Optimizing User Oriented Job Scheduling within TORQUE

We presents major extension of the widely used TORQUE Resource Manager. Unlike common resource managers that use queuing approach, our solution uses planning (job schedule construction) and schedule optimization in order to achieve better predictability, performance and fairness with respect to common queue-based approaches. Moreover, additional important features are supported, e.g., so called multi resource fairness that is used to fairly prioritize users subject to their (highly) heterogeneous demands concerning various system resources. Also, new textual and graphical interfaces allow users to better control and plan computation of their jobs. Our solution is currently undergoing experimental deployment in Czech national Grid MetaCentrum.