A Hardware Approach to Fairly Balance the Inter-Thread Interference in Shared Caches

Abstract

Shared caches have become the common design choice in the vast majority of modern multi-core and many-core processors, since cache sharing improves throughput for a given silicon area. Sharing the cache, however, has a downside: the requests from multiple applications compete among them for cache resources, so the execution time of each application increases over isolated execution. The degree in which the performance of each application is affected by the interference becomes unpredictable yielding the system to unfairness situations. This paper proposes Fair-Progress Cache Partitioning (FPCP), a low-overhead hardware-based cache partitioning approach that addresses system fairness. FPCP reduces the interference by allocating to each application a cache partition and adjusting the partition sizes at runtime. To adjust partitions, our approach estimates during multicore execution the time each application would have taken in isolation, which is challenging. The proposed approach has two main differences over existing approaches. First, FPCP distributes cache ways incrementally, which makes the proposal less prone to estimation errors. Second, the proposed algorithm is much less costly than the state-of-the-art ASM-Cache approach. Experimental results show that, compared to ASM-Cache, FPCP reduces unfairness by 48% in four-application workloads and by 28% in eight-application workloads, without harming the performance.

A Survey of Techniques for Reducing Interference in Real-Time Applications on Multicore Platforms

“…Selfa et al [95] introduced a hardware-based cache partitioning approach that reduces shared cache interference by assigning a private cache partition to each application. The size of each cache partition is dynamically adjusted at runtime according to the requirements of each application during its execution.…”

A State of Art Survey for OS Performance Improvement

“…If cache will be shared the downside will appear which the competition among the requests from various applications for cache resources, therefore over isolated execution, the execution time will be increased of each application. Fair-Progress Cache Partitioning (FPCP) has been suggested by (Vicent Selfa, Sahuquillo, Petit, & Gómez, 2017), this approach is a low-overhead cache partitioning hardware-based which can identify system fairness. By allocating; for all applications; a cache partition, the interference can be decreased with FPCP and modify the partition sizes at runtime.…”

“…Table 1 provides an overview of the different systems explained in section II for measuring and even enhancing the operating system performance. As illustrated in the table there is a number of novel approaches been introduced as in (Li et al, 2017), (Saez et al, 2017), (Seo et al, 2018), (V Selfa, Sahuquillo, Petit, & Gómez, 2017) and (Lu et al, 2017) that cover many metrics. The most important approach is Argobots that proposed by (Seo et al, 2018) and includes several features; low-level threading, low-level tasking, and a lightweight framework.…”

Adaptive Prefetching and Cache Partitioning for Multicore Processors

“…The idea behind cache partitioning is to retain the benefits of a monolithic and shared cache and, at the same time, avoiding the disadvantages of cache interference. By mitigating cache interference, a well-crafted cache partitioning policy can improve performance [87], fairness [78], and isolate applications for security [96] and quality of service (QoS) reasons [8]. Cache partitioning allows system software to divide the cache space between cores, threads or applications.…”

“…Cache partitioning allows system software to divide the cache space between cores, threads or applications. There are different ways to partition the cache, doing set-partitioning from software with page coloring [101] way partition-ing from hardware [78,70] or using probabilistic cache replacement mechanisms [58].…”