Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

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

Juan Carlos Saez, Fernando Castro, Graziano Fanizzi, Manuel Prieto-Matias

TL;DR

LFOC+ tackles shared LLC contention in multicore systems by introducing a fairness-aware OS-level cache-clustering policy that dynamically classifies applications via performance-monitoring counters and couples this with a pairing-based clustering strategy to allow up to two cache-sensitive apps per partition. Implemented as a Linux kernel module, LFOC+ extends the authors' prior LFOC approach and demonstrates substantial fairness improvements over state-of-the-art policies while maintaining comparable throughput on Skylake hardware, approaching the optimal fairness solution in many scenarios. The approach relies on offline simulation (PBBCache) to derive insights and online PMC-driven adaptations (sampling and fairness phases) to react to workload changes, with special provisions for data-parallel multithreaded programs. The work also provides a practical path toward broader adoption via open-source tooling (PMCTrack) and discusses limitations and future directions, including deeper intra-application partitioning and support for AMD architectures with multiple LLCs.

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.

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

TL;DR

LFOC+ tackles shared LLC contention in multicore systems by introducing a fairness-aware OS-level cache-clustering policy that dynamically classifies applications via performance-monitoring counters and couples this with a pairing-based clustering strategy to allow up to two cache-sensitive apps per partition. Implemented as a Linux kernel module, LFOC+ extends the authors' prior LFOC approach and demonstrates substantial fairness improvements over state-of-the-art policies while maintaining comparable throughput on Skylake hardware, approaching the optimal fairness solution in many scenarios. The approach relies on offline simulation (PBBCache) to derive insights and online PMC-driven adaptations (sampling and fairness phases) to react to workload changes, with special provisions for data-parallel multithreaded programs. The work also provides a practical path toward broader adoption via open-source tooling (PMCTrack) and discusses limitations and future directions, including deeper intra-application partitioning and support for AMD architectures with multiple LLCs.

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.
Paper Structure (15 sections, 3 equations, 10 figures, 5 tables, 3 algorithms)

This paper contains 15 sections, 3 equations, 10 figures, 5 tables, 3 algorithms.

Table of Contents

  1. Introduction
  2. Related Work
  3. Cache-Clustering Algorithms
  4. Metrics
  5. Simulation Tool and Application Classes
  6. LFOC's Cache-Partitioning Algorithm
  7. Fairer LLC Distribution among Sensitive Programs
  8. The LFOC+ dynamic partitioning strategy
  9. Design and Implementation
  10. Dealing with Multithreaded Applications and Reducing Contention of Streaming Programs
  11. Experiments
  12. Comparison with Previous Approaches
  13. Overhead Analysis
  14. Workloads Including Multithreaded Applications
  15. Conclusions and future work

Figures (10)

  • Figure 1: Slowdown and LLCMPKC for different way counts
  • Figure 2: Cluster count and breakdown of applications into different categories for each cluster size
  • Figure 3: Distribution of LLC-space made by (a) UCP-Slowdown and (b) Best fairness-wise clustering with clusters of up to 2 applications
  • Figure 4: Sampling mode's LLC space distribution
  • Figure 5: Multiprogram workloads used for our experiments. The "x2" mark indicates that 2 instances of a benchmark are present in a workload
  • ...and 5 more figures