Processor Load Based Mutli-VM Scheduling System

Using virtualization technology, multiple virtual machines can run on a single server. When all virtual machines process data, there may be resources contention, and these virtual machines need more time to finish jobs. On the other hand, some of the servers are in idle status, so computing resources wouldn't be used effectively.Multi virtual machines scheduling system provides tow functions to solve these problems, including server load balance and combining virtual machines. Through monitoring each server and virtual machines we can conclude virtual machines'status, and than analysis if the computing resource can meet the tasks demand. We can change the distribution of virtual machines in the computing system, in order to archive server load balancing. So the computing performance of virtual machines improves, and it needs less time to finish jobs. Combining virtual machines uses minimum migration cost principle, making idle virtual machines running on concentrate servers. This can avoid resource fragment, and let a virtual machine can use maximum computing resource in a single server.Setting virtual machines VCPU number is 6, memory is 128 MB. Running 6 processes in virtual machine, each process computes PI 1,000,000 digits. After using server load balancing strategy, the speedup of finish all jobs is 2.869. In order to archive combining virtual machines, 5 virtual machines need to be migrated, and the time of these operations is 10.8s, the time is acceptable.