Performance evaluation of scheduling policies in symmetric multiprocessing environments

Abstract

Abstract-The shift of hardware architecture towards parallel execution led to a broad usage of multi-core processors in desktop systems and in server systems. The benefit of additional processor cores for software performance depends on the software's parallelism as well as the operating system scheduler's capabilities. Especially, the load on the available processors (or cores) strongly influences response times and throughput of software applications. Hence, a sophisticated understanding of the mutual influence of software behaviour and operating system schedulers is essential for accurate performance evaluations. Multi-core systems pose new challenges for performance analysis and developers of operating systems. For example, an optimal scheduling policy for multi-server systems, such as shortest remaining processing time (SRPT) for single-server systems, is not yet known in queueing theory. In this paper, we present a detailed experimental evaluation of general purpose operating system (GPOS) schedulers in symmetric multiprocessing (SMP) environments. In particular, we are interested in the influence of multiprocessor load balancing on software performance. Additionally, the evaluation includes effects of GPOS schedulers that can also occur in single-processor environments, such as I/Oboundedness of tasks and different prioritisation strategies. The results presented in this paper provide the basis for the future development of more accurate performance models of today's software systems.

Towards software performance engineering for multicore and manycore systems

“…For general purpose operating system schedulers, the effects of load balancing and process priorities on process execution were investigated in [21] and later modeled in [22]. In addition to allocating execution time, operating system scheduling also has a major impact on how processes interact through shared resources at both hardware and software levels.…”

Joint Tracking of Performance Model Parameters and System Behavior Using a Multiple-Model Kalman Filter

“…Two widely used service disciplines, PS and FCFS, are considered. The choice of discipline within a performance model has been shown to depend on the relative size of request service time and the OS scheduling time-slice [8]. As the request service time changes, the close-to-truth service discipline may also change.…”

“…A mode is the true system behavior that is to be approximated by the models, including measurement model, process model, and error covariances. For example, the service discipline of CPU can be approximated by process sharing (PS) discipline when the request service time is significantly larger than the time-slice of the OS scheduler, but it more closely resembles a first-come first-served (FCFS) discipline, when the service time is smaller than the time-slice [8]. The service time of the system may vary smoothly during normal time, but it undergoes rapid changes during bursty time.…”