DISTRIBUTED POSITIONING OF REPLICA USING GRASS GROWING STRUCTURE

Abstract

Data Replication is a common technique used to enhance the performance of data access in distributed systems. There are many efficient replica schemes involved in replication technique such as replication strategy, replica selection strategy to find the best fit replica, replication consistency, and replica positioning mechanisms This thesis focuses on a distributed data replication technique based on Grass Growing Structure to reduce bandwidth consumption and access latency. The replication will be multicast to the nearest nodes within predefined limiting distance using Depth Limit Search algorithm. The proposed algorithm is then compared with centralized, flooding, multi-master and random walk algorithms. Evaluation is carried out by a simulation using OptorSim to run on standard benchmarking testbeds, namely, EU Data Grid Testbed, CMS Testbed and GridPP Testbed. Performance measurement is done based on a number of network metrics such as mean job time, effective network usage, local file access, and Band width usage. Result statistics show that the Grass Growing Structure performs better than all comparative algorithms which will entail energy conservation in network data distribution.