Multi-objective Optimization for a Commercial Datacenter in Paraguay

This work presents the implementation of a Resource Optimizer System for a commercial datacenter in Paraguay, which offers cloud services and uses Citrix Hypervisor (XenServer) as virtualization platform. To solve the VMP problem in a real environment, an automatic data collection module was implemented to periodically obtain the status of the datacenter resources. An especially design multi-objective memetic algorithm was also implemented to recommend improvements on the virtual machine placement, when needed. Three objective functions were considered: (1) power consumption minimization, (2) network traffic minimization and (3) economical revenue maximization. Finally, recommendations were applied in the datacenter and measurements of the used resources were made, minimizing power consumption at 29%, minimizing network traffic up to 50% while maintaining to maximum the economical revenue.