LFOC+: A Fair OS-level Cache-Clustering Policy for Commodity Multicore Systems

Abstract

Commodity multicore systems are increasingly adopting hardware support that enables the system software to partition the last-level cache (LLC). This support makes it possible for the operating system (OS) or the Virtual Machine Monitor (VMM) to mitigate shared-resource contention effects on multicores by assigning different co-running applications to various cache partitions. Recently cache-clustering (or partition-sharing) strategies have emerged as a way to improve system throughput and fairness on new platforms with cache-partitioning support. As opposed to strict cache-partitioning, which allocates separate cache partitions to each application, cache-clustering allows partitions to be shared by a group of applications.In this article we propose LFOC+, a fairness-aware OS-level cache-clustering policy for commodity multicore systems. LFOC+ tries to mimic the behavior of the optimal cache-clustering solution for fairness, which we could obtain for different workload scenarios by using a simulation tool. Our dynamic cache-clustering strategy continuously gathers data from performance monitoring counters to classify applications at runtime based on the degree of cache sensitivity and contentiousness, and effectively separates cache-sensitive applications from aggressor programs to improve fairness, while providing acceptable system throughput. We implemented LFOC+ in the Linux kernel and evaluated it on a real system featuring an Intel Skylake processor, where we compare its effectiveness to that of four previously proposed cache-clustering policies. Our experimental analysis reveals that LFOC+ constitutes a lightweight OS-level policy and improves fairness relative to two other state-of-the-art fairness-aware strategies -Dunn and LFOC-, by up to 22% and up to 20.6%, respectively, and by 9% and 4.9% on average.

Flexible system software scheduling for asymmetric multicore systems with PMCSched: A case for Intel Alder Lake

“…PMCSched is already publicly available at PMCTrack's source code repository; 48 technical documentation on our framework can be found on its official website 54 . As for future work, we plan on exploring the potential of PMCSched for the design and implementation of resource‐management policies by leveraging its API for system‐software guided cache‐partitioning 71 on both symmetric and asymmetric multicore platforms. We also envision leveraging a similar approach to the OS‐runtime interaction for other types of multithreaded programs, as well as to augment flexible AID's functionality with support for malleability.…”

Divide&Content: A Fair OS-Level Resource Manager for Contention Balancing on NUMA Multicores

“…Specifically, cores in a memory node typically share critical resources with the neighboring cores, such as a last-level cache (LLC) and the local DRAM controller. Corunning applications and virtual machines (VMs) may intensively compete with each other for such shared resources, leading to uneven and hard-to-predict application/VM's performance degradation [1], [7]- [9]. The higher latency of remote memory accesses, and the interconnect contention in NUMA, are known to aggravate this problem further [10].…”

“…wide fairness [7], [11], [12] and QoS (Quality of Service) constraints [1], [2], [13]. For instance, an application's completion time and tail latency may largely depend on the application's co-runners [9], [13].…”

“…relative performance degradation w.r.t. an isolated execution) of equal-priority applications that run simultaneously on the system may differ substantially under contention, leading to unfairness and other issues [2], [7], [8], [11]. Contention-induced performance disparities are exacerbated in NUMA multicores -and more generally in systems with multiple LLCs (each one shared by different groups of cores)-, where an application's slowdown may even differ substantially across multiple runs of the same multiprogram workload, depending on the specific thread-tocore mappings.…”

“…For these techniques to be effective, the system software has to be cognizant of the underlying hardware and the memory -and LLC-related behavior of the various applications, which may vary dynamically with program phases [9]. Another popular control mechanism is to partition the LLC, enabling to impose a certain degree of isolation among applications/VMs [1], [7]. After more than two decades of research on cache-partitioning [19], [20], the fairly recent adoption of hardware partitioning extensions in commodity processors [21]- [23] has given rise to a growing interest in the design of LLC-partitioning policies [7], [13], [21], [24].…”