大规模智算互联网络仿真技术和平台

摘要：网络仿真作为高效的评估预测手段，已成为智算基础设施发展不可或缺的工具。总结大规模智算场景下网络仿真三大核心挑战：规模与复杂性、效率与可扩展性、精度与真实性，并系统剖析层次化建模、混合仿真、‌并行仿真、硬件加速、真实‌数据驱动等仿真关键技术。中兴通讯探索仿真新架构，设计实现面向数据流特征驱动的网络仿真平台，并在工程实践中验证了其高效性和可扩展性。

关键词：网络仿真；仿真建模；仿真加速；仿真校准

Abstract: As an efficient evaluation and prediction method, network simulation has become an indispensable tool for the development of intelligent computing infrastructure. This paper summarizes three core challenges of network simulation in large-scale intelligent computing scenarios: scale and complexity, efficiency and scalability, accuracy and fidelity, and systematically analyzes key technologies, including hierarchical modeling, hybrid simulation, parallel simulation, hardware acceleration, and real-data-driven simulation. ZTE Corporation explores a new simulation architecture, designs and implements a network simulation platform driven by data flow characteristics, and verifies its efficiency and scalability in engineering practice.

Keywords: network simulation; simulation modeling; simulation acceleration; simulation calibration