HyGra: Accelerating Network-State Simulation for LLM Training in DCNs via Adaptive Packet-Flow Granularity

2026-03-13

AI Summary

Existing network simulators struggle to simultaneously achieve efficiency and fidelity in large language model (LLM) training: packet-level simulation offers high accuracy but incurs substantial overhead, whereas flow-level simulation is efficient yet introduces significant distortion. This work proposes a hybrid-granularity network state simulator that, for the first time, incorporates an adaptive granularity-switching mechanism—employing packet-level simulation during non-steady-state phases and dynamically transitioning to flow-level simulation during steady-state periods to balance accuracy and speed. The approach requires no specialized hardware, runs on a single machine, and integrates seamlessly with existing simulators. Experiments on real-world LLM workloads, including ChatGPT, DeepSeek, and Qwen, demonstrate speedups of up to 15.4× under single parallelism strategies and 7.8× under hybrid parallelism, all while preserving high simulation fidelity.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Search and Optimization: Sampling/Simulation-based Search
• Natural Language Processing: (Large) Language Models

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

In recent years, large language models (LLMs) have driven substantial intelligent transformation across diverse industries. Commercial LLM training is typically performed over data center networks (DCNs) comprising hundreds to thousands of GPUs, with multiple devices collocated per node. As network scale expands, inter-node communication becomes a primary bottleneck to training efficiency. Network-state simulators therefore play a crucial role by enabling cost-effective evaluation of network configurations and parallelization strategies through faithful emulation of DCN dynamics during LLM training. However, existing simulators are constrained by a efficiency-fidelity tradeoff, as packet-level simulators (PLSs) incur prohibitive runtime overhead, whereas flow-level simulators (FLSs) compromise essential modeling accuracy. In this paper, we develop \texttt{HyGra}, a hybrid-granularity network-state simulator that exploits intrinsic network dynamics in LLM training to adaptively switch simulation granularity. Specifically, \texttt{HyGra} employs packet-level simulation during non-steady phases with transient fluctuations and flow-level simulation during steady phases with periodic patterns, thereby accelerating execution while preserving high fidelity. Moreover, it requires no specialized hardware, supports single-machine deployment, and is compatible with existing simulators. Experiments based representative commercial LLM workloads, including ChatGPT, DeepSeek, and Qwen, show that \texttt{HyGra} achieves up to 15.4$\times$ speedup under single parallelization strategy and 7.8$\times$ under hybrid parallelization strategies while maintaining high accuracy.

Problem

• network-state simulation
• LLM training
• data center networks
• simulation granularity
• efficiency-fidelity tradeoff

Innovation

• hybrid-granularity simulation
• adaptive packet-flow granularity
• network-state simulation
• LLM training acceleration
• data center networks