Mycroft: Tracing Dependencies in Collective Communication Towards Reliable LLM Training

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Deng, Yangtao, Zhang, Lei, Wang, Qinlong, Zhi, Xiaoyun, Zhang, Xinlei, Jiang, Zhuo, Xu, Haohan, Wang, Lei, Song, Zuquan, Liu, Gaohong, Bai, Yang, Wang, Shuguang, Xiao, Wencong, Ye, Jianxi, Yu, Minlan, Xu, Hong
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911136449822720
author Deng, Yangtao
Zhang, Lei
Wang, Qinlong
Zhi, Xiaoyun
Zhang, Xinlei
Jiang, Zhuo
Xu, Haohan
Wang, Lei
Song, Zuquan
Liu, Gaohong
Bai, Yang
Wang, Shuguang
Xiao, Wencong
Ye, Jianxi
Yu, Minlan
Xu, Hong
author_facet Deng, Yangtao
Zhang, Lei
Wang, Qinlong
Zhi, Xiaoyun
Zhang, Xinlei
Jiang, Zhuo
Xu, Haohan
Wang, Lei
Song, Zuquan
Liu, Gaohong
Bai, Yang
Wang, Shuguang
Xiao, Wencong
Ye, Jianxi
Yu, Minlan
Xu, Hong
contents Reliability is essential for ensuring efficiency in LLM training. However, many real-world reliability issues remain difficult to resolve, resulting in wasted resources and degraded model performance. Unfortunately, today's collective communication libraries operate as black boxes, hiding critical information needed for effective root cause analysis. We propose Mycroft, a lightweight distributed tracing and root cause analysis system designed to address previously hidden reliability issues in collective communication. Mycroft's key idea is to trace collective communication states and leverage internal control and data dependencies to resolve reliability problems in LLM training. Mycroft has been deployed at ByteDance for over six months to debug collective communication related issues at runtime. It detected anomalies within 15 seconds in 90% of cases and identified the root cause within 20 seconds in 60% of cases. We also conducted extensive fault injection experiments to demonstrate Mycroft's capability and efficiency.
format Preprint
id arxiv_https___arxiv_org_abs_2509_03018
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Mycroft: Tracing Dependencies in Collective Communication Towards Reliable LLM Training
Deng, Yangtao
Zhang, Lei
Wang, Qinlong
Zhi, Xiaoyun
Zhang, Xinlei
Jiang, Zhuo
Xu, Haohan
Wang, Lei
Song, Zuquan
Liu, Gaohong
Bai, Yang
Wang, Shuguang
Xiao, Wencong
Ye, Jianxi
Yu, Minlan
Xu, Hong
Distributed, Parallel, and Cluster Computing
Machine Learning
Reliability is essential for ensuring efficiency in LLM training. However, many real-world reliability issues remain difficult to resolve, resulting in wasted resources and degraded model performance. Unfortunately, today's collective communication libraries operate as black boxes, hiding critical information needed for effective root cause analysis. We propose Mycroft, a lightweight distributed tracing and root cause analysis system designed to address previously hidden reliability issues in collective communication. Mycroft's key idea is to trace collective communication states and leverage internal control and data dependencies to resolve reliability problems in LLM training. Mycroft has been deployed at ByteDance for over six months to debug collective communication related issues at runtime. It detected anomalies within 15 seconds in 90% of cases and identified the root cause within 20 seconds in 60% of cases. We also conducted extensive fault injection experiments to demonstrate Mycroft's capability and efficiency.
title Mycroft: Tracing Dependencies in Collective Communication Towards Reliable LLM Training
topic Distributed, Parallel, and Cluster Computing
Machine Learning
url https://arxiv.org/abs/2509.03018