Researching Scheduling Algorithm Of Embedded Linux Operating System

Embedded operating system is applied in many areas of the Internet. Linux system is not a perfect real-time operating system. In order to improving the processing power, there are many reforms of the real-time performance in Linux. The methods are the reform of the kernel such as the dual-kernel and preemption kernel approach, and the refinement of the clock granularity. Another method is the kernel's scheduling strategy. The real-time of the system depends on the process scheduling algorithm. We try to improve the real-time of the kernel in Linux by promoting the scheduling algorithm.The scheduling algorithm of the kernel in Linux2.6 is discussed in the paper. And we analyze the principle and the program code of the scheduling algorithm O(1). In this basic, we summarize the improvement of the real-time performance of the kernel in Linux2.6. We choose the scheduling algorithm based on the priority and discuss the problem. EDF scheduling algorithm is one of the widely used dynamic priority scheduling algorithm. The algorithm allocates the priority of the various processes in the ready queue according to their deadlines. But it can not solve the overloaded problem. The overloaded problem makes CPU spend too much time in the scheduling. Hence the performance degenerates quickly. To solve the above-mentioned problem, the study is based on the ISM (Isolation Server Method) in my paper. SLAD (SLACK Donation) algorithm and BACKSLASH (BACK SLACK Donation) algorithm are proposed by Lin and Brandt recently. They are improved algorithms based on the ISM. EDF scheduling is mainly to improve on the slack time. The algorithms reduce the DMR (deadline miss ratio) by making good use of the slack time. The algorithms lower the"domino"effect between the tasks and reduce the DMR without the effect on the overrun of the tasks. The algorithms can solve the overloaded problem. But they are not better than the EDF scheduling in the normal scheduling problem. A new dynamic priority scheduling algorithm is proposed. Under considering the second-time cache of the processor, a scheduling count link is designed and used to record the number of the occurrences by the cache. The tasks occurring frequently are recorded and others are removed. The method may solve the overloaded and normal scheduling problem and its feasibility is understood by the experiments.