Sei sulla pagina 1di 27

Consistency Management Strategies for Data Replication in Mobile Ad Hoc Networks

By. P. Victer Paul Dear, We planned to share our eBooks and project/seminar contents for free to all needed friends like u.. To get to know about more free computerscience ebooks and technology advancements in computer science. Please visit.... http://free-computerscience-ebooks.blogspot.com/ http://recent-computer-technology.blogspot.com/ http://computertechnologiesebooks.blogspot.com/ Please to keep provide many eBooks and technology news for FREE. Encourage us by Clicking on the advertisement in these Blog.

ABSTRACT

In a mobile ad hoc network, data replication drastically improves data availability. Global consistency of data operations on replicas is not desirable by many applications New consistency maintenance based on local conditions such as location and time need to be investigated. This paper attempts to classify different consistency levels according to requirements from applications and provides protocols to realize them

INTRODUCTION

In mobile ad hoc network (MANET), as mobile hosts move freely, disconnections often occur and two separated networks to become inaccessible to each other. To improve data availability, data replication is the most promising solution Replicas of a data item become invalid after the host holding the original updates it. Here we discuss different consistency conditions of data operations on replicas in MANETs.

SYSTEM MODEL

We assume an environment where each mobile host accesses data items held by other mobile hosts in a MANET. Details of the system model are given as follows:
Each mobile host (peer) knows its current location by using some device such as GPS The MANET consists of two kinds of mobile hosts; proxies and peers. Each proxy and peer knows all proxies in the entire network. The set of all regions in the entire network is denoted by R ={R1,R2, . . .,Rl}, where l is the total number of regions Data are handled as a collection of data items. We assign a unique data identifier to each data item located in the system.

CLASSIFICATION OF CONSISTENCY LEVELS

We propose four consistency levels

different

primitive

Global Consistency (GC) Local Consistency (LC) Time-Based Consistency (TC) Peer-Based Consistency (PC) Application-Based Consistency (AC)

Global Consistency

The consistency of data operations on replicas is required in the entire network GC requires that every read operation issued by any peer necessarily reads a replica of the latest version in the entire network Many applications do not desire it.

Local Consistency

The consistency of data operations on replicas is required only in each region of interest. Weakens the strictness of consistency Every read operation reads a replica of the latest version in the region.

Time-Based Consistency

Replicas are valid, if it have not passed a predetermined time (T). A read operation on the replica succeeds when, Vj + T > Aj

Vj be the time at latest write operation on the replica was performed. Aj be time at which read operation is issued.

We assume peers use some logical clocks and time is synchronized

Peer-Based Consistency

The consistency of data operations is required only in each peer. Weakens the strictness of consistency than LC. Every read operation issued by the peer necessarily reads its own replica to which the latest write operation by itself.

Application-Based Consistency

Since different applications running in a MANET might require different consistency levels, hybrid consistency levels can be used.

CONSISTENCY MANAGEMENT PROTOCOLS

We describe protocols to realize the GC consistency level. the system has to guarantee peers to read the latest version in the entire network. To realize GC, the simplest way is read-one write-all,

Quorum system

In a quorum system, read and write operations are performed on only replicas held by mobile hosts that form read and write quorums. There is an intersection between write and read quorums. At least one mobile host in a read quorum holds a replica of the latest version.

Overview

The consistency of data operations on replicas is hierarchically managed at two levels;


among peers in each region (local quorum) and among proxies (global quorum).

We adopt a very simple strategy that calculates the sizes of quorums, |QW|+|QR| >l.
QW is Quorum size of write operation to any data item. QR is Quorum size of read operation to any data item. l is Total number of regions in the entire network.

In each region Ri (i=1,. . .,l ), the quorum size, |QLWij| +|QLRij|>Pij.


QLWij is Quorum size of write operation on data item Dj. QLRij is Quorum size of Read operation on data item Dj. Pij is Total number of peers that holds Dj.

Protocol

We present the protocol to achieve GC. We adopt a three-phase approach for a write (read) operation, which basically consists of three rounds of message exchanges.
1) a lock request/its reply, 2) write operations/its acknowledgment (a request for the latest replica/data transfer), and 3) a message for commit and lock release/its acknowledgment (acknowledgment and lock release).

we describe the details of each of the three phases

Phase 1

When a read (write) operation is issued by a peer in region Ri to data item Dj,
first, the peer unicasts a request to the proxy of the region. peer does not connect to the proxy, the peer tries to find another proxy in a different region The coordinator tries to set global read (write) locks to arbitrary |QR| (|QW|) replicas held by proxies. Each proxy in region Ri that received the request tries to set local read (write) locks to arbitrary |QLRij| (|QLWij|) replicas of data item Dj.

Phase 2: read operation.

In phase 2 of Read Operation,


the requested data operation is performed on replicas with locks. the coordinator unicasts a data transmission request to the closest proxy having the latest version in its region. The peer that received the request transfers its holding replica to the proxy. The proxy forwards the replica to the coordinator. The coordinator that received the latest replica forwards the replica to the request-issuing peer.

Phase 3: read operation

In this phase,
peer that received a latest replica sends an acknowledgment to the coordinator the coordinator sends a commit and lock release message to the proxies involved in the global quorum. Each proxy that received this message forwards it to the peers involved in the local quorum in the region. Then, all the locks set to replicas for this operation are released and the procedure finishes.

Phase 2: write operation

In phase 2 of the write operation,


the coordinator receives the replica of the new version from the request-issuing peer and forwards it to |QW| proxies. Each proxy that received the replica forwards it to the closest |QLWij| peers having replicas with the local lock. If the proxy receives the acknowledgment from all the |QLWij| peers, it sends acknowledgment to the coordinator.

Phase 3: write operation

In phase 3 of the write operation,


the coordinator sends a commit and lock release message to the proxies involved in the global quorum. Each proxy that received this message forwards it to the peers having replicas with the local lock and the procedure finishes.

To ensure the execution of write operations on replicas held by multiple nodes in a distributed system, two phase commit (2PC) is the most popular approach. Achieves by two rounds of message transmissions between the coordinator and all the replica holders;
commit prepare/acknowledgment and commit/acknowledgment.

Future Work

In our work, we assume a simple transaction model in which every transaction will consist of a data operation (read or write) but in a real environment, transaction consists of multiple operations. So, We also plan to extend our protocols to deal with more complex transactions. We plan to consider replication strategies suitable for each consistency level.

Conclusion

Since in MANETs peers disconnection causes frequent network partitioning, it is difficult and, in some cases, not desirable to provide traditional strong consistency of data operations We have classified consistency levels according to applications demand and, then, have designed protocols to achieve GC.

QUERIES

Thank You