TonY: An Orchestrator for Distributed Machine Learning Jobs
This addresses the problem of efficient orchestration for distributed ML training, primarily for organizations like LinkedIn, but it is incremental as it builds on existing orchestration concepts.
The paper tackles the complexity of managing distributed machine learning jobs by introducing TonY, an open-source orchestrator developed at LinkedIn to handle resource contention, configurations, monitoring, and fault tolerance.
Training machine learning (ML) models on large datasets requires considerable computing power. To speed up training, it is typical to distribute training across several machines, often with specialized hardware like GPUs or TPUs. Managing a distributed training job is complex and requires dealing with resource contention, distributed configurations, monitoring, and fault tolerance. In this paper, we describe TonY, an open-source orchestrator for distributed ML jobs built at LinkedIn to address these challenges.