BootSeer: Analyzing and Mitigating Initialization Bottlenecks in Large-Scale LLM Training

TL;DR

This paper investigates startup overhead in large-scale LLM training, characterizes its components, and introduces BootSeer, a system that reduces startup delays by 50% through innovative techniques, improving efficiency in industrial settings.

Contribution

It provides the first detailed analysis of LLM startup overhead and proposes BootSeer, a novel system with techniques that significantly reduce startup time in production environments.

Findings

Startup overhead accounts for over 3.5% GPU time in large LLM training.

BootSeer reduces startup overhead by 50% in real workloads.

Analysis of startup components informs targeted optimizations.

Abstract

Large Language Models (LLMs) have become a cornerstone of modern AI, driving breakthroughs in natural language processing and expanding into multimodal jobs involving images, audio, and video. As with most computational software, it is important to distinguish between ordinary runtime performance and startup overhead. Prior research has focused on runtime performance: improving training efficiency and stability. This work focuses instead on the increasingly critical issue of startup overhead in training: the delay before training jobs begin execution. Startup overhead is particularly important in large, industrial-scale LLMs, where failures occur more frequently and multiple teams operate in iterative update-debug cycles. In one of our training clusters, more than 3.5% of GPU time is wasted due to startup overhead alone. In this work, we present the first in-depth characterization of LLM training startup overhead based on real production data. We analyze the components of startup cost, quantify its direct impact, and examine how it scales with job size. These insights motivate the design of Bootseer, a system-level optimization framework that addresses three primary startup bottlenecks: (a) container image loading, (b) runtime dependency installation, and (c) model checkpoint resumption. To mitigate these bottlenecks, Bootseer introduces three techniques: (a) hot block record-and-prefetch, (b) dependency snapshotting, and (c) striped HDFS-FUSE. Bootseer has been deployed in a production environment and evaluated on real LLM training workloads, demonstrating a 50% reduction in startup overhead.