Performance enhancement in quality of service for a data intensive grid

Abstract

The new generation scientific applications such as high energy newlinephysics molecular modelling and earth sciences involve the production of newlinelarge data sets from simulations or large scale experiments The analysis of newlinethese data sets and its dissemination among researchers located over a wide newlinegeographic area requires high capacity resources which include newlinesupercomputers high bandwidth networks and mass storage systems A grid newlinecomputing is prevalently used to support these computation intensive and newlinedata intensive applications which require distributed services Different types newlineof grids have been developed to support such applications and are classified newlineas computational grids and data grids The efficient use of grids requires newlineovercoming the challenges in security resource management scheduling and newlinedata management The major focus of this thesis is to present solutions to newlinechallenges especially in scheduling distributed resources newlineTraditionally scheduling algorithms objective is to manage the newlinetotal time required for computing all the jobs in a set and also address the load newlinebalancing factor The major challenge involved here is to distribute the load newlineamong the machines so that no machine is either idle or overburdened newlineScheduling algorithms with economic objectives try to maximize the newlineeconomic utilities for the user while minimizing the cost of executing a job newlineand at the same time maintain the quality of service QoS requirements newlinespecified by the users newline