Evaluating performance attributes of layered software architecture

Abstract

The architecture of a software system is the highest level of abstraction whereupon useful analysis of system properties is possible. Hence, performance analysis at this level can be useful for assessing whether a proposed architecture can meet the desired performance specifications and can help in making key architectural decisions. In this paper we propose an approach for performance evaluation of software systems following the layered architecture, which is a common architectural style for building software systems. Our approach initially models the system as a Discrete Time Markov Chain, and extracts parameters for constructing a closed Product Form Queueing Network model that is solved using the SHARPE software package. Our approach predicts the throughput and the average response time of the system under varying workloads and also identifies bottlenecks in the system, suggesting possibilities for their removal. © Springer-Verlag Berlin Heidelberg 2005.