Search for a command to run...
Using smart grids to build low carbon networks is one of the most challenging topics in ICT (Information and Communication Technologies) industry. One of the first worldwide initiatives is the GreenStar Network, completely powered by renewable energy sources such as solar, wind and hydroelectricity across Canada. Smart grid techniques are deployed to migrate data centers among network nodes according to energy source availabilities, thus CO <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sub> emissions are reduced to minimal. Such flexibility requires a scalable resource management support, which is achieved by virtualization technique. It enables the sharing, aggregation, and dynamic configuration of a large variety of resources. A key challenge in developing such a virtualized management is an efficient resource description and discovery framework, due to a large number of elements and the diversity of architectures and protocols. In addition, dynamic characteristics and different resource description methods must be addressed. In this paper, we present an ontology-based resource description framework, developed particularly for ICT energy management purpose, where the focus is on energy-related semantic of resources and their properties. We propose then a scalable resource discovery method in large and dynamic collections of ICT resources, based on semantics similarity inside a federated index using a Bayesian belief network. The proposed framework allows users to identify the cleanest resource deployments in order to achieve a given task, taking into account the energy source availabilities. Experimental results are shown to compare the proposed framework with a traditional one in terms of GHG emission reductions.