Answers are grounded in this paper and cite their sources.
This work considers a semi-online version of scheduling on m identical machines, where the objective is to minimize the makespan. In the variant studied here, jobs are presented sorted by non-increasing sizes, and a buffer of size k is available for storing at most k jobs. Every arriving job has to be either placed into the buffer until its assignment, or else it has to be assigned immediately to a machine. We prove a lower bound greater than 1 on the competitive ratio of the problem for any m and any buffer size. To complement this negative result, we design a simple algorithm for any m whose competitive ratio tends to 1 as the buffer size grows. Using those results, we show the best possible competitive ratio is 1+Θ (m/k) 1 + Θ ( m k ) . We provide additional bounds for small values of m . In particular, we show that for $$m=2$$ m = 2 the case $$k=1$$ k = 1 is not different from the case without a buffer, while $$k=2$$ k = 2 admits an improved competitive ratio.
No takes yet. Share an insight, caveat, or question.
Epstein et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: