Handling Resource Failure Towards Load Balancing in Computational Grid Environment

摘要

A computational grid environment consists of several loosely coupled heterogeneous resources though a network. The resources may geographically reside in any part of the planet and its interaction with other components in the grid is independent of its location. The grid generally follows a Client-Broker-Resource System, in which the broker acts as an intermediary between the clients and the resources. The broker relays a client's request to the relevant available resource retrieves the response and relays the response back to the client. One of the major concerns in computational grid environment arises when there is a failure of any resource belonging to any grid site and that particular resource was already assigned a job by the resource broker then the situation becomes worse. The objective of this paper is to propose an approach to tackle this type of problematic situation in grid. This paper also emphasizes on a central control unit, known as resource broker, which is responsible for all the communications between the client/s and resource/s. This paper also proposes an approach for the optimum use of the available resources in the grid, so that the load across the grid is balanced. The major objective of this paper is to handle resource failure scenario in already proposed NDFS algorithm and to present the experimental results of enhanced NDFS algorithm implemented using UNICORE.