Research Of Task Scheduling For Cloud Computing Based On Dynamic Priority

In Cloud computing, multiple resources which distributed in datacenter are permanently stored in a shared resource pool through virtualization technology. Users can obtain corresponding services according to their requirement and just need to pay for the actual consumption of resource. At present, the objective of almost task scheduling algorithms in cloud computing is single, and they are difficult to meet various needs. Therefore, how to take both users and cloud service providers into account and meet various needs become the key problem to be solved in the cloud environment.Through analyzing the characteristics of the real-time tasks with deadlines, we establish a cloud computing task scheduling model based on dynamic priority, this paper emphatically has done the following works:(1) The article describes the definition, service types, architecture and features of the cloud computing.(2) The scheduling objectives are proposed from two perspectives. We should minimize the completion time as well as ensure the urgent tasks being finished before their deadline. In addition, the computing resources in cloud computing should remain evenly balanced to maximize the resource utilization and increase the profit of cloud service providers.(3) Generally, some existing scheduling algorithm which based on priority only consider unilateral features. To meet various scheduling objectives, we take into consideration both task value and time urgent of tasks and propose a task scheduling strategy based on dynamic priority. In this case, tasks will be scheduled according to priority order and be executed at right time.(4) The ACO-LB (Load balancing optimization algorithm based on ant colony algorithm) is proposed based on basic ant colony algorithm combined with the roulette algorithm. The pheromone of virtual machines can be updated in real time through optimizing the pheromone updating rule. In addition, some excellent allocation schemes will be retained through the collaboration of the ants. Also, a load balancing adjustment factor in applied in ACO-LB scheduling algorithm to update to the heuristic factor.(5) The Preemptive scheduling algorithm based on Dynamic Priority(DPP) is proposed to avoid high-priority tasks being in long waiting period in non-preemptive task scheduling scheme. This scheme ensures priorities to execute high-priority tasks to avoid these tasks missing their deadlines. In addition, we analyse the relationship between the slack time of the high-priority tasks and the remaining execution time of low-priority tasks, corresponding scheduling strategies are given according to the different condition to reduce the times of unnecessary preemption as far as possible and improve the success completion rate.(6) The open source cloud computing simulation tool CloudSim is used to be the experimental platform to simulate the improved ACO-LB and DPP. The simulation results show that the improved ACO-LB algorithm not only ensures the task be completed as soon as possible but also enables the virtual machines to maintain in a relatively balanced state. In addition, the DPP algorithm ensures urgent tasks complete their executions in time and increase the revenue of cloud service provider.