Dynamic, capability-driven scheduling of DAG-based real-time jobs in heterogeneous clusters
International Journal of High Performance Computing and Networking
| Hi-index | 0.00 |
The priority-based weighted fair queueing (PWFQ) schedule for real-time networks can be used to decouple session delay and its allocated bandwidth share. The idea is to introduce the concept of sliding window on the packet's virtual finish time. For packets whose finish time is within the window, the fixed priority assigned to each communication session is used to determine which packet should be sent out first. We apply the sliding window mechanism to aperiodic job scheduling with the total bandwidth server algorithm. When job arrivals are modeled by the leaky bucket model, we show that the algorithm can be used to improve the response time of aperiodic jobs.