Research On Task Scheduling Strategy For Service Quality Based On Cloud Robot Platform

With the rapid development of cloud computing technology,cloud robot has become a key research direction in home service robots.Robot will send a variety of tasks to the cloud based on users' requests during task execution process.In the multi-machine coexistence environment,different robots will send different requests to the cloud at the same time.Therefore,how to effectively dispatch users' tasks is the core problem that needs to be solved in cloud robot systems.Based on environment characteristics in service robot laboratory of Shandong University,a small IaaS(Infrastructure as a Service)cloud management platform is built to realize the control and monitoring to the equipment.Then,task scheduling in cloud is studied deeply and corresponding scheduling strategies are designed for both batch scheduling and online scheduling respectively.The main works of this thesis are listed as follows.(1)CloudStack,a cloud management platform is set up.With the number of infrastructure increases gradually in laboratory environment,CloudStack is chosen as the basic framework for unified management and monitoring in small private cloud.CloudStack is relatively friendly,it can be installed easily and offers great convenience for secondary development.In this structure,one physical machine acts as a management node and the MySQL database is configured.NFS(Network File System)is also set up on management node as the main storage.Another machine serves as a KVM(Kemel-based Virtual Machine)host.The management node collects requests from users,with the help of MySQL database,and it can record all kinds of information about the system,monitor states of the computing node,storage and VM(Virtual Machine),help administrator and users understand the running situations of each part and deal with emergency conveniently.Computing node obtains relevant information from the management node and finally completes all the requests.Functional test shows that the system can run normally and provide both computing and storage resources to users.(2)The batch scheduling problem in cloud environment is studied and a corresponding strategy is proposed.A service-oriented static hybrid scheduling algorithm is put forward aimed at large user tasks.The strategy considers differences of tasks on diverse resource nodes,and weights are gotten through combination of the differences with defined priorities.Tasks are executed according to their weights and the greedy algorithm is integrated to avoid imbalance.Simulation results show that the proposed algorithm can deal with complex multitask in cloud environment,improve the performance and service quality of cloud robot system effectively.(3)The task scheduling algorithm under online scheduling mode is designed.Aiming at the uncertainties of the arriving time of user tasks,an improved dynamic ant colony algorithm for service quality in cloud environment is proposed.In the algorithm,a task's priority is measured by the combination of its arriving time and resource demands.During the process of allocation,inherent execution ability,current load situation of each node and some other factors are considered.Simulation results indicate that this new algorithm can not only adapt to dynamic cloud computing environment,but also realize the balance of load and the improvement of service quality.