An Automated Performance Tuning Scheduling Framework for Computational Jobs in Desktop Grid
Tuning the performance of applications is a well studied field for parallel systems where the underlying architecture is known along with the interconnection pattern. The challenge of developers lies in effective utilization of application characteristics on specific architecture that leads to efficient deployment. But this process is highly manual demanding expertise to identify performance bottleneck, identify the cause from performance data by correlating run time behavior with program characteristics. Unfortunately few efforts have focused on tuning the performance of desktop grid scheduling at run time.