Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Energy-Efficient QoS-Aware Scheduling for S-NUCA Many-Cores

Sudam M. Wasala, Jurre Wolff, Yixian Shen, Anuj Pathania, Clemens Grelck, Andy D. Pimentel

TL;DR

This work tackles quality-of-service (QoS) maintenance in S-NUCA many-core processors where cache-access heterogeneity makes performance unpredictable. It introduces a reactive QoS management policy driven by application-level Heartbeat performance, using DVFS and thread migrations to keep Heart Rate (HR) within a predefined range and then minimize energy once QoS is achieved. The authors extend the HotSniper simulator with an integrated Heartbeat framework and a novel HR-based scheduler, demonstrating stable QoS and up to 18.7% energy savings versus state-of-the-art schedulers. This approach provides a practical pathway to QoS-aware, energy-efficient scheduling in highly non-uniform multicore architectures.

Abstract

Optimizing performance and energy efficiency in many-core processors, especially within Non-Uniform Cache Access (NUCA) architectures, remains a critical challenge. The performance heterogeneity inherent in S-NUCA systems complicates task scheduling due to varying cache access latencies across cores. This paper introduces a novel QoS management policy to maintain application execution within predefined Quality of Service (QoS) targets, measured using the Application Heartbeats framework. QoS metrics like Heartbeats ensure predictable application performance in dynamic computing environments. The proposed policy dynamically controls QoS by orchestrating task migrations within the S-NUCA many-core system and adjusting the clock frequency of cores. After satisfying the QoS objectives, the policy optimizes energy efficiency, reducing overall system energy consumption without compromising performance constraints. Our work leverages the state-of-the-art multi-/many-core simulator {\em HotSniper}. We have extended it with two key components: an integrated heartbeat framework for precise, application-specific performance monitoring, and our QoS management policy that maintains application QoS requirements while minimizing the system's energy consumption. Experimental evaluations demonstrate that our approach effectively maintains desired QoS levels and achieves 18.7\% energy savings compared to state-of-the-art scheduling methods.

Energy-Efficient QoS-Aware Scheduling for S-NUCA Many-Cores

TL;DR

This work tackles quality-of-service (QoS) maintenance in S-NUCA many-core processors where cache-access heterogeneity makes performance unpredictable. It introduces a reactive QoS management policy driven by application-level Heartbeat performance, using DVFS and thread migrations to keep Heart Rate (HR) within a predefined range and then minimize energy once QoS is achieved. The authors extend the HotSniper simulator with an integrated Heartbeat framework and a novel HR-based scheduler, demonstrating stable QoS and up to 18.7% energy savings versus state-of-the-art schedulers. This approach provides a practical pathway to QoS-aware, energy-efficient scheduling in highly non-uniform multicore architectures.

Abstract

Optimizing performance and energy efficiency in many-core processors, especially within Non-Uniform Cache Access (NUCA) architectures, remains a critical challenge. The performance heterogeneity inherent in S-NUCA systems complicates task scheduling due to varying cache access latencies across cores. This paper introduces a novel QoS management policy to maintain application execution within predefined Quality of Service (QoS) targets, measured using the Application Heartbeats framework. QoS metrics like Heartbeats ensure predictable application performance in dynamic computing environments. The proposed policy dynamically controls QoS by orchestrating task migrations within the S-NUCA many-core system and adjusting the clock frequency of cores. After satisfying the QoS objectives, the policy optimizes energy efficiency, reducing overall system energy consumption without compromising performance constraints. Our work leverages the state-of-the-art multi-/many-core simulator {\em HotSniper}. We have extended it with two key components: an integrated heartbeat framework for precise, application-specific performance monitoring, and our QoS management policy that maintains application QoS requirements while minimizing the system's energy consumption. Experimental evaluations demonstrate that our approach effectively maintains desired QoS levels and achieves 18.7\% energy savings compared to state-of-the-art scheduling methods.

Paper Structure

This paper contains 10 sections, 2 equations, 6 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Heartbeat Simulation using HotSniper
  3. Static Non-Unifrom Cache Access (S-NUCA) Architecture
  4. Heartbeat Integration in HotSniper
  5. Related Work
  6. QoS Management Algorithm
  7. Evaluation
  8. Evaluation of Performance
  9. Evaluation of Energy Consumption
  10. Conclusion

Figures (6)

  • Figure 1: HotSniper tool-flow with heartbeat extensions in blue.
  • Figure 2: Control flow of the Qos management policy
  • Figure 3: States of the System characterized based on the current HR with respect to the target range.
  • Figure 4: Performance of our policy compared to HPM and PCMiG. (top left: blackscholes (3 threads), top right: blackscholes (15 threads) middle left: canneal (3 threads), middle right: canneal (16 threads), bottom left: dedup (2 threads), bottom right: dedup (10 threads))
  • Figure 5: Four parallel blackscholes (12 threads) applications with different target ranges
  • ...and 1 more figures