Energy-Aware Massively Distributed Cloud Facilities: The DISCOVERY Initiative

International audienceInstead of the current trend consisting of building larger and larger data centers (DCs) in few strategic locations, the DISCOVERY initiative proposes to leverage any network point of presences (PoP, i.e., a small or medium-sized network center) available through the Internet. The key idea is to demonstrate a widely distributed Cloud platform that can better match the geographical dispersal of users and of renewable energy sources. This involves radical changes in the way resources are managed, but leveraging computing resources around the end-users will enable to deliver a new generation of highly efficient and sustainable Utility Computing (UC) platforms, thus providing a strong alternative to the actual Cloud model based on mega DCs (i.e., DCs composed of tens of thousands resources). This poster will present the DISCOVERY initiative efforts towards achieving energy-aware massively distributed cloud facilities. To satisfy the escalating demand for Cloud Computing (CC) resources while realizing economy of scale, the production of computing resources is concentrated in mega data centers (DCs) of ever-increasing size, where the number of physical resources that one DC can host is limited by the capacity of its energy supply and its cooling system. To meet these critical needs in terms of energy supply and cooling, the current trend is toward building DCs in regions with abundant and affordable electricity supplies or in regions close to the polar circle to leverage free cooling techniques [1]. However, concentrating Mega-DCs in only few attractive places implies different issues. First, a disaster in these areas would be dramatic for IT services the DCs host as the con-nectivity to CC resources would not be guaranteed. Second, in addition to jurisdiction concerns, hosting computing resources in a few locations leads to useless network overheads to reach each DC. Such overheads can prevent the adoption of the UC paradigm by several kinds of applications such as mobile computing or big data ones