m4: A Learned Flow-level Network Simulator
Chenning Li 1 , Anton A. Zabreyko 1 , Arash Nasr-Esfahany 1 , Kevin Zhao 2 , Prateesh Goyal 3 , Mohammad Alizadeh 1 , and Thomas Anderson 2 1 MIT CSAIL, 2 University of Washington, 3 Microsoft Research · 2025
Read m4: A Learned Flow-level Network Simulator by Chenning Li 1 , Anton A. Zabreyko 1 , Arash Nasr-Esfahany 1 , Kevin Zhao 2 , Prateesh Goyal 3 , Mohammad Alizadeh 1 , and Thomas Anderson 2 1 MIT CSAIL, 2 University of Washington, 3 Microsoft Research (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-01770.txt.
Description
Chenning Li and six co-authors learn to simulate networks in twenty twenty-five. Data center networks are usually modeled at flow level for speed. Flow simulators treat traffic as continuous streams with assigned rates. That abstraction skips queuing, congestion control, and retransmissions, so it can be inaccurate.
Their simulator, m4, uses machine learning to close the accuracy gap. It learns the dynamics of the network of interest directly. Its architecture splits state updates into spatial and temporal parts. Each part gets a neural network suited to it. Dense supervision comes from predicting intermediate network metrics, not just final outcomes. The result stays scalable like a flow simulator while tracking packet-level effects. Readers get simulation that learns the network it models.
Catalog record
| CREATOR | Chenning Li 1 , Anton A. Zabreyko 1 , Arash Nasr-Esfahany 1 , Kevin Zhao 2 , Prateesh Goyal 3 , Mohammad Alizadeh 1 , and Thomas Anderson 2 1 MIT CSAIL, 2 University of Washington, 3 Microsoft Research |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 61 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2503-01770