A Pressure-Aware Policy for Contention Minimization on Multicore Systems

Abstract

Modern Chip Multiprocessors (CMPs) are integrating an increasing amount of cores to address the continually growing demand for high-application performance. The cores of a CMP share several components of the memory hierarchy, such as Last-Level Cache (LLC) and main memory. This allows for considerable gains in multithreaded applications while also helping to maintain architectural simplicity. However, sharing resources can also result in performance bottleneck due to contention among concurrently executing applications. In this work, we formulate a fine-grained application characterization methodology that leverages Performance Monitoring Counters (PMCs) and Cache Monitoring Technology (CMT) in Intel processors. We utilize this characterization methodology to develop two contention-aware scheduling policies, one static and one dynamic , that co-schedule applications based on their resource-interference profiles. Our approach focuses on minimizing contention on both the main-memory bandwidth and the LLC by monitoring the pressure that each application inflicts on these resources. We achieve performance benefits for diverse workloads, outperforming Linux and three state-of-the-art contention-aware schedulers in terms of system throughput and fairness for both single and multithreaded workloads. Compared with Linux, our policy achieves up to 16% greater throughput for single-threaded and up to 40% greater throughput for multithreaded applications. Additionally, the policies increase fairness by up to 65% for single-threaded and up to 130% for multithreaded ones.

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

“…Kundan et. al [36] propose a scheduler tailored explicitly to oversubscription scenarios on UMA. These are the same scenarios addressed by other recent works [16], [37], which -unlike our proposal-leverage coscheduling on UMA platforms instead of NUMA-aware thread placement.…”

“…Determining these optimal solutions (described in detail later) requires extensive exploration of the search space. Unfortunately, the complexity of the optimal thread placement problem combined with the NP-hard nature of the optimal cachepartitioning/clustering problems [7], [9], [27], [36] requires the exploration of a vast search space, whose size grows exponentially with the number of applications and LLC ways. This constitutes an important challenge.…”

Interference Analysis and Resource Management in Server Processors: from HPC to Cloud Computing

“…Kundan et al [58] characterize the pressure that applications inflict on both LLC and memory bandwidth to schedule applications to the processor cores, prioritizing those applications that made the least progress. However, offline information (IPC of the application alone) is required to estimate the progress.…”