ClusterFlow - Um ambiente de apoio à realização de experimentos científicos para um cluster de computadores de alto desempenho

Abstract

Computer-supported simulation of scientific experiments has been generating an increasing amount of data. This demands the use of computer clusters as they have high processing power whereas lower costs than mainframes. However, clusters need to be shared among scientists as they are not present in every research laboratory. Nevertheless, scientists still face barriers to use clusters such as lack of knowledge to use them and the geographic distance between research laboratories and clusters. Scientific workflows and web services has been studied as feasible solutions to support the planning and execution of experiments in clusters. The activities of scientific workflows may involve web services and human activities. This work presents an environment, named ClusterFlow, to support the definition and execution of experiments in a cluster of computers. ClusterFlow aims at supporting scientists in the definition, execution, sharing and reuse of scientific workflows exported to the WS-BPEL language. The main contributions of this work are: (i) to present a documented design of an environment to support scientific experiments in clusters based on web services; (ii) to propose a support mechanism for the management of human tasks in scientific workflows; and, (iii) to offer a mechanism for sharing and reuse scientific workflows. The design of ClusterFlow was based on a prototype development and an example of application.