From the design of first microprocessor to today, real-time embedded systems have been improving their performance, functionality, applicability in various ways. They have changed from cumbersome products to mobile portable ones with improved efficiency. These products' processing and storage capacity are far beyond their size, so they have to shrink every part of the artifact as well as battery also. Reducing the battery's size made it difficult to provide enough energy for processing. Therefore, minimizing the power consumption of a system is the only feasible way to reduce the size of the product. As a result, power scheduling has become one of the issues in the design of any embedded system. Multiple research reports describe scheduling algorithm minimizing the total energy of the system. However, very few of them describe scheduling techniques for which the energy consumption is optimal. Here optimal means the lowest energy that any system could consume in the same circumstances.;In power scheduling technique the major problem arise with deadline. When we make an optimal energy schedule we have to give a considerable though about deadline restriction. This work presents a novel solution of making a better schedule while considering the deadline of every task. The aim of the project was to produce a scheduling technique providing a minimum power schedule on both uni-processor platforms and multiprocessor platforms running non-preemptive jobs. We have used the conventional EDF and LLF scheduling algorithms to produce the initial schedule for our algorithm. We also developed two different algorithms to provide the optimal power schedule for two different situations. The only condition is that the input job set has to be EDF or LLF schedulable.;Our analysis shows a significant improvement in energy consumptions for both uniprocessor and multiprocessor platform. In our observation, multiprocessor platform is more convenient for the execution large number of tasks with greater accuracy. The percentage of energy minimization does not depend on a particular factor, so we were unable to establish the specific percentage of energy minimization value for any algorithms. For instance, for our algorithm, we could reduce the energy consumption by almost 1/3 of the maximum energy consumed by any schedule. Although a suitable number of processors was needed to archive that minimum energy.
展开▼
