Extending Task Parallelism For Frequent Pattern Mining
| Title | Extending Task Parallelism For Frequent Pattern Mining |
| Publication Type | Conference Paper |
| Year of Publication | 2009 |
| Authors | Kambadur, P., A. Ghoting, A. Gupta, and A. Lumsdaine |
| Conference Name | Proceedings of the International Conference on Parallel Computing (ParCO) |
| Date Published | 09/2009 |
| Conference Location | Lyon, France |
| Keywords | OSL |
| Abstract | Frequent pattern mining, a popular approach in informatics, places unique requirements that are not met by the existing tools for parallelism. In particular, this approach is characterized by extremely large data sets and irregular memory access patterns. For efficient parallelization of these applications, it is necessary to support highly dynamic load balancing along with scheduling mechanisms that allow users to exploit data locality. Given these requirements, task parallelism is the most promising of the available parallel programming models. However, existing implementations of task parallelism do not allow users to specify custom scheduling policies that allow them to exploit data locality. In this paper we use frequent pattern mining as a test case to demonstrate the requirements for efficiently parallelizing some informatics applications. We also present PFunc, a novel task parallel library that offers features such as customized task scheduling and task affinity that satisfy some of these requirements. |
|
Copyright © 2012 The Trustees of Indiana University | Copyright Complaints