With the development of computing technology, the types of realtime applications increasing number of expanding the range of applications of realtime systems, the increasing complexity of the system, especially with the development of network technology to promote the network for realtime applications, such as network multimedia, distance learning, telesurgery etc., task completion of these applications have time constraints, but such time constraints neither as stringent as hard realtime, unlike the weak realtime as illdefined, but based on the quality of service requirements. In order to better cope with the task types of realtime systems, complex constraints with instantaneous overload characteristics, the weak hard realtime theory emerged. Weakly hard real theory as a specification, perfect to enrich the theory of realtime systems, unified description of the the original various types of realtime system, hard realtime and soft realtime are actually weakly hard realtime systems, a special case. Weakly hard realtime to meet the needs of the new characteristics of the realtime system. Because of its hard realtime, weak realtime unified, and therefore easier to deal with many types of task scheduling; weakly hard realtime quality of service requirements of the two parameters describing the task, in order to more clearly define and distinguish the quality of service of the task ; system overload by weakly hard realtime scheduling algorithms provide a slow degradation of the quality of service. In this paper, to further enrich the weakly hard realtime constraints specification and weakly hard realtime scheduling algorithm is carried out indepth research, productive work and innovative made in the following aspects: 1) This paper enriches the weak hard realtime constraints specification proposed (p, k) constraints, and prove the equivalence constraint (m, p), and thus can be exported relationship with the other weakly hard realtime constraints. (M, p) constraints, and (p, k) constraint of the different emphasis, the former projecting the number of consecutive lost deadline, which highlights the user to consider the minimum window. Further, the loss rate (p, k) is defined, and is given by (p, k) lost rate satisfies (p, k) constraints of the necessary conditions to provide a theoretical basis for classification selection algorithm. 2) In this paper, a class cutbased scheduling algorithm for solving variablelength window constraint violation discriminant cropped weak hard realtime scheduling (CutDown Based Scheduling constraints based on the (m, p), CDBS ) algorithm. Effective cut through the execution of the task sequence, and the introduction of the concept of the turning point, so that the constraint on the (m, p) satisfies the discriminant complexity is greatly reduced, and has nothing to do with the sequence length. The clipping algorithm correctness proof is given in the text, and its effectiveness is verified by experiment. In this paper, a simple priority allocation strategy, allocate tasks based on the the distance m consecutive losing deadline distance priority scheduling, and the combination of the four states of the task may appear. Finally, the algorithms and EDF, the DWCS, DBP algorithm compares the CDBS algorithm in terms of dynamic loss of efficiency and minimum success rate provides the appropriate compromise, has considerable performance compared with other algorithms. 3) This paper presents an arbitrary window constrained scheduling algorithm, assurance issues, research in services provide fair and carrier case different from the loss rate of the variablelength window. Design constraints based on the (p, k), an arbitrary window constraint scheduling (Any Window Constraint Scheduling, AWCS) algorithm, analyzing AWCS complexity of the algorithm, and in accordance with its sharp increase in complexity not suitable for scheduling rethrough load conditions proposed simplified algorithm K window constrained scheduling (KWindow Constraint Scheduling, KWCS) algorithm. The experiment showed that KWCS having the AWCS similar performance and complexity is substantially reduced, and therefore KWCS more suitable for the actual system applications. Fairness and differential analysis of AWCS (KWCS) provide further define the success rate deviates from, and given scheduling algorithm delay upper bound of General said method. Finally, the algorithm and other weak hard realtime scheduling algorithm compares AWCS (KWCS) The results show that rethrough overload situations better than the other algorithms, and can make the task of QoS degradation slowly, providing a fair but differentiated services . 4) the practical application of the K window constrained scheduling algorithm rich expansion. Proposed KWCS and DBP, hybrid algorithm, the system overload conditions is divided into mild overcontained critical overupload and respent upload, dynamic monitoring system of the state, and according to the different overloads take different scheduling policies, and thus solve the KWCS in mild before upload the problem of poor performance in the case; for KWCS algorithm needs to save the history of the state and not conducive to the expansion of issues, proposed classification selection algorithm, based on the (p, k) loss ratio, to be classified on the (p, k) flow, thereby improving the algorithm scalability; proposed multihop K the window constraint scheduling (Multihop KWCS, MKWCS) algorithm, to solve the KWCS application in an endtoend system. 5) weakly hard realtime scheduling algorithm attempts in the exploration of new application areas. Proposed multiprocessor weakly hard realtime scheduling algorithm to solve the multiprocessor, multiclass task scheduling, and consider the resources shared / exclusive access; proposed simple feedbackbased hybrid static / dynamic energy saving weakly hard realtime scheduling algorithm for the actual execution time of the task is usually far less than the actual situation of the worstcase execution time of mixed static / dynamic energy saving weakly hard realtime scheduling algorithm to improve the introduction of the division of tasks, estimated task execution time of the feedback mechanism, in order to obtain lower the speed of execution, to achieve better energysaving effect. The experiments show that, when the average execution time is less than the worstcase execution time is large, the new algorithm is superior to the original algorithm, energy savings up to 60% to 70%, minimum energy savings of about 10%. Inadequacies of the algorithm is that when the average execution time is close to the worstcase execution time, the new algorithm is more energyconsuming than the original algorithm. Finally, the general summary of the full text, and pointed out that the issue needs further study and perfect.
