Research On Real-time Elastic Task Scheduling Algorithm Based On Feedback Mechanism

Real-time system often refers to the system which can provide the required levelof service in the specified time. It needs an effective response to the external changesand a wide range of flexibility, which can timely response to variable system loads.One of the key contributors to decide the system performance and successfulcompletion of real-time tasks is the scheduling policies, so to design a properscheduling policy is the key problem in real-time systems.Elastic scheduling is one effective policy to solve the scheduling problem forflexible real-time systems. It adjusts the period of real-time tasks by the periodadjustment algorithm, and adjusts system loads to adapt to flexibility requirement ofreal-time systems. This paper focuses on the problem of elastic scheduling. In order toensure the scheduling success ratio and rational use of system resources, based on theanalysis of the existing real-time elastic task scheduling algorithms, we improve thebasic elastic scheduling algorithm where task deadlines are equal to task periods, andthe general elastic scheduling algorithm where tasks deadlines are less than periods.My contributions are as follows:Because of the fact that elastic scheduling mainly research on the soft real-timetasks, usually can not handle hard real-time tasks and result in system performancemutation, we presents an improved basic elastic scheduling algorithm by hardreal-time tasks with resource reservation, called basic period adjustment algorithmbased on resource reservation, to accommodate the coexistence of hard and softreal-time tasks. Based on the original performance functions, this paper gives a newfunction of the changes in resource utilization to adjust task periods, and to achievethe flexibility of variable loads and the lower performance changes.Because of elastic scheduling assumed the task execution time is pre-determined,resulting in a lower success rate, this paper presents a scheduling method for real-timeelastic tasks based on feedback mechanism. Using the dynamic feedback mechanismto estimate the task execution time, this method improved the basic elastic schedulingalgorithm and generalized real-time elastic task scheduling algorithm. By reflectingresource utilization of real-time tasks, the improved basic scheduling algorithmavoids the pessimistic estimates of system resources, increases the utilization ofsystem resources and scheduling success rate. On the other hand, the improvedgeneralized scheduling algorithm relaxes the conditions of the general elastic scheduling algorithm, and improves the success rate of scheduling algorithm.In order to estimate the effectiveness of the algorithms, this paper presents thesimulation comparison experiments between the proposed algorithms and the classicalalgorithms, by randomly generating real-time elastic task sets. The result reveals thatthe improved basic real-time elastic task scheduling algorithm has strong adaptability,while improving the success rate of task scheduling algorithm and system throughput.The improved generalized real-time elastic task scheduling algorithm greatlyimproves the success rate, but little effect on system throughput.