Data partitioning for networked parallel processing

Author

Crandall, Phyllis E. ; Quinn, Michael J.

Author_Institution

Dept. of Comput. Sci., Oregon State Univ., Corvallis, OR, USA

fYear

1993

fDate

1-4 Dec 1993

Firstpage

376

Lastpage

379

Abstract

The workstation model of parallel processing presents specific challenges caused by the latency of the communications network and the workload imbalance that arises from the heterogeneity of the nodes. Data partitioning is critically important for parallel processing in this environment. We mathematically characterize the communication costs for four data decomposition schemes: scatter, contiguous point, contiguous row, and block. These methods are analyzed in terms of problem size, number of processors, network speed, and communication pattern. Bounds are established for the performance of these decomposition schemes that can be used to make better-informed data partitioning decisions

Keywords

communication complexity; distributed memory systems; local area networks; parallel processing; performance evaluation; resource allocation; communication costs; communication pattern; communications network; data partitioning; latency; network speed; networked parallel processing; problem size; workload imbalance; workstation model; Communication networks; Computer architecture; Computer science; Costs; Delay; Ethernet networks; Parallel processing; Pattern analysis; Scattering; Workstations;

fLanguage

English

Publisher

ieee

Conference_Titel

Parallel and Distributed Processing, 1993. Proceedings of the Fifth IEEE Symposium on

Conference_Location

Dallas, TX

Print_ISBN

0-8186-4222-X

Type

conf

DOI

10.1109/SPDP.1993.395508

Filename

395508