Research On Optimizing A 2-part Schedulability Analysis Method

Task schedulability analysis in real-time multiprocessor systems is an important topic in the field of real-time systems.Among all the methods,Response Time Analysis(RTA)is an important and promising technique for analyzing the schedulability of real-time tasks under both Global Fixed-Priority(G-FP)scheduling and Global Earliest Deadline First(GEDF)scheduling.Although widely adopted by many RTA-based works,the interference calculation technique is too conservative that most existing RTA methods using this technique for tasks under global scheduling are dominated by the schedulability analysis methods for tasks under partitioned scheduling.Recently,a new technique called 2Part is proposed that addresses this pessimism at the cost of efficiency.Although effective,this technique introduces a large amount of extra calculations and therefore has a bad performance in terms of efficiency.To this end,this paper first gives a case study exploring the key factors that affect the efficiency of 2Part and proposes an idea to solve the efficiency problem.Then by analyzing the relationship between the total interference suffered by the target task with its real execution time in the first phase,it is concluded that there are too many extra calculations in 2Part.By reducing these calculations,a novel RTA method ETP(Efficient Two-Part)is proposed,which improves the efficiency of 2Part to an acceptable degree without sacrificing the accuracy of the schedulability test.A set of experiments with different parameters is conducted to test the performance of the proposed execution framework ETP.The result shows that ETP can generally reduce the enumeration amount and the execution time of 2Part.ETP performs much better upon the real-time systems with less processors,where the efficiency of 2Part is improved by up to96%.It follows that the ETP method can effectively solve the efficiency problem of 2Part and make it applicable in real-time systems.