Optimizing Reconfigurable Optical Datacenters: The Power of Randomization

TL;DR

This paper presents a randomized online algorithm for optimizing reconfigurable optical datacenter topologies, significantly improving performance over deterministic methods and validated through real-world workload simulations.

Contribution

It introduces a $O( ext{log} b)$-competitive randomized algorithm for online topology optimization, proving its asymptotic optimality and outperforming deterministic approaches.

Findings

The randomized algorithm is exponentially better than deterministic ones.

The algorithm is asymptotically optimal.

Simulations confirm practical effectiveness with real workloads.

Abstract

Reconfigurable optical topologies are a promising new technology to improve datacenter network performance and cope with the explosive growth of traffic. In particular, these networks allow to directly and adaptively connect racks between which there is currently much traffic, hence making an optimal use of the bandwidth capacity by avoiding multi-hop forwarding. This paper studies the dynamic optimization of such reconfigurable topologies, by adapting the network to the traffic in an online manner. The underlying algorithmic problem can be described as an online maximum weight $b$-matching problem, a~generalization of maximum weight matching where each node has at most $b \geq 1$ incident matching edges. We make the case for a randomized approach for matching optimization. Our main contribution is a $O(\log b)$-competitive algorithm and we show that it is asymptotically optimal. This algorithm is hence exponentially better than the best possible deterministic online algorithm. We complement our theoretical results with extensive trace-driven simulations, based on real-world datacenter workloads.