Research Of Task Scheduling And Results Recovery Strategy In Cloud Service

Cloud computing is the development of traditional computer technology such as parallel computing, distributed computing, grid computing and utility computing, which adapts to the rapid development of the network technology and rapid change of business model, and satisfying users’ growing network demand for the computing power and business service quality. It is a new model of commercial calculation.However, there are still many performance-related issues need to be improved in Cloud computing, such as task scheduling and results recovery strategies. These two problems are also two difficulties of cloud computing research. Cloud computing is a calculating model of internet based access to gain shared resource instantly. It has the characteristics of large extent, virtualization, high reliability, universality, high scalability, on-demand service, and cost-effective etc. On studying of existing task scheduling and results recovery algorithm, this paper finds there are such defects as long execution time, uneven load, and expensive costs etc. Aiming at these problems, this paper works out a new mode of calculating, the IGOA, which improve the accuracy of calculation and avoid the local optimumand and precocious phenoment. According to the defect that results of recovery for AI planning cannot subtasks to meet the requirements efficiently and accurately in the task library, this paper designs results recovery strategy which is based on user’s demand. According to the numbers sequence of decomposition, this strategy combines the subtask together reasonably through the task matching way, which not only improves the recovery efficiency, but also satisfies the users’ demand.At last, based on CloudSim, this paper conduct a serious of experiment, which simulates the cloud task in the length of task, the size of input and output file, the changing of bandwidth, while other conditions remain unchanged. By comparing the experimental status, we can see that the improved algorithm can effectively save the execution time, improve the load balance, improve the utilization rate of resources, and achieve the expected results. Nevertheless, the algorithm can only adapted to the simulation phase, which means the task parameters setting of the algorithm and dynamic recovery also need further improvement. Furthermore, the improvement of algorithm in cloud computing platform of the open source for the verification of practical application is our next task.