Daniel Lombraña González

Characterizing fault-tolerance of genetic algorithms in desktop grid systems

This paper presents a study of the fault-tolerant nature of Genetic Algorithms (GAs) on a real-world Desktop Grid System, without implementing any kind of fault-tolerance mechanism. The aim is to extend to parallel GAs previous works tackling fault-tolerance characterization in Genetic Programming. The results show that GAs are able to achieve a similar quality in results in comparison with a failure-free system in three of the six scenarios under study despite the system degradation. Additionally, we show that a small increase on the initial population size is a successful method to provide resilience to system failures in five of the scenarios. Such results suggest that Paralle GAs are inherently and naturally fault-tolerant.

Evolutionary Algorithms on Volunteer Computing Platforms: The MilkyWay@Home Project

Evolutionary algorithms (EAs) require large scale computing resources when tackling real world problems. Such computational requirement is derived from inherently complex fitness evaluation functions, large numbers of individuals per generation, and the number of iterations required by EAs to converge to a satisfactory solution. Therefore, any source of computing power can significantly benefit researchers using evolutionary algorithms. We present the use of volunteer computing (VC) as a platform for harnessing the computing resources of commodity machines that are nowadays present at homes, companies and institutions. Taking into account that currently desktop machines feature significant computing resources (dual cores, gigabytes of memory, gigabit network connections, etc.), VC has become a cost-effective platform for running time consuming evolutionary algorithms in order to solve complex problems, such as finding substructure in the Milky Way Galaxy, the problem we address in detail in this chapter.

Characterizing fault tolerance in genetic programming

Evolutionary Algorithms (EAs), and particularly Genetic Programming (GP), are techniques frequently employed to solve difficult real-life problems, which can require up to days or months of computation. One approach to reduce the time to solution is to use parallel computing on distributed platforms. Distributed platforms are prone to failures, and when these platforms are large and/or low-cost, failures are expected events rather than catastrophic exceptions. Therefore, fault tolerance and recovery techniques often become necessary. It turns out that Parallel GP (PGP) applications have an inherent ability to tolerate failures. This ability is quantified via simulation experiments performed using failure traces from real-world distributed platforms, namely, desktop grids (DGs), for two well-known GP problems. A simple technique is then proposed by which PGP applications can better tolerate the different, and often high, failures rates seen in different platforms.

Optimal tiling for the RNA base pairing problem

Dynamic programming is an important combinatorial optimization technique that has been widely used in various fields such as control theory, operations research, computational biology and computer science. Many authors have described parallel dynamic programming algorithms for the family of multistage problems. More scarce is the literature for the more general class of problems where dependences appear between non-consecutive stages. Among the important problems falling in this class is the RNA base pairing problem. In this study we propose a new parallel scheme for a large class of recurrences with triangular iteration space and nonuniform dependences that includes the RNA base pairing problem. We derive two different instances of this scheme that correspond to an horizontal and a vertical traverse of the iteration domain. We develop and extend the tiling approach for this particular class. We formulate and analytically solve the optimization problem determining the tile size that minimizes the total execution time of the tiled program on a distributed memory parallel machine. Our analyze is based on the BSP model, which assures the portability of the obtained results. The computational experiments carried out on the CRAY T3E behave according to the predictions of our theoretical model.

A peer-to-peer approach to genetic programming

This paper proposes a fine-grained parallelization of the Genetic Programming paradigm (GP) using the Evolvable Agent model (EvAg). The algorithm is decentralized in order to take full-advantage of a massively parallel Peer-to-Peer infrastructure. In this context, GP is particularly demanding due to its high requirements of computational power. To assess the viability of the approach, the EvAg model has been empirically analyzed in a simulated Peer-to-Peer environment where experiments were conducted on two well-known GP problems. Results show that the spatially structured nature of the algorithm is able to yield a good quality in the solutions. Additionally, parallelization improves times to solution by several orders of magnitude.

Customizable execution environments for evolutionary computation using BOINC + virtualization

Evolutionary algorithms (EAs) consume large amounts of computational resources, particularly when they are used to solve real-world problems that require complex fitness evaluations. Beside the lack of resources, scientists face another problem: the absence of the required expertise to adapt applications for parallel and distributed computing models. Moreover, the computing power of PCs is frequently underused at institutions, as desktops are usually devoted to administrative tasks. Therefore, the proposal in this work consists of providing a framework that allows researchers to massively deploy EA experiments by exploiting the computing power of their instituions’ PCs by setting up a Desktop Grid System based on the BOINC middleware. This paper presents a new model for running unmodified applications within BOINC with a web-based centralized management system for available resources. Thanks to this proposal, researchers can run scientific applications without modifying the application’s source code, and at the same time manage thousands of computers from a single web page. Summarizing, this model allows the creation of on-demand customized execution environments within BOINC that can be used to harness unused computational resources for complex computational experiments, such as EAs. To show the performance of this model, a real-world application of Genetic Programming was used and tested through a centrally-managed desktop grid infrastructure. Results show the feasibility of the approach that has allowed researchers to generate new solutions by means of an easy to use and manage distributed system.

Characterizing Fault-Tolerance in Evolutionary Algorithms

This chapter presents a study of the fault-tolerant nature of some of the best known Evolutionary Algorithms, namely Genetic Algorithms (GAs) and Genetic Programming (GP), on a real-world Desktop Grid System. We study the situation when no fault-tolerance mechanisms is employed. The results show that when parallel GAs and GPs are run on non-reliable distributed infrastructures -thus suffering degradation of available hardware- they can achieve results of a similar quality when compared with a failure-free platform in three of the six scenarios under study. Additionally, we show that increasing the initial population size is a successful method to provide resilience to system failures in five of the scenarios. Such results suggest that Parallel GAs and GPs are inherently and naturally fault-tolerant.

Volunteer Clouds and Citizen Cyberscience for LHC Physics

Computing for the LHC, and for HEP more generally, is traditionally viewed as requiring specialized infrastructure and software environments, and therefore not compatible with the recent trend in volunteer computing, where volunteers supply free processing time on ordinary PCs and laptops via standard Internet connections. In this paper, we demonstrate that with the use of virtual machine technology, at least some standard LHC computing tasks can be tackled with volunteer computing resources. Specifically, by presenting volunteer computing resources to HEP scientists as a volunteer cloud, essentially identical to a Grid or dedicated cluster from a job submission perspective, LHC simulations can be processed effectively. This article outlines both the technical steps required for such a solution and the implications for LHC computing as well as for LHC public outreach and for participation by scientists from developing regions in LHC research.

On the Intrinsic Fault-Tolerance Nature of Parallel Genetic Programming

In this paper we show how parallel genetic programming can run on a distributed system with volatile resources without any lack of efficiency. By means of a series of experiments, we test whether parallel GP - and consistently evolutionary algorithms - are intrinsically fault-tolerant. The interest of this result is crucial for researchers dealing with real-life problems in which parallel and distributed systems are required for obtaining results on a reasonable time. In that case, parallel GP tools will not require the inclusion of fault-tolerant computing techniques or libraries when running on meta-systems undergoing volatility, such us desktop grids offering public resource computing. We test the performance of the algorithm by studying the quality of solutions when running over distributed resources undergoing processors failures, when compared with a fault-free environment. This new feature, which shows its advantages, improves the dependability of the parallel genetic programming algorithm

Laser Dynamics Modelling and Simulation: An Application of Dynamic Load Balancing of Parallel Cellular Automata

This chapter reviews the application of a biologically inspired heuristic technique - Cellular Automata (CA) - for developing high performance simulations of a well known complex system: the laser.