The Design And Implementation Of The Virtual Machine Scheduling Algorithm Over Multiple Datacenters In The Cloud System

The cloud service providers usually build multiple datacenters geographically distributed from all over the world to serve users from different place or guarantee the quality of the service. Generally, the cloud system will let users make the decision where to deploy the virtual machines. This method has two disadvantages: first, the uses usually cannot make the best decision lacking experience and information about the users of the virtual machines; second, for the cloud system, the resource utilization and the degree of load balance may be harmed. The advantage that multiple datacenters can cooperate to deploy the virtual machines requested by the users and guarantee the quality of these virtual machines and make better use of the resource in the cloud system is rarely researched and applied.To take the advantage of the cooperation of the multiple datacenters providing better service and better resource utilization, we propose one virtual machine scheduling system based on the agent model and one deploying method based on combinatorial auction and one two phase virtual machine adjusting algorithm. This scheduling system will deploy virtual machines requested by the users and adjust the position and size of these virtual machines dynamically to accommodate the load changes of the applications of the uses and guarantee the resource utilization and load balance in the total cloud system. The effectiveness of the scheduling system is verified by the software simulation.