An genetic algorithm approach for profiling computational performance measures
This paper present a Genetic Algorithm(GA) approach for clustering data metric of computational performance measures collected from vmstat and sar tools. The proposed work models the genes, chromosomes, species and environment based on the dataset and presents an algorithm to analyze patterns and classify the records. The proposed method submits the performance information to an N-Dimensional Histogram in order to obtain the distribution of data that is used as input to the cluster initialization. The individual from each species undergoes successive crossover, mutation and selection operations to improve and evolve the initial population to a given environment state. The fitness-function is determined by the N-Dimensional Euclidean distance. The selection method is based on the Roulette-Wheel Selection, Elitist Selection and Truncation Selection. The results presented were obtained from seven test scenarios.