Load Prediction-Based Automatic Scaling Cloud Computing

It is vital to cloud computing that the elastic control technology can dynamically adjust resources on demand by changing the size or number of virtual machines. It will improve the utilization and realize the cost savings. However, there are still some shortcomings for elastically scaling technology. On the one hand, the elastic scaling of resources will take some time, so changing resource requirements can not be responded timely. On the other hand, whenever the resources do not meet the demand, it can not properly allocate resources based on the size of demand. Therefore, this paper proposes an algorithm about automatic scaling of resources based on load prediction. By using the algorithm combining linear regression and the improved Knuth-Morris-Pratt match to predict the next moment load, complete automatic extension before the changes in resource requirements and reduce resource adjustment time. In addition, according to the results, integer programming algorithm is used for solving specific resource expansion. The experiment results indicated that the proposed method can increase resource utilization and reduce cost of cloud computing resources while meeting the changing demand.