Communication-Efficient Distributed Computing Through Combinatorial Multi-Access Models

TL;DR

This paper introduces a novel distributed computing framework called MADC that uses combinatorial designs to reduce communication costs between mapper and reducer nodes, improving efficiency over traditional methods.

Contribution

It presents a new coding scheme based on t-designs for MADC, achieving minimal computation load and flexible reducer node selection, with reduced number of reducers compared to existing schemes.

Findings

Achieves computation load of 1 using combinatorial t-designs.

Reduces number of reducer nodes relative to existing schemes.

Demonstrates flexibility in choosing the number of reducers.

Abstract

This paper explores the multi-access distributed computing (MADC) model, a novel distributed computing framework where mapper and reducer nodes are distinct entities. Unlike traditional MapReduce frameworks, MADC leverages coding-theoretic techniques to minimize communication overhead without necessitating file replication across mapper nodes. We introduce a new approach utilizing combinatorial designs, specifically t-designs, to construct efficient coding schemes that achieve a computation load of 1. By establishing a connection between t-designs and MapReduce Arrays, we characterize the achievable communication loads and demonstrate the flexibility of our method in selecting the number of reducer nodes. The proposed scheme significantly reduces the number of reducer nodes relative to existing combinatorial topology schemes, at the expense of increased communication cost.