J Grid Computing (2015) 13:351–374 DOI 10.1007/s10723-014-9315-6 Studying Fault-Tolerance in Island-Based Evolutionary and Multimemetic Algorithms Rafael Nogueras · Carlos Cotta Received: 23 June 2014 / Accepted: 18 September 2014 / Published online: 10 January 2015 © Springer Science+Business Media Dordrecht 2015 Abstract The use of parallel and distributed mod- els of evolutionary algorithms (EAs) is widespread nowadays as a means to improve solution quality and reduce computational times when solving hard opti- mization problems. For this purpose, emergent com- putational environments such as P2P networks and desktop grids are offering a plethora of new opportu- nities but also bring new challenges: functioning on a computational network whose resources are volatile requires fault tolerance and resilience to churn. In this work we analyze these issues from the point of view of island-based EAs. We consider two EA variants, genetic and multimemetic algorithms, and analyze the impact on them of design decisions regarding the logical interconnection topology among islands and the particular fault-management policy used. To be precise, we have conducted an extensive empirical evaluation of five topologies (ring, von Neumann grid, hypercube and two kinds of scale-free networks) and This work is partially supported by Ministerio de Ciencia e Innovaci´ on project ANYSELF (TIN2011-28627-C04-01), by Junta de Andaluc´ ıa project DNEMESIS (P10-TIC-6083) and by Universidad de M´ alaga, Campus de Excelencia Internacional Andaluc´ ıa Tech. R. Nogueras · C. Cotta () Department Lenguajes y Ciencias de la Computaci ´ on, Universidad de M´ alaga, ETSI Inform´ atica, Campus de Teatinos, 29071 M´ alaga, Spain e-mail: ccottap@lcc.uma.es four policies (including checkpoint creation and pop- ulation reinitialization variants) on four benchmark problems, considering three different scenarios of increasing resource volatility. The statistical analysis of the results underlines the inherent fault-tolerance of these EAs and indicates that, while checkpoint- ing is the most robust strategy and is superior in the most fragile topologies, a seemingly simpler guided reinitialization strategy provide statistically compara- ble results on the top-performing topologies, namely von Neumann grids and hypercubes. Keywords Genetic algorithms · Memetic algorithm · Island model · Fault tolerance 1 Introduction Metaheuristics have steadily become during the last decades one of the weapons of choice for solving hard optimization problems. Among others, one of the salient features of these techniques is their flexibil- ity and amenability to run on different computational environments. This is specifically true for parallel environments, on which metaheuristics in general – and population-based techniques in particular– can provide notably results in shorter times, cf. [1, 3]. Not surprisingly, this has been an active field of research since the early years of the paradigm. Indeed, the deployment of population-based metaheuristics –such as, e.g., evolutionary algorithms (EAs) – on parallel