Title :
Development of an MPI for Power System Distributed Parallel Computing in Wide Area Network with P2P Technology
Author :
Huang, Qi ; Yi, Jianbo ; Jing, Shi ; Huang, Ke
Author_Institution :
Sch. of Autom. Eng., Univ. of Electron. Sci. & Technol. of China, Sichuan
Abstract :
The analysis of modern power system becomes more computation intensive, and has to be performed in a dispersed IT infrastructure due to the inherently distributed topology of power system. Grid computing is potentially an ideal solution for electric power system and power network infrastructure also provides potential application field for grid computing because grid computing is most efficient in a large organization. This paper presents a new version of MPI (message passing interface) based on P2P, WD-MPI, in order to exploit the maximum performance in a heterogeneous environment for power system distributed computing. The problems of traditional MPI and some versions of modified MPI to improve its performance on heterogeneous and wide area network are analyzed. The system structure is presented and the design of the system, including asynchronous iterative algorithm, task and resource partition, is presented. The power flow calculation problem of an IEEE 300 bus system is used to compare the performances of the proposed scheme and previous solutions. The test is performed in a grid environment built in campus, which can simulate the behavior of wide area network to some extent. The test results show that the P2P based WD-MPI can give better performance than that of MPICH, MPICH-G2 and Globus solution
Keywords :
application program interfaces; message passing; parallel processing; peer-to-peer computing; power system analysis computing; wide area networks; IEEE 300 bus system; IT infrastructure; MPI development; asynchronous iterative algorithm; electric power system analysis; electric power system distributed parallel computing; grid computing; heterogeneous environment; message passing interface; peer-to-peer technology; power flow calculation problem; power network infrastructure; resource partitioning; wide area network; Distributed computing; Grid computing; Message passing; Network topology; Parallel processing; Performance analysis; Power system analysis computing; Power systems; Testing; Wide area networks; (MPI); (P2P); Distributed parallel computing.; Electric power system; Grid computing; Message Passing Interface; Peer-to-peer;
Conference_Titel :
Parallel Computing in Electrical Engineering, 2006. PAR ELEC 2006. International Symposium on
Conference_Location :
Bialystok
Print_ISBN :
0-7695-2554-7
DOI :
10.1109/PARELEC.2006.30