Reliability-Oriented Scheduling for Electricity–Computing Coordination in Data Centers with Adjustable Computing Workloads

Abstract

Data centers can provide demand-side flexibility by shifting adjustable computing workloads, but price-aware scheduling must preserve service completion and control delay loss. This paper proposes reliability-oriented electricity-computing coordination for data centers. Under a seven-day minute-level setting, an FCFS-like baseline is compared with a price-aware model that constructs start-time candidates for urgent and normal tasks, minimizes combined energy and delay cost, and refines the schedule through local search and rolling-window mixed-integer linear programming. A Cost-Adjusted Operational Reliability Index (CORI) combines workload completion with energy and delay-related service costs. Both models complete the same workload and consume 22975.21 kWh. The collaborative model reduces energy cost from 9555.65 RMB to 9228.17 RMB, delay cost from 4950.00 RMB to 614.44 RMB, and total task cost from 14505.65 RMB to 9842.61 RMB (32.15%). CORI increases from 100.00 to 147.38, indicating improved scheduling resilience and cost-adjusted operational reliability under the adopted deadline and penalty settings.