Optimization of Virtual Machines in Cloud Environment

The proposed VMP-LR (Virtual Machine Placement-Load Rebalancing) has three main components, namely, Resource Request Handling Component, Placement Component and Load Monitoring Component. VMP-LR is designed using a two-phase methodology, where the first phase handles the tasks involved with Resource Request Handling and Placement Components, while Phase II handles the tasks of Load Monitoring Component. The algorithms proposed in both the phases are designed to handle multiple resource requests. Dur ing non-rush hours (low traffic), as the number of requests is minimal, the VMP-LR uses a simple enhanced round-robin method to place virtual machines (VMs) to physical machines (PMs). During rush hours (heavy traffic), the three queues created are handled using three separate hybrid scheduling and load- balancing algorithms to perform placement operation efficiently and accommodate high resource demands. To solve this issue, the load monitoring component is used. For this purpose, a hybrid algorithm that combines ant colony optimiza tion (ACO) with an artificial bee colony (ABC) algorithm is used. The proposed algorithms are implemented using CloudSim Simulator and evaluated using seven performance metrics. They are throughput, response time, resource utilization rate, power usage, load unbalancing rate, SLA Violation rate and migra tion rate. The average SLA violation rate of VMP-LR is 0.6% less as compared with FF (1.26%), BF (1.1%), RR (1.36%), GA (0.98%), ACO (1.16%) and PSO (1.01%). This shows that VMP-LR is a much-improved version among existing load balancing algorithm.