Efficient relocation of virtual spare resources over optical backbone networks

Network virtualization has become an important feature for cloud-based services. Virtual networks can be set up over high-capacity optical networks. They consist of virtual links and nodes, which are mapped over physical resources. Operational virtual networks get gradually upgraded to avoid capacity exhaustion and create resources to handle traffic growth. These upgrades require leasing of new virtual resources (e.g., additional bandwidth for virtual links) and might be costly to network operators. The additional capital expenditure (CapEx) to upgrade the network can be reduced or eliminated by relocating spare virtual resources which were already-leased in the network, but are unused (e.g., due to optimistic forecasts or excessive overprovisioning). Relocating, here, is basically release some of the unused bandwidth from a virtual link and lease the same amount of bandwidth for another virtual link that will be upgraded. We investigate a cost-effective scheme to relocate spare virtual capacity in a virtual network over an optical WDM backbone network to (i) reduce or eliminate upgrade costs, (ii) avoid capacity exhaustion (by migrating resources from underutilized links to overutilized links), and (iii) improve network robustness (by migrating resources from links supported by unreliable lightpaths to links supported by reliable lightpaths). We propose an efficient relocation scheme to avoid capacity exhaustion and improve network robustness while minimizing or even eliminating requirement of new bandwidth. We build our problem into a mathematical formulation which turns out to be an Integer Linear Program (ILP). Our results show that network operator can significantly reduce CapEx while upgrading/improving a virtual network by using our relocation scheme.