Fair Resource Sharing for Dynamic Scheduling of Workflows on Heterogeneous Systems

This chapter presents a review of offline and online concurrent workflow scheduling. It compares five algorithms for online scheduling of a heterogeneous system in terms of turnaround time ratio (TTR), percentage of wins, and normalized turnaround time (NTT) when the goal was to maximize the user quality of service (QoS) defined by the completion time of the individual submitted jobs. The five algorithms are fairness dynamic workflow scheduling (FDWS), online workflow management (OWM), rank hybrid (Rank_Hybd), online min-min, and online max-min, which can all handle multiple workflow scheduling in dynamic situations. Based on the experiments, FDWS leads to better performance in terms of TTR, win(%), and NTT, showing better QoS characteristics for a range of time intervals from 10 to 90. For the time interval of zero, which is equivalent to offline scheduling, Rank_Hybd also performed well, but the schedules produced by FDWS had better QoS characteristics.