Research On Task Scheduling Optimization Algorithm In Cloudlets

The development of cloud computing in recent years provides good technical sup-port for users to perform resource-intensive tasks quickly.However,many emerging applications require not only computing resources but also extremely low latency.Un-der the limitation of the WAN,the centralized cloud data center structure will lead to high service delays,especially for real-time and high-bandwidth services,resulting in poor user experience while iterating with cloud data center.The proposal of Cloudlets solves this problem.Cloudlets is a small cloud data center composed of a group of servers,and distributed deployed at the edge of the network close to the user.Then the user can access the service in cloudlets with high bandwidth and low delay,which greatly reduces service delay.Cloudlts theoretically can well satisfy the interactive experience of many emerging applications.However users may be far away from cloudlets due to the mobility of mobile users,which leads to a poor network connectivity,thus,user experience will be poor.While a user moves across multiple cloudlets areas,issues of service scheduling between cloudlets to better support user experience become important.On the other hand,due to the distributed deployment of cloudlets,It will easily lead to lots of user requests,that is,greater load.However,too many tasks will result in lower response time of user request,which is difficult to accept for many applications.Therefore,it is worth investigating that scheduling tasks in cloudlet to make full use of resouces in cloudlets and reduce the response time of task.This dissertation makes a detailed study and analysis of the task scheduling un-der the scenes of mobile users and heterogeneous loads.A Mobility Prediction-based Markov Decision Process algorithm is proposed for task scheduling in user mobility scenarios.The algorithm aims at optimizing the average response time of the request task in the whole process of the user's movement.Based on the result of the user's mobility prediction,this algorithm optimizes this goal through the Markov decision process model.Comparison experiments with other algorithms show that the algorithm can ensure the user experience by effectively reducing the average response time of the request task.In the end,a task scheduling-based two-stage scheduling optimization algorithm is proposed for task scheduling under the uneven load of cloudlets.The algorithm aims to minimize the maximum task response time in the cloudlets system,and approaches the optimum through a two-stage process.The simulation results show that compared with other algorithms,this algorithm can achieve smaller maximum task response time.