A Unified Coding Framework for Distributed Computing with Straggling Servers
09 February 2017
We propose a unified framework for distributed computing incorporating recent coding ideas in [1] and [2] to speed up a distributed matrix multiplication application. We utilize MDS codes to create redundant computations to deal with straggling server nodes, and also apply coded multicasting during the data shuffling to reduce the total communication load. The proposed framework provides a flexible approach to trade computation latency for communication load.
