HammingMesh: A Network Topology for Large-Scale Deep Learning

TL;DR

HammingMesh is a new network topology designed to improve data movement efficiency and flexibility in large-scale deep learning training systems, addressing current bottlenecks and supporting high bandwidth needs.

Contribution

We introduce HammingMesh, a novel network topology optimized for large-scale deep learning, offering high bandwidth, low cost, and flexible job scheduling capabilities.

Findings

Supports full bandwidth and isolation for deep learning jobs

Provides high global bandwidth for generic traffic

Enables scalable large-scale deep learning systems

Abstract

Numerous microarchitectural optimizations unlocked tremendous processing power for deep neural networks that in turn fueled the AI revolution. With the exhaustion of such optimizations, the growth of modern AI is now gated by the performance of training systems, especially their data movement. Instead of focusing on single accelerators, we investigate data-movement characteristics of large-scale training at full system scale. Based on our workload analysis, we design HammingMesh, a novel network topology that provides high bandwidth at low cost with high job scheduling flexibility. Specifically, HammingMesh can support full bandwidth and isolation to deep learning training jobs with two dimensions of parallelism. Furthermore, it also supports high global bandwidth for generic traffic. Thus, HammingMesh will power future large-scale deep learning systems with extreme bandwidth requirements.