Scalable algorithms for QoS-aware virtual network mapping for cloud services

Both business and consumer applications increasingly depend on cloud solutions. Yet, many are still reluctant to move to cloud-based solutions, mainly due to concerns of service quality and reliability. Since cloud platforms depend both on IT resources (located in data centers, DCs) and network infrastructure connecting to it, both QoS and resilience should be offered with end-to-end guarantees up to and including the server resources. The latter currently is largely impeded by the fact that the network and cloud DC domains are typically operated by disjoint entities. Network virtualization, together with combined control of network and IT resources can solve that problem. Here, we formally state the combined network and IT provisioning problem for a set of virtual networks, incorporating resilience as well as QoS in physical and virtual layers. We provide a scalable column generation model, to address real world network sizes. We analyze the latter in extensive case studies, to answer the question at which layer to provision QoS and resilience in virtual networks for cloud services.