Load Balancing Algorithm in a Virtualized Cloud Environment

In this work, an approach to the design of load balancing algorithm that dynamically allocates and reallocates resources using the Shortest Time Path in order to reduce CPU response time was proposed. To exemplify the approach, a system is presented which dynamically monitors the state of the VM, their utilisation capacities and eliminates idle VM and those that contribute too little to CPU utilisation and reallocate processes to VM to leverage on their maximum capacity. The African adage that says 'too many cook spoils the soup' was utilised to eliminate non-contributing VM to CPU utilisation. The performance of this algorithm was assessed by setting up two data centers using VMware v2.0.2 and VMware ESX v50.05. WebSphere V6 was used as the application server. Three PMs each having three VMs and 2GB of memory. Each VM had one core with 512MB RAM. ESX server had three PMs with four VMs. Each PM had 4GB of memory and each one core assigned 1GB RAM. The experimental results reveal that it helped to slightly reduce the CPU response time when servicing and reallocating resources as evident in the dynamic nature of VM activation and de-activation.