HAADAS : An enhanced approach for Re-
allocation of Fragments in Peer-to-Peer
Distributed Databases
Dr. D.I. George Amalarethinam
Director-MCA, Associate Professor
Department of Computer Science
Jamal Mohamed College (Autonomous)
Tiruchirappalli, India
di_george@ymail.edu
Dr. C. Balakrishnan
Assistant Professor
Alagappa Institute of Skill Development
Alagappa University
Karaikudi, India
balasjc@gmail.com
Abstract- The process of allocating fragments into the appropriate
sites is a NP-Complete problem in nature. It is a crucial step in
the design of Pee-to-Peer Distributed Databases. This paper
proposes a methodology for fragment re-allocation by using
Assignment algorithm of optimization principle, namely,
Hungarian Assignment Algorithm based Data Allocation Strategy
(HAADAS). The objective of the work is to reduce the number of
un-successful queries that are forwarded to other sites of the
network. The effective allocation of fragments directly influences
the efficiency in executing the user queries. The simulation
results confirms that the re-allocation of fragments in suitable
sites is influencing the communication cost while transaction.
Keywords: ElasticPeerDB, Peer-to-Peer databases, fragmentation,
allocation, optimization technique
I. INTRODUCTION
The characteristics of Distributed data processing
provide an efficient way of improving the performance,
reliability, availability and modularity. Adhering with this idea,
the Distributed Database System (DDBS) is a collection of
multiple, logically interrelated databases distributed over a
computer network [1]. There will be a possibility of improved
response times to queries and upgrading system capacity or
performance incrementally.
The conceptually simplest distribution scheme is to
distribute at the table level: any given table is stored in its
entirety at some site or it may be partitioned as logical units
and stored in different sites, this technique is called as
Fragmentation and Allocation. These fragments contain
sufficient information to allow reconstruction of the original
relation.
Fragmentation is desirable for performance reasons.
There are horizontal, vertical and mixed or hybrid
fragmentations. The horizontal fragment is a subset of the
tuples with all attributes in that relation. Horizontal
fragmentation splits the relation „horizontally‟ by assigning
each tuple or group (subset) of tuples of a relation to one or
more fragments, where each tuple or subset has a certain
logical meaning. Vertical fragmentation splits the relation by
decomposing „vertically‟ by columns (attributes). It requires to
include the primary key (candidate key) attributes in every
fragment so that the full relation can be reconstructed from the
fragments. In vertical fragmentation, system-provided „tuple-
ID‟ (or TID) is used as the primary key attribute along with
stored relation as address for linking tuples. Horizontal (or
vertical) fragmentation of a relation, followed by further
vertical (or horizontal) fragmentation of the same fragments, is
called mixed fragmentation. Allocation refers the process of
allocating fragment or each copy of a fragment into a particular
site of the distributed system. Fragment placement strategies
may be centralized, partitioned or fragmented and replicated.
The replication of fragments helps to improve availability,
performance of retrieval of global queries and reliability.
Peer-to-Peer (P2P) Database has distinction
characteristic to the client-server model. Each node of the
network would act as both client and as well as server, towards
requesting data from nearby nodes, routing, transporting and
serving data to other nodes. The characteristics of P2P
technology makes suitable for storing multiple copies of data
in several nodes, to offer reliable access of data and to
distribute the load of requests. These features of P2P
technology promise a network which is dynamic, reliable and
scalable.
The Peer-to-Peer concept having the potential to
handle the data in an effectual manner. This paper explains a
cluster based architecture for Distributed Databases, namely,
ElasticPeerDB. This addresses the fragmentation and
allocation phases in database design. ElasticPeerDB considers
the works done on the predicate based fragmentation and
introduced the clustering approach to draft the database model.
This paper proposes an optimized strategy for allocating the
data among the sites.
The paper is organized as follows. The section II
presents literature reviews on fragmentation, allocation,
clustering. Section III describes the ElasticPeerDB architecture
and allocation strategy. In Section IV an improved allocation
methodology is presented. The Section V describes the
simulation results of the new methodology. Finally Section VI
give conclusion and future research guidelines.
II. LITERATURE REVIEW
Bellatreche et al., [2] made a study on horizontal
fragmentation in the object-oriented model. Huang and Chen
International Journal of Applied Engineering Research, ISSN 0973-4562 Vol. 10 No.82 (2015)
© Research India Publications; httpwww.ripublication.comijaer.htm
315