Energy minimization techniques for real-time scheduling on multiprocessor platforms

The scheduling of systems of periodic tasks upon multiprocessor systems with processors of di ering speeds is considered. Computation speeds may di er because the system is comprised of di erent types of processors | i.e. it may be a uniformmultiprocessor platform. Alternatively, the platform may be comprised of identical processors with speeds that di er because some processors may have their voltage levels reduced in order to minimize energy consumption. Given a periodic task system, a condition is derived for minimizing the computing capacity of the multiprocessor system while ensuring all deadlines of the task system are met when scheduled using the earliest deadline rst (EDF) algorithm. A method of selecting the voltages of identical processors in order to meet this condition while minimizing the energy consumption is then presented. The resulting system is a platform of identical processors with di erent computation rates on which the task set can be scheduled using EDF.