Achieving the autonomic adaptation of resources in virtualized network environments

The infrastructure that supports the digital world is adopting network and system virtualization to raise its reliability and flexibility. However, the allocation of such virtual resources is still performed by human administrators, which limits the agility of such systems to respond to events and new requirements. In this paper we propose an architecture that exploits the capabilities offered by virtualization techniques to achieve the prompt adaptation of network and computing resources to the needs imposed by their changing environments by automating control and management operations in an autonomic manner. It collects observations from any of element of the environment, including resources and sensors, and characterizes them as semantic observations. Such observations are then analyzed to determine if any corrective or perfective change is required to overcome a possible problem. The whole process is automated but guided by control statements provided by administrators, so the actions it can take have totally predictive results. As it provides a closed control loop, the expected effects of such actions are continuously checked. Finally, we discuss the benefits of this system in an emergency scenario and demonstrate its feasibility by presenting the results obtained from a proof-of-concept implementation of the proposed approach.