Research And Implementation Of Real-Time Energy-saving Scheduling Technology Based On Improved OLDVS Algorithm In Linux Kernel

In recent years, energy consumption is more and more taken seriously in the field of real-time systems, the use of dynamic voltage scaling (DVS) technology has become the primary means to adjust the system energy consumption of processor, and gradually applied to the design of real-time system. OLDVS-AQ (Online Dynamic Voltage Scaling Assisted Queue) is a partitioned scheduling algorithm which makes use of DVS technique to schedule real-time system of saving energy. It has achieved the target that reduces system energy consumption meeting the real-time requirement of changing voltage or frequency reasonably, which still has good real-time performance and high energy saving, so it has become a real hot issue in the field of energy conservation scheduling.Currently, energy-saving scheduling algorithm is a lack of implementation in real operating systems. All kinds of energy consumption algorithms are based on theoretical analysis and lack the real data, which limits the development of its practical application in the field of real-time systems. Therefore establishing a real-time operating system experimental platform that can verify actual energy consumption in a variety of energy-efficient scheduling algorithm is to be today’s urgent needs. LITMUSRT (Linux Testbed for Multiprocessor Scheduling in Real-Time Systems) is currently popular real-time scheduling experiment platform, but it only supports limited real-time scheduling algorithm and cannot meet this demand.In this paper, a kernel scheduler is developed which supports OLDVS-AQ real-time scheduling strategy and is based on the version of Linux kernel 3.0.0 with integrated Litmus 2012-2 patch. The four key technologies have been studied and developed, which include processes preemption and the runtime states transition of real-time task routines, the dynamic CPU frequence adjustment to the process that will be scheduled and the efficient dynamic priority queues management within the full life of the real-time tasks process. The OLDVS-AQ algorithm model is modified to fit the real adjustment mode of CPU processor and the auxiliary queue is designed to calculate the CPU frequency of the process that will be scheduled. Meanwhile an efficient real-time application is also developed to execute on this OLDVS-AQ scheduler and a fixed frequency of EDF scheduler in order to make kernel be able to record the execution and scheduling information of each real-time task process. A large number of test results and compared real-time scheduling experiments show that this real-time scheduler is correct, effective and energy saving to schedule real-time task. This work will be more conducive to real-time scheduling algorithm for energy saving analysis and comparison, and provides a useful technical reserves and abundant experimental data support for carrying out a more in-depth application of energy-efficient scheduling strategy with the method of dynamic voltage scaling. It has played a catalytic role in promoting energy-efficient scheduling theoretical research to practical system application.