Adaptive Management of Virtual Network Resources

The past few years have witnessed a rapid emergence of large-scale, geographically dispersed, clouds offering in the form of an Infrastructure-as-a-Service (IaaS). The adoption of these services requires the deployment of new networking technologies. This in turn, ensures the performance of the offered cloud services. Network virtualization has been proposed as a key attribute of the future inter-networking paradigm, providing efficient resource management solutions. Among the challenges that need yet to be addressed is the necessity to provide dynamic quality differentiated network services. In addition, it is required to guarantee the availability of network resources in response to workload fluctuations. Finally, it is necessary to periodically re-optimize the resource provisioning to be able to provide efficient resource utilization. These challenges are the motivation behind this work which aimed at developing a novel adaptive resource management model based on network virtualization. First, the proposed work describes a novel Virtual-Network-as-a-Service (VNaaS) model offering differentiated network-aware cloud services, resulting in a guarantee for the quality of the offered applications. This is achieved by enabling the cloud application providers to accurately express their dynamic needs, demand constraints and their network latency tolerance. The proposed work also enables the infrastructure provider to offer Elasticity-as-a-Service (EaaS) for the communication links by estimating and reserving the adequate pool of resources needed to fulfill the network workload fluctuations. This EaaS is offered at differentiated levels according to the hosted applications bandwidth-sensitivity. Finally, the proposed work employs a novel network resource re-optimization technique. The latter efficiently performs rearrangement for the VN portions contributing to the fragmentation of the underlying network. Simulation results demonstrate the effectiveness of the presented work and the significant gains achieved in terms of better adaptive network resource management