Congestion Control System Optimization with Large Language Models

2025-08-22

AI Summary

Existing congestion control algorithms exhibit poor generalizability and high tuning costs in heterogeneous networks. This paper proposes the first large language model (LLM)-driven automated congestion control optimization framework. It employs structured prompt engineering to guide LLMs in generating executable algorithm code, integrates with network simulation platforms (e.g., ns-3) for closed-loop evaluation across diverse scenarios and metrics—including throughput, latency, and fairness—and incorporates a statistical surrogate model to enable efficient Bayesian optimization. The approach reduces traditional manual tuning—typically requiring months—to just several days, without relying on domain expertise or prior knowledge. When deployed in real-world QUIC stacks, the LLM-generated optimal algorithm achieves a 27% average throughput improvement over BBRv3 and demonstrates significantly enhanced robustness across varying bandwidth, round-trip delay, and packet loss rate conditions.

Technology Category

• Search and Optimization: Algorithm Configuration
• Planning, Routing, and Scheduling: Planning with Language Models
• Machine Learning: Optimization

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Congestion control is a fundamental component of Internet infrastructure, and researchers have dedicated considerable effort to developing improved congestion control algorithms. However, despite extensive study, existing algorithms continue to exhibit suboptimal performance across diverse network environments. In this paper, we introduce a novel approach that automatically optimizes congestion control algorithms using large language models (LLMs). Our framework consists of a structured algorithm generation process, an emulation-based evaluation pipeline covering a broad range of network conditions, and a statistically guided method to substantially reduce evaluation time. Empirical results from four distinct LLMs validate the effectiveness of our approach. We successfully identify algorithms that achieve up to 27% performance improvements over the original BBR algorithm in a production QUIC implementation. Our work demonstrates the potential of LLMs to accelerate the design of high-performance network algorithms and paves the way for broader applications in networking systems.

Problem

• Optimizing congestion control algorithms using large language models
• Addressing suboptimal performance across diverse network environments
• Reducing evaluation time through statistically guided methods

Innovation

• LLMs optimize congestion control algorithms automatically
• Emulation-based evaluation covers diverse network conditions
• Statistically guided method reduces evaluation time significantly