EFFECT OF FREQUENCY SCALING ON POWER CONSUMPTION IN EMBEDDED SYSTEMS

Optimization of power consumption in a real time embedded systems is a critical design parameter. As most of low power embedded systems are battery operated and have very limited source of power. Therefore the optimization of power utilization is discussed in this paper establishing its relationship with execution frequency of core processing unit in multitasking environment. The number of processes required to be executed in multitasking embedded systems may vary with time for an application, and hence the dynamic power consumed also vary accordingly. Each process in real time are defined with respective absolute deadlines and their relative deadlines depend on the number of processes schedules at a time. The relative deadline parameter can be used to estimate the required operating frequency of processing unit and can be scaled accordingly to optimize the power consumption. The relative deadlines needs to be less than the absolute deadline for any process to complete the task successfully. Therefore the operating frequency can be dynamically scaled such that the relative deadlines are just within absolute deadlines. The hard-real time systems are designed to meet the deadline of all processes under maximum process load, hence the operating frequency of the processing unit is chosen much higher to complete make sure all the processes within deadline under peak load condition. However, scaling down the frequency under other conditions will improve the thermal stability and also optimize the power consumption. The analysis of frequency requirement is done for two most widely used scheduling methods, Earliest Deadline First (EDF) and Rate Monotonic (RM) scheduling algorithms for comparison.