Exact and Heuristic Resource Mapping Algorithms for Distributed and Hybrid Clouds

This paper addresses the problem of cloud and networking resources mapping in distributed and hybrid cloud environments. In this context private and public resources are acquired and combined to set up tenant dedicated virtual infrastructures to fulfil distributed applications requirements. An exact algorithm is proposed to map jointly nodes and links of the requested virtual infrastructure graph to the physical graph from multiple providers (data centers and network providers). We view the problem as a virtual network mapping and use integer linear programming to find optimal solutions. To address complexity and scalability for large virtual and physical networks of thousands of nodes, an efficient heuristic algorithm, relying on topology patterns and bipartite matching, is used to provide close to optimal solutions and reduce mapping delays by three to four orders of magnitude.