Research On Hybrid Task Scheduling Based On Users' QoS Requirements In Cloud Computing Environment

Cloud computing collect computing resources by the mature virtualization technology and these resources can constitute a shared and dynamically configure IT resource pool. Users who use the resources of the pool need to pay the corresponding cost according to usage. At present the task scheduling model and the optimization goal of the cloud computing is relatively single,at the same time task scheduling can't take the QoS (Quality of Service) requirements of tasks submitted by users into account in the process of scheduling, for this reason, in the cloud computing form the perspective of the users research of hybrid task scheduling based on the user's QoS requirements is of great importance.In this dissertation, according to the diversity of the task model, the hybrid task scheduling model is put forward, and the model not only can achieve multiple scheduling goals,but also can ensure the QoS requirements of the task when scheduling. The main work content is as follows:(1) Considering the diversity of tasks and the users can focus on the quality of services of the task in the task execution, a hybrid task scheduling model based on the user's QoS requirements is presented, and the model contains both the workflow and independent task.(2) Aimed at the problem that the task scheduling time in the process of scheduling is too long, and the service cost is too much, the improved particle swarm optimization is put forward. Based on the improvement of the particle swarm optimization, the improved particle swarm optimization is able to control the search step length when searching and can filter out high quality particles when each iteration, and ensures that it can juggle tasks scheduling time and task scheduling cost when the task scheduling.(3) The improved ant colony optimization is put forward which combine the basic ant colony optimization and the improved particle swarm optimization. It adjusts the combination of the stimulating factor of ant colony algorithm dynamically by improved particle swarm optimization, and improve the certainty when the ant choose the path and adjust the pheromone of trails to ensure that it can satisfy the user's QoS requirements when scheduling.(4) The simulation simulator CloudSim of the cloud environment is used to simulate the scheduling of FIFO (First Input First Output), Genetic Algorithm, two basic scheduling algorithm and two kinds of improved algorithm. Comparing the task execution time and execution cost of two scheduling algorithms before and after improvement, it proves that the improved algorithm can not only ensure the user's QoS requirements but also improved in terms of scheduling performance.