Mesh-Attention: A New Communication-Efficient Distributed Attention with Improved Data Locality

TL;DR

Mesh-Attention introduces a novel 2D tiled distributed attention algorithm that significantly reduces communication overhead and improves scalability for large language models on multi-GPU systems.

Contribution

The paper presents Mesh-Attention, a new matrix-based distributed attention method with a 2D tiling approach, outperforming Ring-Attention in efficiency and scalability.

Findings

Achieves up to 3.4x speedup over existing methods.

Reduces communication volume by up to 85.4%.

Maintains high performance as system scales to 256 GPUs.

Abstract

Distributed attention is a fundamental problem for scaling context window for Large Language Models (LLMs). The state-of-the-art method, Ring-Attention, suffers from scalability limitations due to its excessive communication traffic. This paper proposes a new distributed attention algorithm, Mesh-Attention, by rethinking the design space of distributed attention with a new matrix-based model. Our method assigns a two-dimensional tile -- rather than one-dimensional row or column -- of computation blocks to each GPU to achieve higher efficiency through lower communication-computation (CommCom) ratio. The general approach covers Ring-Attention as a special case, and allows the tuning of CommCom ratio with different tile shapes. Importantly, we propose a greedy algorithm that can efficiently search the scheduling space within the tile with restrictions that ensure efficient communication among GPUs. The theoretical analysis shows that Mesh-Attention leads to a much lower communication complexity and exhibits good scalability comparing to other current algorithms. Our extensive experiment results show that Mesh-Attention can achieve up to 3.4x speedup (2.9x on average) and reduce the communication volume by up to 85.4% (79.0% on average) on 256 GPUs. Our scalability results further demonstrate that Mesh-Attention sustains superior performance as the system scales, substantially reducing overhead in large-scale deployments. The results convincingly confirm the advantage of Mesh-Attention.