Abstract:

In order to achieve efficient resource usages and load balancing in the cloud computing system, we need to schedule the cloud computing system on the virtual machine granularity and to migrate the virtual machine from a high load physical node to a low load physical node by the live migration technique. Combining the load forecasting techniques and the virtual machine dynamic scheduling technique, we propose a load forecast scheduling algorithm, which can predict future load changes of the virtual machine via historical load data, and then schedule the virtual machine according to the  load forecast results. The proposed algorithm can effectively avoid the occurrence of high load physical nodes in the cloud computing system, realize load balancing, and improve resource utilization.

Key words: cloud computing, load forecast, dynamic scheduling, live migration