Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

AgentRM: An OS-Inspired Resource Manager for LLM Agent Systems

Jianshu She

Abstract

Large Language Model (LLM) agent systems have experienced rapid adoption across diverse domains, yet they suffer from critical user experience problems that limit their practical deployment. Through an empirical analysis of over 40,000 GitHub issues from six major agent frameworks (OpenClaw, AutoGen, CrewAI, LangGraph, Codex, Claude Code), we identify two fundamental resource management challenges: (1) scheduling failures leading to system unresponsiveness due to blocking, zombie processes, and rate limit cascades, and (2) context degradation causing agent "amnesia" from unbounded memory growth and poor retention policies. Drawing inspiration from decades of operating systems research, we present AgentRM, a middleware resource manager that treats agent resources analogously to OS resources. AgentRM employs a Multi-Level Feedback Queue (MLFQ) scheduler with zombie reaping and rate-limit-aware admission control, coupled with a three-tier Context Lifecycle Manager that implements adaptive compaction and hibernation mechanisms. Our evaluation demonstrates significant improvements: AgentRM-MLFQ reduces P95 latency by 86%, decreases lane waste by 96%, and increases throughput by 168% while eliminating zombie agents (0 vs. 29 baseline). AgentRM-CLM achieves 100% key information retention with 95% quality score compared to 65.1% retention and 87% quality for existing approaches, albeit with higher compaction costs (34,330 vs. 17,212 tokens).

AgentRM: An OS-Inspired Resource Manager for LLM Agent Systems

Abstract

Large Language Model (LLM) agent systems have experienced rapid adoption across diverse domains, yet they suffer from critical user experience problems that limit their practical deployment. Through an empirical analysis of over 40,000 GitHub issues from six major agent frameworks (OpenClaw, AutoGen, CrewAI, LangGraph, Codex, Claude Code), we identify two fundamental resource management challenges: (1) scheduling failures leading to system unresponsiveness due to blocking, zombie processes, and rate limit cascades, and (2) context degradation causing agent "amnesia" from unbounded memory growth and poor retention policies. Drawing inspiration from decades of operating systems research, we present AgentRM, a middleware resource manager that treats agent resources analogously to OS resources. AgentRM employs a Multi-Level Feedback Queue (MLFQ) scheduler with zombie reaping and rate-limit-aware admission control, coupled with a three-tier Context Lifecycle Manager that implements adaptive compaction and hibernation mechanisms. Our evaluation demonstrates significant improvements: AgentRM-MLFQ reduces P95 latency by 86%, decreases lane waste by 96%, and increases throughput by 168% while eliminating zombie agents (0 vs. 29 baseline). AgentRM-CLM achieves 100% key information retention with 95% quality score compared to 65.1% retention and 87% quality for existing approaches, albeit with higher compaction costs (34,330 vs. 17,212 tokens).
Paper Structure (42 sections, 5 equations, 6 figures, 9 tables, 2 algorithms)

This paper contains 42 sections, 5 equations, 6 figures, 9 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Background and Motivation
  3. Empirical Study of Agent System Failures
  4. Problem Classification
  5. System Model and Problem Formulation
  6. Formal Definitions
  7. Scheduling Problem Formulation
  8. Context Management Problem Formulation
  9. AgentRM Architecture
  10. Overview
  11. Agent Scheduler
  12. Multi-Level Feedback Queue (MLFQ)
  13. Zombie Reaper
  14. Rate Limit-Aware Scheduling
  15. Fairness Mechanisms
  16. ...and 27 more sections

Figures (6)

  • Figure 1: Architecture comparison: (a) Traditional agent systems with direct model access, (b) AgentRM with centralized resource management.
  • Figure 2: P95 latency comparison across scheduling algorithms and workload scenarios.
  • Figure 3: Comprehensive scheduling performance comparison across all test scenarios.
  • Figure 4: Zombie analysis showing lane hold times, recovery patterns, and resource waste across scenarios.
  • Figure 5: Context quality vs. utilization trade-offs for different management strategies.
  • ...and 1 more figures