Application of Open-Source Single-System Image Parallel Computing Platform for Many-Objective Tribological Design

The goal of this study is to develop a cluster computing scheme for tribological and engineering optimization using the Single System Image (SSI) system on a cluster of servers (computing nodes). A configured SSI system can automatically balance the computing load among the network-connected nodes, which can significantly simply the application of a cluster of computers. A recent published multiobjective optimization scheme (Group Inching Fortification, GIF) is to be modified to adapt the available many-core computing environment for an air foil bearing design. The parallel programming paradigm to be used is the popular OpenMP and the message passing among the computing nodes is handled by the SSI system. The distributed memory system is much different from the author’s recent studies which are mainly applying multicore/GPU processing in a shared memory system. In this study, the technique to manipulate a cluster computers with many processor cores is to be developed. This kind of system is scalable, which and can be further expanded to include more computing nodes. On the contrary, the computing power of a shared-memory system is basically limited to its initial setup. If the intended goal of this study can be acquired, the basic ability (manipulating high performance computing power) to perform big-data analyses and artificial intelligence study for complex problems can be obtained. The main objectives of this study are: (1) Modify the GIF method for an efficient cluster computing. The main bottleneck is the load balancing among the computing nodes for the proposed multiobjective optimization. (2) Find an effective way to obtain the critical number of computing cores in the numerical model for the system, which has a large number of processor cores. (3) Compare the effectiveness of using the hybrid (in the optimization-algorithm and iterative-solution levels) computation in a cluster of computers with a shared-memory multicore computing.

Project ID:PB10708-1778
External Project ID:MOST107-2221-E182-038