Policy Driven Negotiation to Improve the QoS in Data Grid

Policy Driven Negotiation to Improve the QoS in Data Grid

Ghalem Belalem (University of Oran (Es Senia), Algeria)
DOI: 10.4018/978-1-61520-611-7.ch105
OnDemand PDF Download:
$37.50

Abstract

Data grids have become an interesting and popular domain in grid community (Foster and Kesselmann, 2004). Generally, the grids are proposed as solutions for large scale systems, where data replication is a well-known technique used to reduce access latency and bandwidth, and increase availability. In splitting of the advantages of replication, there are many problems that should be solved such as, • The replica placement that determines the optimal locations of replicated data in order to reduce the storage cost and data access (Xu et al, 2002); • The problem of determining which replica will be accessed to in terms of consistency when we need to execute a read or write operation (Ranganathan and Foster, 2001); • The problem of degree of replication which consists in finding a minimal number of replicas without reducing the performance of user applications; • The problem of replica consistency that concerns the consistency of a set of replicated data. This consistency provides a completely coherent view of all the replicas for a user (Gray et al 1996). Our principal aim, in this article, is to integrate into consistency management service, an approach based on an economic model for resolving conflicts detected in the data grid.
Chapter Preview
Top

Introduction

Data grids have become an interesting and popular domain in grid community (Foster and Kesselmann, 2004). Generally, the grids are proposed as solutions for large scale systems, where data replication is a well-known technique used to reduce access latency and bandwidth, and increase availability. In splitting of the advantages of replication, there are many problems that should be solved such as,

  • The replica placement that determines the optimal locations of replicated data in order to reduce the storage cost and data access (Xu et al, 2002);

  • The problem of determining which replica will be accessed to in terms of consistency when we need to execute a read or write operation (Ranganathan and Foster, 2001);

  • The problem of degree of replication which consists in finding a minimal number of replicas without reducing the performance of user applications;

  • The problem of replica consistency that concerns the consistency of a set of replicated data. This consistency provides a completely coherent view of all the replicas for a user (Gray et al 1996).

Our principal aim, in this article, is to integrate into consistency management service, an approach based on an economic model for resolving conflicts detected in the data grid.

The reminder of the article is organized as follows. The next section describes the fundamental principles of pessimistic and optimistic consistency approaches. Section 3, is devoted to the description of the model used in our consistency management service. In section 4, we describe our consistency management service and its algorithms for replicas in data grid. It is based on the economic model (Buyya and Vazhkudai, 2001). An evaluation and comparison of our proposition with other approaches are presented in Section 5. Section 6 briefly presents related pioneering works for resolution of the conflicts among the divergent replicas. Finally, Section 7 concludes this work by some future tracks.

Top

Approaches To Consistency Management

Data replica is made up of multiple copies, on separate computers. It is a significant technology which improves availability and execution. But the various replica of the same object should be coherent. There are many consistency models, which neither offer the same performances nor impose the same constraints to the application programmers.

The replica consistency management can be done either in a synchronous way by using what is known as pessimistic algorithms, or in an asynchronous way by employing what is designated as optimistic algorithms.

Key Terms in this Chapter

Multi-Master Strategy: In this strategy, a system supporting several masters per object.

Quorum: In general allow writes to be recorded only at a subset (a write quorum) of the up nodes, so long as reads are made to query a subset (a read quorum) that is guaranteed to overlap the write quorum.

Data Grid: A data grid is a grid computing system that deals with data — the controlled sharing and management of large amounts of distributed data.

Economic model: An approach can be employed for managing resources in data grid for services that end-user consumes. Pricing based on the demand of users and the supply of resources is the main driver in the competitive.

Single Master Strategy: In this strategy, a system supporting one master per object.

Dutch Auction: A Dutch auction referred specifically to a type of auction that starts with a high price that keeps going down until the item sells. This is the opposite process to regular auctions, where an item starts at a minimum price and bidders wrestle over it by increasing their offers.

English Auction: Bidding starts with a low price, and is raised incrementally as progressively higher bids are solicited, until either the auction is closed or no higher bids are received.

Complete Chapter List

Search this Book:
Reset