A uniform parallel optimization method for knowledge discovery grid

Abstract

Grid is a new solution to computationally and data intensive computing problems. Since the distributed knowledge discovery process is both data and computational intensive, the Grid is a natural platform for deploying a high performance data mining service. In order to improve the performance of data mining applications, an effective method is task parallelization. Existing mechanisms of data mining parallelization are based on NOW or SMP, it is necessary to develop new parallel mechanism for grid feature. In this paper, we present a framework for high performance DDM applications in Computational Grid environments called Data Mining Grid, with the function for decomposing data mining application into subtasks and then combine those subtasks to form directed acyclic graph. This kind of parallel mechanism decomposes application according to the actual computation power of each node in dynamic Grid environment. © 2008 Springer-Verlag Berlin Heidelberg.