A Hybrid Meta-heuristic-based Linear Regression Algorithm for Live Virtual Machine Migration in Cloud Datacenters

Cloud computing provides services to its subscribers. Since there are different kinds of users that request computing resources for their applications, their resource usages obey from variable usage patterns. To manage cloud datacenters regarding to both users and providers’ viewpoints, a new system architecture for cloud datacenters is presented. This architecture embeds different modules. The core is a broker module which records resource requests for predetermined time-intervals. Then, the whale optimization algorithm (WOA) is combined with linear regression (WOALR) model that is a machine learning (ML) prediction tool to robust forecasting resource requests for short-term near future. Once the prediction model recognizes the system is subject to over-loaded or under-loaded circumstances, the live virtual machine (VM) migration approach is triggered to transfer appropriate VMs for meeting the needs. To assess the proposed prediction model, the linear regression model is hybridized with the most successful meta-heuristic algorithms. The proposed WOALR outperforms against other hybrid approaches in terms of service level agreement level (SLA) violation rates, total power usage, precluding number of additional VM migrations because of less squared error of proposed hybrid prediction model.