Bandwidth Management For Distributed Systems

Abstract

The method of finding point of bending on efficiency curve in a distributed systems has been presented. The performance of distributed environment was studied as a cost function of mean network throughput and memory buffers availability. The method can be used then to build up an autonomic bandwidth management. The bounding point on efficiency curve was found as well analytically as for testing facility of a distributed laboratory environment. The influence of data replication and scheduling strategy on final results has been included. The computer simulation has been supported by analytical calculation. Two regions of different behavior on universal hyperbolic like dependence of the performance versus mean network throughput was shown up. A method of finding bounding point has been discuss in order to optimize efficiency of the considered system by such parameters like memory buffers and network bandwidth.