The Research on MPC-WS, a Web Service for the Simulation of Metal Powder Compaction Process

摘要

Compaction is a nonlinear procedure commonly used to form powder material. Simulating the compaction process is often time-consuming. In this paper, a parallel computation system for the simulation of metal powder compaction process was established on a Linux cluster that consists of 32 IBM eServer xSeries machines connected over 1 Gb Ethernet. Test problem of the compaction of a gear was simulated. Density distribution as well as other useful results was obtained. Performance of the system was evaluated with different parallel schema used. It shows that four processors produce maximum time saving. Efficiency tends to decline if parallel scale exceeds the limit. Incoming network traffic of the master node was measured for two different cases where eight and ten processors are employed respectively. It shows that when the number of processors increases, more network traffic is created, which decreases the platform's efficiency. An MPC-WS that can choose approximate number of processors for user jobs and serve user requests concurrently was offered.