MINIMIZATION OF NETWORK TRAFFIC IN DISTRIBUTED FILE SYSTEMS: A CLUSTER BASED APPROACH

Abstract

As the availability of cheap and faster computer networks has grown, the demand for distributed systems is gaining momentum as a medium for information sharing. In distributed systems file system, is the most important design issue. Distributed file system offers reliability and availability through file replication. Communication overhead caused due to accessing of replicated files has a profound impact on the performance of the distributed system. Reducing this overhead shows a significant improvement in the performance of the system. In this dissertation an attempt has been made to minimize r.. the communication overhead so as to enhance the performance of the distributed system. In order to achieve this objective an algorithm "Optimal Program Execution Location Algorithm" has been developed and tested. To optimize the communication costs the proposed algorithms considers remote execution cost, costs for read and write accesses of data files and I/O costs, for each Initiated program. Based on the knowledge of file replication and program file characteristics, it calculates the best location for the program execution. A cluster based appropch has been used to reduce the complexity of the algorithm. A discrete event simulation has been carried out to examine the implications of the algorithm. The Inter cluster traffic caused by program execution in the optimal cluster has always been less than that when the program executes in the initiating cluster. The simulation study was made on TATA Elxsi running IRIX. The code required for the simulation was written in C++.