PARALLEL COMPUTER SYSTEM, PARALLEL COMPUTING METHOD, AND PROGRAM FOR PARALLEL COMPUTER

摘要

PPROBLEM TO BE SOLVED: To provide a parallel computer capable of automatically detecting troubles caused during computation using the parallel computer, and automatically handle the troubles, in an environment requiring a long time for a single computation, such as design optimization using an evolutionary algorithm. PSOLUTION: A parallel computer system includes a plurality of computing nodes for executing a computing program and a master node connected to the computing nodes through a network, and executes parallel computation processing in an environment requiring a long time for a single computation, such as design optimization using an evolutionary algorithm. The system further includes a trouble handling means for automatically performing a series of processings for monitoring a crash or hang-up of the computing program periodically or for every processing unit, aborting execution of the computation program in the node with a trouble detected therein, and having the computing program executed by another one of the computing nodes. PCOPYRIGHT: (C)2008,JPO&INPIT