LAGC: Lazily Aggregated Gradient Coding for Straggler-Tolerant and Communication-Efficient Distributed Learning
This work addresses efficiency and robustness issues in distributed learning systems, particularly for applications using Parameter Server architectures, by providing a unified analysis and incremental improvements over existing techniques.
The paper tackled the problem of straggler delays and communication bottlenecks in distributed learning by introducing novel strategies, Lazily Aggregated Gradient Coding (LAGC) and Grouped-LAG (G-LAG), which combine gradient coding and grouping with adaptive selection to improve performance. Results showed that G-LAG achieved the best wall-clock time and communication performance while maintaining low computational cost for two representative worker computing time distributions.
Gradient-based distributed learning in Parameter Server (PS) computing architectures is subject to random delays due to straggling worker nodes, as well as to possible communication bottlenecks between PS and workers. Solutions have been recently proposed to separately address these impairments based on the ideas of gradient coding, worker grouping, and adaptive worker selection. This paper provides a unified analysis of these techniques in terms of wall-clock time, communication, and computation complexity measures. Furthermore, in order to combine the benefits of gradient coding and grouping in terms of robustness to stragglers with the communication and computation load gains of adaptive selection, novel strategies, named Lazily Aggregated Gradient Coding (LAGC) and Grouped-LAG (G-LAG), are introduced. Analysis and results show that G-LAG provides the best wall-clock time and communication performance, while maintaining a low computational cost, for two representative distributions of the computing times of the worker nodes.