Virtualizing performance asymmetric multi-core systems

Abstract

Performance-asymmetric multi-cores consist of heterogeneous cores, which support the same ISA, but have different computing capabilities. To maximize the throughput of asymmetric multi-core systems, operating systems are responsible for scheduling threads to different types of cores. However, system virtualization poses a challenge for such asymmetric multi-cores, since virtualization hides the physical heterogeneity from guest operating systems. In this paper, we explore the design space of hypervisor schedulers for asymmetric multi-cores, which do not require asymmetry-awareness from guest operating systems. The proposed scheduler characterizes the efficiency of each virtual core, and map the virtual core to the most area-efficient physical core. In addition to the overall system throughput, we consider two important aspects of virtualizing asymmetric multi-cores: performance fairness among virtual machines and performance scalability for changing availability of fast and slow cores.We have implemented an asymmetry-aware scheduler in the open-source Xen hypervisor. Using applications with various characteristics, we evaluate how effectively the proposed scheduler can improve system throughput without asymmetry-aware operating systems. The modified scheduler improves the performance of the Xen credit scheduler by as much as 40% on a 12-core system with four fast and eight slow cores. The results show that even the VMs scheduled to slow cores have relatively low performance degradations, and the scheduler provides scalable performance with increasing fast core counts.

How Small Can it Be?: The Design of a Cost-Effective Side-core for I/O Virtualization

“…Running workloads that have frequent memory accesses on a small core is more energy efficient [10], [12], [13], [17]. Thus, on an AMP system, scheduling VM I/O tasks on small core(s) while keeping big cores occupied with complex applications provides better energy efficiency and performance.…”

“…There are many in-depth studies about efficient workload scheduling techniques on AMP systems [10]- [13], [17]. They developed performance models to characterize workloads, and proposed asymmetry aware core scheduling techniques according to the characteristics of a workload.…”

“…They developed performance models to characterize workloads, and proposed asymmetry aware core scheduling techniques according to the characteristics of a workload. Among them, Kazempour et al [11] and Kwon et al [17] studied the AMP aware scheduling in a virtualized environment. However, They did not study VM I/O, and explored the scheduling when workloads are running on shared cores.…”

“…Kazempour et al [11], Shelepov et al [12], Kwon and et al [17] used DVFS to emulate AMP systems. However, emulating an AMP system only with DVFS can be an over-simplification for some workloads, because frequency sensitive workloads always run slow on a core at a low frequency [13].…”

“…There are some studies about efficient asymmetry aware scheduling on AMPs [11], [12], [17]. They did not study VM I/O, and the workloads on their studies were independent of each other, while I/O intensive workloads are highly dependent to VM I/O tasks.…”