Analysis of Real-Time Applications Feasibility through Simulation

Abstract

An approach to estimate feasibility of a real-time multi-task application with various combinations of the scheduling mode and the protocol of access to shared informational resources when run on a multi-core platform is described. The application structure is specified through a simple formalized profile consisting of segments of three types and specifying access to informational resources shared among application tasks, the amount of the required computing resource being estimated for each segment. The approach is based on the notion of application density introduced by the authors which characterizes the use of computational resource by this application and is derived from estimation of the application feasibility for various values of processor performance and the number of its cores in case of a multi-core platform. The overall structure of a simulation tool for estimation of the task response time (and therefore, application feasibility) is described, which provides more exact data compared to the known analytical methods where they are applicable. Two dissimilar implementations of this tool were developed and run on a number of benchmarks, including Liu-Layland configurations specified in the described formalism for application structure; the results in form of charts are presented along with their analysis and interpretation. The suggested approach allows to indentify an optimal combination of the scheduling mode and access protocol for the given multi-task application structure.