Title:Scheduling on a machine with varying speed: Minimizing cost and energy via dual schedules

Abstract:We study two types of problems related with scheduling on a machine of varying speed. In a static model, the speed function is (through an oracle) part of the input and we ask for a cost-efficient scheduling solution. In a dynamic model, deciding upon the speed is part of the scheduling problem and we are interested in the tradeoff between scheduling cost and speed-scaling cost, that is energy consumption. Such problems are relevant in production planning, project management, and in power-management of modern microprocessors.
We consider scheduling to minimize the total weighted completion time. As our main result, we present a PTAS for the static and the dynamic problem of scheduling on a single machine of varying speed. As a key to our results, we re-interprete our problem within the folkloric two-dimensional Gantt chart: instead of the standard approach of scheduling in the time-dimension, we construct scheduling solutions in the weight-dimension. We also give complexity results, more efficient algorithms for special cases, and a simple (2+epsilon)-approximation for preemptive dynamic speed-scaling with release dates. Our results also apply to the closely related problem of scheduling to minimize generalized global cost functions.