 7th International
Baltic Conference on Databases and Information Systems (DB&IS), 2006,
fragments among the sites of the distributed system. As we have pp. 69–76.
only taken highest valued attribute from ALP table, no unwanted [11] M. AlFares et al, “Vertical Partitioning for Database Design: A Grouping
fragments were created. Other relations of the distributed Algorithm”, in Proc. International Conference on Software Engineering and
banking database can be fragmented in the same way like Data Engineering (SEDE), 2007, pp. 218-223.
[12] F. F. Marwa, I. E. Ali, A. A. Hesham, “A heuristic approach for horizontal
Accounts. fragmentation and alllocation in DOODB,” in Proc. INFOS2008, 2008, pp.
For simplicity we have considered only four sites of the system 9-16.
for allocation. It is worth mentioning that our fragmentation [13] E. S. Abuelyaman, “An optimized scheme for vertical partitioning of a
technique will work in the same way for large number of sites of distributed database,” Int. Journal of Computer Science & Network
Security, Vol. 8, No.1, 2008.
any distributed system. [14] H. Mahboubi and J. Darmont, “Enhancing XML Data Warehouse Query
Performance by Fragmentation,” in Proc. ACM SAC09, 2009, pp.1555-
1562.
IV. CONCLUSION [15] Haroun Rababaah, Dr. H. Hakimzadeh ,“Distributed
Databases:Fundamentals and research”,Advanced Database – B561. Spring
In this report, we presented an introduction to 2005.
distributed database system through a study that targeted two [16] M. Tamer Özsu, Patrick Valduriez, “Distributed Database Systems: Where
Are We Now?” Appeared in IEEE Computer, Vol. 24, No. 8, August 1991.
main parts: in the first part we presented an exploration of
distributed database environment and types of fragmentation. In
the second part, we explore the horizontal fragmentation AUTHORS
technique of a relation according to locality of precedence of its
attributes.
First Author – Ms. Priyanka R. Bhuyar, M.E.-I Year (CSE),
Making proper fragmentation of the relations and
Sipna college of Engineering & Technology, Amravati, India,
allocation of the fragments is a major research area in distributed
[email protected]
databases. Many techniques have been proposed by the

www.ijsrp.org
International Journal of Scientific and Research Publications, Volume 2, Issue 5, May 2012 7
ISSN 2250-3153
Second Author – Dr. A.D.Gawande, HOD (CMPS), Department Third Author – Prof.A.B.Deshmukh, Professor(CMPS),
of computer science & Engineering, Sipna College of Department of computer science & Engineering, Sipna College
Engineering & Technology, Amravati, India, of Engineering & Technology, Amravati, India.
[email protected]

www.ijsrp.org

View publication stats