Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Mixture-of-Schedulers: An Adaptive Scheduling Agent as a Learned Router for Expert Policies

Xinbo Wang, Shian Jia, Ziyang Huang, Jing Cao, Mingli Song

TL;DR

The paper tackles the limitation of static OS schedulers in heterogeneous and dynamic workloads by introducing the Adaptive Scheduling Agent (ASA), which dynamically routes workloads to a portfolio of expert schedulers. ASA combines a perception–decision–action loop with a three-stage offline preparation pipeline and a sched_ext-based runtime switch, enabling rapid adaptation to new hardware without retraining the core model. Its contributions include a time-weighted voting mechanism for stable workload recognition, a mapping table for scenario-specific policy selection, and a user-experience–oriented evaluation framework. Empirical results show ASA outperforms the default Linux scheduler in most scenarios, approaches oracle-level performance, and generalizes well to unseen hardware, suggesting practical impact for intelligent, adaptive OS scheduling.

Abstract

Modern operating system schedulers employ a single, static policy, which struggles to deliver optimal performance across the diverse and dynamic workloads of contemporary systems. This "one-policy-fits-all" approach leads to significant compromises in fairness, throughput, and latency, particularly with the rise of heterogeneous hardware and varied application architectures. This paper proposes a new paradigm: dynamically selecting the optimal policy from a portfolio of specialized schedulers rather than designing a single, monolithic one. We present the Adaptive Scheduling Agent (ASA), a lightweight framework that intelligently matches workloads to the most suitable "expert" scheduling policy at runtime. ASA's core is a novel, low-overhead offline/online approach. First, an offline process trains a universal, hardware-agnostic machine learning model to recognize abstract workload patterns from system behaviors. Second, at runtime, ASA continually processes the model's predictions using a time-weighted probability voting algorithm to identify the workload, then makes a scheduling decision by consulting a pre-configured, machine-specific mapping table to switch to the optimal scheduler via Linux's sched_ext framework. This decoupled architecture allows ASA to adapt to new hardware platforms rapidly without expensive retraining of the core recognition model. Our evaluation, based on a novel benchmark focused on user-experience metrics, demonstrates that ASA consistently outperforms the default Linux scheduler (EEVDF), achieving superior results in 86.4% of test scenarios. Furthermore, ASA's selections are near-optimal, ranking among the top three schedulers in 78.6% of all scenarios. This validates our approach as a practical path toward more intelligent, adaptive, and responsive operating system schedulers.

Mixture-of-Schedulers: An Adaptive Scheduling Agent as a Learned Router for Expert Policies

TL;DR

The paper tackles the limitation of static OS schedulers in heterogeneous and dynamic workloads by introducing the Adaptive Scheduling Agent (ASA), which dynamically routes workloads to a portfolio of expert schedulers. ASA combines a perception–decision–action loop with a three-stage offline preparation pipeline and a sched_ext-based runtime switch, enabling rapid adaptation to new hardware without retraining the core model. Its contributions include a time-weighted voting mechanism for stable workload recognition, a mapping table for scenario-specific policy selection, and a user-experience–oriented evaluation framework. Empirical results show ASA outperforms the default Linux scheduler in most scenarios, approaches oracle-level performance, and generalizes well to unseen hardware, suggesting practical impact for intelligent, adaptive OS scheduling.

Abstract

Modern operating system schedulers employ a single, static policy, which struggles to deliver optimal performance across the diverse and dynamic workloads of contemporary systems. This "one-policy-fits-all" approach leads to significant compromises in fairness, throughput, and latency, particularly with the rise of heterogeneous hardware and varied application architectures. This paper proposes a new paradigm: dynamically selecting the optimal policy from a portfolio of specialized schedulers rather than designing a single, monolithic one. We present the Adaptive Scheduling Agent (ASA), a lightweight framework that intelligently matches workloads to the most suitable "expert" scheduling policy at runtime. ASA's core is a novel, low-overhead offline/online approach. First, an offline process trains a universal, hardware-agnostic machine learning model to recognize abstract workload patterns from system behaviors. Second, at runtime, ASA continually processes the model's predictions using a time-weighted probability voting algorithm to identify the workload, then makes a scheduling decision by consulting a pre-configured, machine-specific mapping table to switch to the optimal scheduler via Linux's sched_ext framework. This decoupled architecture allows ASA to adapt to new hardware platforms rapidly without expensive retraining of the core recognition model. Our evaluation, based on a novel benchmark focused on user-experience metrics, demonstrates that ASA consistently outperforms the default Linux scheduler (EEVDF), achieving superior results in 86.4% of test scenarios. Furthermore, ASA's selections are near-optimal, ranking among the top three schedulers in 78.6% of all scenarios. This validates our approach as a practical path toward more intelligent, adaptive, and responsive operating system schedulers.

Paper Structure

This paper contains 33 sections, 2 equations, 8 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Background & Related Work
  3. The Enduring Challenge of OS Scheduling
  4. Scenario-Specific Static Policies
  5. AI-Assisted Dynamic Policies
  6. The Rise of Extensible Scheduling Frameworks
  7. ASA: Adaptive Scheduling Agent
  8. Agent-Based Architecture
  9. Perception module
  10. Decision module
  11. Action module
  12. Offline Preparation
  13. Cross-Platform Scheduling Challenge and Solution
  14. User Experience Oriented Evaluation Criteria
  15. Model Training & Scheduler Mapping Selection
  16. ...and 18 more sections

Figures (8)

  • Figure 1: Core Philosophy of ASA: A Paradigm Shift from Designing a Single Optimal Policy to Dynamically Selecting the Best Policy from a Portfolio of Experts.
  • Figure 2: Core Framework of the Adaptive Scheduling Agent
  • Figure 3: The Workflow within a Single Offline Preparation Stage.
  • Figure 4: The performance evaluation curve for the User Experience Oriented Evaluation Criteria
  • Figure 5: Global heatmap (scenarios × devices, relative to EEVDF)
  • ...and 3 more figures