Scheduling real-time parallel applications in SaaS clouds in the presence of transient software failures
A proper scheduling algorithm is critical for the performance of distributed systems. One of task scheduling algorithms for parallel and distributed systems is gang scheduling. In this paper we study gang scheduling performance in the presence of soft real-time periodic jobs which can tolerate deadline misses by bounded amounts. We employ a simulation model to address performance issues associated with gang scheduling in distributed systems for various workloads. Simulated results indicate that the performance of gang scheduling policies depends on gang workload, the period of the soft real-time jobs and also on the size of the bounded tardiness.