ScaleSim: Serving Large-Scale Multi-Agent Simulation with Invocation Distance-Based Memory Management

TL;DR

ScaleSim is a memory-efficient system that improves large-scale multi-agent LLM simulations by using invocation distance to optimize memory management, enabling faster performance and better scalability.

Contribution

We introduce invocation distance, a novel abstraction, and develop ScaleSim, a system that enhances memory management and performance in large-scale multi-agent LLM simulations.

Findings

Achieves up to 1.74x speedup over SGLang.

Effectively manages GPU memory for large agent populations.

Supports diverse agent-specific memory with modular design.

Abstract

LLM-based multi-agent simulations are increasingly adopted across application domains, but remain difficult to scale due to GPU memory pressure. Each agent maintains private GPU-resident states, including models, prefix caches, and adapters, which quickly exhaust device memory as the agent count grows. We identify two key properties of these workloads: sparse agent activation and an estimable agent invocation order. Based on an analysis of representative workload classes, we introduce invocation distance, a unified abstraction that estimates the relative order in which agents will issue future LLM requests. Leveraging this abstraction, we present ScaleSim, a memory-efficient LLM serving system for large-scale multi-agent simulations. ScaleSim enables proactive prefetching and priority-based eviction, supports diverse agent-specific memory through a modular interface, and achieves up to 1.74x speedup over SGLang on simulation benchmarks.