A framework for simulation of concurrency control policy in real-time systems

Abstract

Control of concurrency is a critical aspect of both performance and correctness of real-time systems. Use of formal policies and mechanisms for concurrency control, in the style of transaction processing systems, would make it easier to reason about the logical correctness of concurrently executing processes in real-time systems. Adaptation of transaction processing techniques to real-time environments is a goal of our research. In this paper we describe a simulator used to study the interactions among scheduling policy, overload management, and concurrency control policy and the effects on the performance of real-time applications. The structure and capabilities of the simulator are described and its use in the evaluation of some concurrency control options for a real-time application is demonstrated. Through this example, we show how the simulation is customized for different processing models, scheduling options, and concurrency control policies.