Understanding Stragglers in Large Model Training Using What-if Analysis

TL;DR

This paper investigates the causes and patterns of stragglers in large language model training, using a comprehensive five-month trace and what-if analysis to understand their impact and origins.

Contribution

It introduces a detailed study of stragglers in LLM training with a novel what-if analysis approach based on real trace data.

Findings

Stragglers significantly impact training performance.

Stragglers exhibit both temporal and spatial patterns.

Root causes include complex factors beyond hardware failures.

Abstract

Large language model (LLM) training is one of the most demanding distributed computations today, often requiring thousands of GPUs with frequent synchronization across machines. Such a workload pattern makes it susceptible to stragglers, where the training can be stalled by few slow workers. At ByteDance we find stragglers are not trivially always caused by hardware failures, but can arise from multiple complex factors. This work aims to present a comprehensive study on the straggler issues in LLM training, using a five-month trace collected from our ByteDance LLM training cluster. The core methodology is what-if analysis that simulates the scenario without any stragglers and contrasts with the actual case. We use this method to study the following questions: (1) how often do stragglers affect training jobs, and what effect do they have on job performance; (2) do stragglers exhibit temporal or spatial patterns; and (3) what are the potential root causes for stragglers?