Today, as a result of electronic business development and controlling facility requirements, embedded system importance and applications increased, as it became one of the most important research areas in computer science in recent years. In most of the times, embedded systems operations should be done before hitting a specific deadline, so almost all of embedded systems are real time. Military and industrial equipment, cellphone and other commercial applications like ATM and intelligent systems are samples of these systems. In addition to real time feature, low energy consumption is another main feature of embedded system, which is a main issue for digital system designers. Task scheduling and executing on present cores is the key problem in multi-core systems. In contrast to single core, that scheduling only defined as a matter of time, in multi-core systems the scheduling problem is a two dimensional issue, because the execution space comes into consideration in addition of time. It means that we should decide when and where a task should execute to achieve processing power, performance and execution time efficiency. Here we focus on four aspects of these types of systems: energy consumption, system efficiency, system performance, and system response time. Load distribution pattern on available resources (multi-core processors) affects all of the above aspects. Inefficient load distribution causes high energy consumption and decrease in performance and efficiency of the whole system. Most of proposed methods, dispatch tasks among processors regardless of the task type, and they only focus on dynamic voltage/frequency scaling mechanisms. The algorithm has been composed of three levels. At the first level a new method has been proposed to separate between periodic and aperiodic tasks with respect to the number of available cores. Second level divided to two sections. In the first section a new algorithm used to dispatch periodic tasks among respective cores. In the last section another new algorithm proposed for aperiodic tasks. At the third level a new algorithm has been developed to regulate voltage/frequency level according to deadline. Simulation results reveal that our proposed algorithm causes efficiency in energy consumption, in addition to increasing system performance and efficiency, and decrease in aperiodic tasks response time. Our algorithm is able to provide more quality in comparison to the other studied methods, due to satisfying more deadlines for periodic tasks and decreasing response time for aperiodic tasks in a reasonable order of execution time. Keywords : scheduling, real-time task, multi-core processors, embedded systems
