Hybrid performance modeling approach for network intensive distributed software

Abstract

When designing and evaluating software architectures and network facilities for hosting demanding distributed applications, taking performance considerations into account is essential. A key factor in assessing the performance of such a distributed system is the network latency and its relation to the application behaviour. In this respect, it is important to include the performance impact of the network into the performance models used during the entire design cycle of the system.A framework is proposed that allows to model both the software and the network components separately and extracts a single set of performance estimates for the entire system. This has the advantage of allowing the network and software aspects to be modeled separately using the modeling languages and tools most suited to those system aspects. A case study is presented to illustrate the use of the framework and its usefulness in predicting system performance.