PerfTracker: Online Performance Troubleshooting for Large-scale Model Training in Production
This addresses a critical problem for organizations running large-scale AI training in production, offering a novel solution for performance troubleshooting in complex GPU environments.
The paper tackles the challenge of diagnosing performance issues in large-scale model training on GPU clusters by introducing PerfTracker, an online troubleshooting system that uses fine-grained profiling and differential observability to localize root causes in hardware and software, scaling to clusters with O(10,000) GPUs.
Troubleshooting performance problems of large model training (LMT) is immensely challenging, due to unprecedented scales of modern GPU clusters, the complexity of software-hardware interactions, and the data intensity of the training process. Existing troubleshooting approaches designed for traditional distributed systems or datacenter networks fall short and can hardly apply to real-world training systems. In this paper, we present PerfTracker, the first online troubleshooting system utilizing fine-grained profiling, to diagnose performance issues of large-scale model training in production. PerfTracker can diagnose performance issues rooted in both hardware (e.g., GPUs and their interconnects) and software (e.g., Python functions and GPU operations). It scales to LMT on modern GPU clusters. PerfTracker effectively summarizes runtime behavior patterns of fine-grained LMT functions via online profiling, and leverages differential observability to localize the root cause with minimal production impact. PerfTracker has been deployed as a production service for large-scale GPU clusters of O(10, 000) GPUs (product homepage https://help.aliyun.com/zh/pai/user-guide/perftracker-online-performance-analysis-diagnostic-tool). It has been used to diagnose a variety of difficult performance issues.