Ethane: A Heterogeneous Parallel Search Algorithm for Heterogeneous Platforms

TL;DR

Ethane introduces a novel heterogeneous parallel search algorithm inspired by chemical structures, effectively leveraging diverse hardware to improve search efficiency and robustness over traditional algorithms.

Contribution

The paper proposes Ethane, a chemical-inspired heterogeneous island model, and a schema HydroCM for designing diverse parallel metaheuristics on heterogeneous platforms.

Findings

Ethane outperforms traditional panmitic algorithms in speed.

Ethane demonstrates increased robustness in search problems.

The HydroCM schema facilitates designing diverse algorithms for heterogeneous systems.

Abstract

In this paper we present Ethane, a parallel search algorithm specifically designed for its execution on heterogeneous hardware environments. With Ethane we propose an algorithm inspired in the structure of the chemical compound of the same name, implementing a heterogeneous island model based in the structure of its chemical bonds. We also propose a schema for describing a family of parallel heterogeneous metaheuristics inspired by the structure of hydrocarbons in Nature, HydroCM (HydroCarbon inspired Metaheuristics), establishing a resem- blance between atoms and computers, and between chemical bonds and communication links. Our goal is to gracefully match computers of different power to algorithms of different behavior (GA and SA in this study), all them collaborating to solve the same problem. The analysis will show that Ethane, though simple, can solve search problems in a faster and more robust way than well-known panmitic and distributed algorithms very popular in the literature.