Communication-Efficient (Weighted) Reservoir Sampling from Fully Distributed Data Streams

TL;DR

This paper introduces communication-efficient algorithms for weighted and unweighted reservoir sampling from distributed data streams, demonstrating scalability and speedups in large-scale distributed systems.

Contribution

It proposes novel fully distributed algorithms for reservoir sampling that are communication-efficient and scalable to large distributed data streams.

Findings

Experimental results show good speedups on up to 256 nodes.

Theoretical analysis indicates potential for larger-scale deployment.

Algorithms perform well in distributed streaming environments.

Abstract

We consider communication-efficient weighted and unweighted (uniform) random sampling from distributed data streams presented as a sequence of mini-batches of items. This is a natural model for distributed streaming computation, and our goal is to showcase its usefulness. We present and analyze fully distributed, communication-efficient algorithms for both versions of the problem. An experimental evaluation of weighted reservoir sampling on up to 256 nodes (5120 processors) shows good speedups, while theoretical analysis promises further scaling to much larger machines.