Online Non-preemptive Scheduling on Unrelated Machines with Rejections

When a computer system schedules jobs there is typically significant cost associated with preempting job during execution. This can be incurred from the expensive task of saving memory’s state or loading data into and out memory. Thus, it desirable to schedule non-preemptively avoid costs preemption. There need for non-preemptive schedulers desktops, servers, centers. Despite this need, gap between theory practice. Indeed, few online are known have strong theoretical guarantees. likely due lower bounds on any algorithm popular objectives. typical worst-case analysis approaches, even resource-augmented approaches such as speed augmentation, result in all algorithms having poor performance article considers scheduling problems rejection model where allowed reject small fraction jobs. By rejecting only jobs, shows that circumvented. approach used discover algorithmic policies Specifically, presents following three objectives: minimizing total flow-time, weighted flow-time plus energy convex function, under deadline constraints. The first two constant competitive ratio while For last problem, we present without rejection. Beyond specific results, asserts alternative models beyond augmentation should explored aid discovery good face requirement being non-preemptive.