On Competitiveness of Dynamic Replication for Distributed Data Access

TL;DR

This paper analyzes the online cost optimization for dynamic data replication in distributed storage, proving bounds on algorithm competitiveness and evaluating performance with real-world traces.

Contribution

It corrects a previous algorithm's claimed performance, establishes new competitive ratio bounds, and introduces an improved online algorithm for data replication.

Findings

Previous algorithm's claimed ratio of 2 is incorrect.

No deterministic online algorithm can have a ratio better than 2.

Proposed algorithm achieves a ratio of max{2, min{γ, 3}}.

Abstract

This paper studies an online cost optimization problem for distributed storage and access. The goal is to dynamically create and delete copies of data objects over time at geo-distributed servers to serve access requests and minimize the total storage and network cost. We revisit a recent algorithm in the literature and show that it does not have a competitive ratio of $2$ as claimed by constructing a counterexample. We further prove that no deterministic online algorithm can achieve a competitive ratio bounded by $2$ for the general cost optimization problem. We develop an online algorithm and prove that it achieves a competitive ratio of $\max\{2, \min\{\gamma, 3\}\}$, where $\gamma$ is the max/min storage cost ratio among all servers. Examples are given to confirm the tightness of competitive analysis. We also empirically evaluate algorithms using real object access traces.