Experimental Analysis of the Tardiness of Parallel Tasks in Soft Real-time Systems

International audienceA parallel application is defined as the application that can be executed on multiple processors simultaneously. In software, parallelism is a useful programming technique to take advantage of the hard-ware advancement in processors manufacturing nowadays. In real-time systems, where tasks have to respect certain timing constraints during execution, a single task has a shorter response time when executed in parallel than the sequential execution. However, the same cannot be trivially applied to a set of parallel tasks (taskset) sharing the same processing platform, and there is a negative intuition regarding parallelism in real-time systems. In this work, we are interested in analyzing this statement and providing an experimental analysis regarding the effect of parallelism soft on real-time systems. By performing an extensive simulation of the scheduling process of parallel taskset on multiprocessor systems using a known scheduling algorithm called the global Earliest-Deadline First (gEDF), we aim at providing an indication about the effects (positive or negative) of parallelism in real-time scheduling