Abstract:
Today, in the cloud computing context, the network is starting to arise as a true resource itself and not just as a required connectivity add-on. The ability to define ne...Show MoreMetadata
Abstract:
Today, in the cloud computing context, the network is starting to arise as a true resource itself and not just as a required connectivity add-on. The ability to define network resources (e.g. routing/switching elements, bandwidth) in this context is still scarce, but there are clear evidences that this is a future reality. In this article, we consider that cloud infrastructure services will allow the definition of complete infrastructures, to which we refer as Virtual Infrastructures (VIs). These VIs comprise computing, storage and network resources. This article specifically tackles the VI embedding problem, which is known to be NP-hard. Having in mind this fact, different embedding strategies based on Integer Linear Programming (ILP) formulation are presented. The problem is addressed by proposing strategies that target load balancing and energy consumption of physical resources along with re-optimization strategies. A thorough evaluation of the different strategies is performed, studying the VI acceptance ratio, physical infrastructure energy consumption, and the impact of reconfigurations. Results show that there is a clear tradeoff between improving the VI acceptance ratio and reducing the physical infrastructure usage without allowing re-optimizations. We study in-depth different ways of combined strategies.
Date of Conference: 03-06 July 2017
Date Added to IEEE Xplore: 04 September 2017
ISBN Information: