A Computation vs Communication Tradeoff in Distributed Matrix Multiplication over Finite Fields

Malladi, TS and Rajan, BS (2019) A Computation vs Communication Tradeoff in Distributed Matrix Multiplication over Finite Fields. In: 2019 IEEE International Conference on Communications, ICC 2019, 20 - 24 May 2019, Shanghai.

PDF ICC_2019.pdf - Published Version Restricted to Registered users only Download (16MB) | Request a copy

Abstract

One of the key problems affecting the performance of algorithms running on a distributed storage system is straggler nodes. There have been works on reducing the recovery threshold (i.e., the minimum number of workers the master needs to wait for, in order to compute the final output) in the case of massive matrix multiplication problems. These works generally consider matrices over arbitrary field i.e., matrices over fields of characteristic both zero and prime. In this paper, we focus on multiplication of matrices over finite fields in a distributed storage system and exploit some properties of finite fields to achieve a fractional improvement in the recovery threshold, with a tradeoff in computational complexity. The proposed coding idea is applicable without restriction on the number of workers and the field size.

Item Type:	Conference Paper
Publication:	IEEE International Conference on Communications
Publisher:	Institute of Electrical and Electronics Engineers Inc.
Additional Information:	The copyright for this article belongs to Institute of Electrical and Electronics Engineers Inc.
Keywords:	Multiprocessing systems, Computation vs communications; Distributed matrix; Distributed storage system; Field size; Finite fields; MAtrix multiplication; Performance of algorithm; Recovery thresholds, Matrix algebra
Department/Centre:	Division of Electrical Sciences > Electrical Communication Engineering
Date Deposited:	28 Dec 2022 05:07
Last Modified:	28 Dec 2022 05:07
URI:	https://eprints.iisc.ac.in/id/eprint/78594

Actions (login required)

View Item