Adaptive Worker Grouping For Communication-Efficient and Straggler-Tolerant Distributed SGD
This work addresses communication efficiency and straggler tolerance for distributed machine learning systems, representing an incremental improvement over existing adaptive methods.
The paper tackles the problem of performance degradation in distributed SGD due to stragglers and high communication load by proposing G-CADA, a grouping-based scheme that reduces wall-clock convergence time by 30% and communication load by 40% compared to benchmarks.
Wall-clock convergence time and communication load are key performance metrics for the distributed implementation of stochastic gradient descent (SGD) in parameter server settings. Communication-adaptive distributed Adam (CADA) has been recently proposed as a way to reduce communication load via the adaptive selection of workers. CADA is subject to performance degradation in terms of wall-clock convergence time in the presence of stragglers. This paper proposes a novel scheme named grouping-based CADA (G-CADA) that retains the advantages of CADA in reducing the communication load, while increasing the robustness to stragglers at the cost of additional storage at the workers. G-CADA partitions the workers into groups of workers that are assigned the same data shards. Groups are scheduled adaptively at each iteration, and the server only waits for the fastest worker in each selected group. We provide analysis and experimental results to elaborate the significant gains on the wall-clock time, as well as communication load and computation load, of G-CADA over other benchmark schemes.