Patent attributes
A parallel computer system automatically detects various troubles that may occur during computations using the parallel computer system so as to automatically cope with those troubles in such environment as in a design optimization using an evolutionary optimization in which a long time is required for one computation. A parallel computer system, includes a plurality of computing nodes for executing a calculation program and a master node connected to the computing nodes through networks, for performing a parallel computation process in an environment where a long time is required for one time computation as in a design optimization using an evolutionary algorithm. The system includes abnormality handling unit for automatically performing a series of processes of monitoring, periodically or on the basis of process unit, crash or hang-up of the calculation program, suspending the execution of the calculation program in the computing node in which abnormality has been detected and making another computing node execute the relevant calculation program.
