Task response time for real-time distributed systems with resource contentions

Abstract

Response time is an important performance measure for real-time distributed systems. It is affected by such factors as interprocessor communications, precedence relationships, module assignments, and processor scheduling policies. Furthermore, resources (other than the processors) such as memory and data files are often shared among the modules-thus additional contention delays are incurred. To consider these factors, we propose an analytic model for estimating task response times in distributed systems with resource contentions. The model consists of two submodels. The first submodel is an extended queueing network model used for approximating module response times. This submodel is solved by a decomposition technique, which reduces the computation complexity by a two to three order of magnitude when compared with a direct approach. The second submodel is a weighted control-flow graph model from which task response time can be obtained by aggregating module response times in accordance with the precedence relationships. Simulation is used to validate the assumptions used in the model. Task response times estimated by the analytic model compare closely with simulation results. Our study reveals that resource contention delays depend on the availability or resources as well as the invocation rates and response times of those modules that use these resources. The model proposed in this paper can be used for studying the complex trade-offs among module assignments, scheduling policies, interprocessor communications, and resource contentions in distributed processing systems.