Abstract
While multicore architectures are used in the whole product range from server systems to handheld computers, the deployed software still undergoes the slow transition from sequential to parallel. This transition, however, is gaining more and more momentum due to the increased availability of more sophisticated parallel programming environments. Combined with the ever increasing complexity of multicore architectures, this results in a scheduling problem that is different from what it has been, because concurrently executing parallel programs and features such as non-uniform memory access, shared caches, or simultaneous multithreading have to be considered. In this paper, we compare different ways of scheduling multiple parallel applications on multicore architectures. Due to emerging parallel programming environments, we primarily consider applications where the parallelism degree can be changed on the fly. We propose TACO, a topology-aware scheduling scheme that combines equipartitioning and coscheduling, which does not suffer from the drawbacks of the individual concepts. Additionally, TACO is conceptually compatible with contention-aware scheduling strategies. We find that topology-awareness increases performance for all evaluated workloads. The combination with coscheduling is more sensitive towards the executed workloads and NUMA effects. However, the gained versatility allows new use cases to be explored, which were not possible before.
Highlights
Since the availability of multicore systems to the mass market, the development of parallel applications has changed
We address the issue of resource contention on multicore architectures and outline a method to combine existing contention-aware scheduling strategies with TACO, essentially making them fit for parallel programs – or our approach contentionaware, depending on the point of view
In current operating systems, scheduling of applications is done from within the operating system scheduler that keeps all details, like system load or resource status, hidden from user-level applications, following the concept of separation of concerns. In these days with emerging many-core systems and an increasing count of parallel applications, new challenges arise when it comes to scheduling targeting high and efficient CPU utilization in non-HPC environments, which might require a change in this policy
Summary
Since the availability of multicore systems to the mass market, the development of parallel applications has changed. Schönherr et al / TACO: A scheduling scheme for parallel applications on multicore architectures plication level instead of thread level Both approaches assign computational resources at application level – a fact that applications can take advantage of at design time: low latency communication is possible, busy waiting and static load balancing can be used. In the absence of other objectives, such as job importance or job scalability, it is normally desired to distribute the available CPU time between multiple parallel programs in a fair manner, giving each program an equal share In the past, this has led to equipartitioning [16]: the available computational resources are split in space evenly between the running jobs.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call