Battery-aware dynamic voltage scaling in multiprocessor embedded system

In a battery powered system, a primary design consideration is the battery lifetime. The profile of current drawn from a battery determines its lifetime. Recently, dynamic voltage scaling has been applied to alter the battery load current profile in distributed systems to reduce battery charge consumption (Chowdhury, P. and Chakrabarti, C., IEEE Workshop on Sig. Process. Systems, SIPS ´02, p.201-6, 2002). The load current profile is changed by utilizing the slack in the execution of scheduled tasks. We propose a new dynamic voltage scaling procedure that alters the load current profile by considering the total battery current instead of the method of Chowdhury and Chakrabarti that considers the current dawn by an individual task with the latest finish times in the schedule. The task schedule is partitioned into steps and the load currents during selected steps are targeted for reduction by scaling the supply voltage of the processing elements. Experimental results on a large set of task graphs show that battery charge consumption reductions of up to 89.80% are achieved by the new algorithm.