📝 SummarXiv

Abstract

Distributed training of large deep-learning models often leads to failures, so checkpointing is commonly employed for recovery. State-of-the-art studies focus on frequent checkpointing for fast recovery from failures. However, it generates numerous checkpoints, incurring substantial costs and thus degrading training performance. Recently, differential checkpointing has been proposed to reduce costs, but it is limited to recommendation systems, so its application to general distributed training systems remains unexplored. This paper proposes LowDiff, an efficient frequent checkpointing framework that \textit{reuses} compressed gradients, serving as differential checkpoints to reduce cost. Furthermore, LowDiff incorporates a batched gradient write optimization to persist these differentials to storage efficiently. It also dynamically tunes both the checkpoint frequency and the batching size to maximize performance. We further enhance LowDiff with a layer-wise gradient reusing and snapshotting approach and a CPU-based asynchronous persistence strategy, enabling frequent checkpointing without gradient compression. Experiments on various workloads show that LowDiff can achieve checkpointing frequency up to per iteration with less than 3.1\% runtime overhead.