Allocating jobs of different priorities to a distributed system with heterogeneous resources

摘要

The allocation of heterogeneous computational resources to arriving jobs of different priorities is a challenging task when both performance and workload balancing are considered. This paper studies the effect of job migration on the performance of a distributing computing system when allocating computational resources of different computational capabilities to arriving jobs of different priorities defined by users. The system that is examined is modelled according to queueing theory and its performance is evaluated via simulation. The main objective is to utilize efficiently the available computational resources while maintaining low response times by taking into account the types of submitted jobs. Simulation results showed that job migrations improved the performance metrics of the high priority jobs at the expense of a slight increase in overall mean response time.