Efficiency-Loss of Greedy Schedules in Non-Preemptive Processing of Jobs with Decaying Value

TL;DR

This paper analyzes the efficiency loss of greedy scheduling algorithms for non-preemptive jobs with decaying value, providing performance bounds and demonstrating near-optimal performance in practical scenarios.

Contribution

It establishes provable performance bounds for greedy schedules in decaying-value job processing and shows their effectiveness in realistic applications.

Findings

Greedy schedule achieves within a factor of 2 + E[max_j sigma_j](min_j E[sigma_j]) of optimal.

Performance bound improves to a factor of 2 for identically distributed service times.

Greedy schedules perform close to optimal in practical situations despite ignoring long-term reward decay.

Abstract

We consider the problem of dynamically scheduling J jobs on N processors for non-preemptive execution where the value of each job (or the reward garnered upon completion) decays over time. All jobs are initially available in a buffer and the distribution of their service times are known. When a processor becomes available, one must determine which free job to schedule so as to maximize the total expected reward accrued for the completion of all jobs. Such problems arise in diverse application areas, e.g. scheduling of patients for medical procedures, supply chains of perishable goods, packet scheduling for delay-sensitive communication network traffic, etc. Computation of optimal schedules is generally intractable, while online low-complexity schedules are often essential in practice. It is shown that the simple greedy/myopic schedule provably achieves performance within a factor 2 + E[max_j sigma_j](min_j E[sigma_j]) from optimal. This bound can be improved to a factor of 2 when the service times are identically distributed. Various aspects of the greedy schedule are examined and it is demonstrated to perform quite close to optimal in some practical situations despite the fact that it ignores reward-decay deeper in time.