Autonomic computing to manage green core networks with quality of service

In a context where data and computing services are moving to external specialized datacenters the manual management of these systems is becoming an issue. Human administrators have to deal with hardware resources optimization while meeting the users' needs. In our approach we propose to reconfigure both a set of applications deployed in a datacenter (by adapting their behaviors using autonomic computing) and the wired network (by switching on and off its equipments like routers, modules). We take into account both the energetic costs with the network equipments and the quality of service provided to the end user by the deployed applications. The main contribution of the proposed model is to consider a compromise between the total power consumption of the network equipments and the application quality of service. We validated our approach by simulating deployed applications on the Grid'Mip infrastructure similar to a small core network made up of Cisco routers (part of Grid'5000 project).