You are on page 1of 10

OBJECTIVE:

 Develop a replication control algorithm that


allows as many transactions as possible to
meet the deadlines and at the same time
maintaining the consistency of the replicated
data.
 Epsilon-serializability (ES) is employed as
correctness criterion, offering the possibility of
maintaining mutual consistency of replicated
data asynchronously.
Existing system:
 The existing system provides a high degree of
concurrency and faster response time but
violates data consistency.
 To synchronize concurrent data access of
transactions and to ensure identical replica
values, the transactions are blocked or aborted.
 Blocking may cause priority inversion.
 Aborting the lower priority transactions wastes
the work done.
Existing System:
 For database applications, the availability of
disk-resident data files is perhaps the major
concern.
 Most of the DB systems use a combination of
disk-based log with periodic check pointing of
memory resident data to insure the availability
of database.
 Due to system failures, the recovery time can
be long because of the data unavailability.
Existing System:
 To achieve a high degree of data availability,
two basic techniques are currently used.
 In the first, multiple copies of same data item
are stored.
 When one copy fails, the other copy can be
used.
 Unless both the copies fail simultaneously, the
data will be available.
Existing System:
 In the second approach, the data will be stored
in an array of disk drives along with the
redundant error detection/correction
information.
 When errors are discovered, the redundant
information can be used to restore the data.
 Here the application programs can continue
using that data with minimal interruption.
Proposed System (Abstract):
 The proposed algorithm employs ES
correctness criteria to control the inconsistency
of transactions and the mechanism for conflict
resolution of real-time transactions.
 If no consistency constraints are specified
explicitly by the user, then the ES correctness
criterion is employed to maintain the
consistency.
Proposed System (Abstract):
 Three strategies are proposed to cope up with
the situation when local database buffer has no
available space for a new object.
 The strategies used are
- strategy 1: no replacement
- strategy 2: LRU
- strategy 3: LFU
 The real-time scheduling part of the proposed scheme
has three components:
i) a policy to determine which transactions are
eligible for service,
ii) a policy for assigning priorities to transactions,
and
iii) a policy for resolving conflicts between two
transactions that wants to access the same data object.
 The proposed data allocation and replication algorithm
is made to fundamentally work together with any of
these strategies at the highest level.
MODULE DESCRIPTION

The phases involved in this project are:


 DISTRIBUTED SYSTEM
 LOCALIZED TRANSACTION AND
REPLICA OBJECTS
 TIME STAMP AND DATA
CONSISTENCY
 EPSILON SERIALIZABILITY
 CONFLICT RESOLUTION
References:
 IEEE transactions on parallel and distributed systems,
Vol.17 , NO.9, September 2006 “Practically
realizable Efficient Data allocation and Replication
strategies for distributed Databases with Buffer
Constraints” , Xin Gu, wujuan Lin and Bharardwaj
veeravalli.
 O.wolfson and S.Jajodia,”Distributed Algorithm for
Dynamic Replication of Data”, Proc.11th ACM symp,
Principles of Database Systems,pp.149-163, June
1992.

You might also like