Hypergraph partitioning for multiple communication cost metrics: model and methods
Deveci, Mehmet and Kaya, Kamer and Uçar, Bora and Çatalyürek, Ümit V. (2015) Hypergraph partitioning for multiple communication cost metrics: model and methods. Journal of Parallel and Distributed Computing, 77 . pp. 69-83. ISSN 0743-7315 (Print) 1096-0848 (Onilne)
Official URL: http://dx.doi.org/10.1016/j.jpdc.2014.12.002
We investigate hypergraph partitioning-based methods for efficient parallelization of communicating tasks. A good partitioning method should divide the load among the processors as evenly as possible and minimize the inter-processor communication overhead. The total communication volume is the most popular communication overhead metric which is reduced by the existing state-of-the-art hypergraph partitioners. However, other metrics such as the total number of messages, the maximum amount of data transferred by a processor, or a combination of them are equally, if not more, important. Existing hypergraph-based solutions use a two phase approach to minimize such metrics where in each phase, they minimize a different metric, sometimes at the expense of others. We propose a one-phase approach where all the communication cost metrics can be effectively minimized in a multi-objective setting and reductions can be achieved for all metrics together. For an accurate modeling of the maximum volume and the number of messages sent and received by a processor, we propose the use of directed hypergraphs. The directions on hyperedges necessitate revisiting the standard partitioning heuristics. We do so and propose a multi-objective, multi-level hypergraph partitioner called UMPa. The partitioner takes various prioritized communication metrics into account, and optimizes all of them together in the same phase. Compared to the state-of-the-art methods which only minimize the total communication volume, we show on a large number of problem instances that UMPa produces better partitions in terms of several communication metrics.
Repository Staff Only: item control page