On the evaluation of gridification effort and runtime aspects of JGRIM applications

Abstract

Nowadays, Grid Computing has been widely recognized as the next big thing in distributed software development. Grid technologies allow developers to implement massively distributed applications with enormous demands for resources such as processing power, data and network bandwidth. Despite the important benefits Grid Computing offers, contemporary approaches for Grid-enabling applications still force developers to invest much effort into manually providing code to discover and access Grid resources and services. Moreover, the outcome of this task is usually software that is polluted by Grid-aware code, as a result of which the maintainability suffers. In a previous article we presented JGRIM, a novel approach to easily gridify Java applications. In this paper, we report a detailed evaluation of JGRIM that was conducted by comparing it with Ibis and ProActive, two platforms for Grid development. Specifically, we used these three platforms for gridifying the k -Nearest Neighbor algorithm and an application for restoring panoramic images. The results show that JGRIM simplifies gridification without resigning performance for these applications.