Skip to Main Content
The wide-area vector meta computing infrastructure named a vector computing cloud has been proposed as a next generation high-performance computing infrastructure. However, in the vector computing cloud, the difference in site policies between organizations causes inefficient usage of vector computing resources. To achieve fairness and efficient job scheduling on the vector computing cloud, this paper presents a history-based job scheduling mechanism for a queue system. The proposed mechanism estimates the time to start the job execution in a queue system from the history of job-execution on vector supercomputers. Based on the estimation, the job scheduling mechanism automatically allocates the job to an appropriate site, which can execute the job earlier. The simulation results show that the proposed job scheduling mechanism improves the utilization efficiency of vector computing resources, compared to the conventional round-robin scheduling mechanism. In addition, the experiment using a prototype of the vector computing cloud indicates that the proposed job scheduling mechanism has enough potential for transparently executing jobs between the two SX-9 systems.
Date of Conference: 19-23 July 2010