Research on Survivability Strategies of Virtual Network

Abstract

Virtualization facilitates heterogeneous cloud applications to share the same physical infrastructure with admirable flexibility, while resource efficiency and survivability are critical concerns for virtual network embedding (VNE). As more and more internet applications migrate to the cloud, the resource efficiency and the survivability of VNs, such as single link failure or large-scale disaster survivability, have become crucial issues. Separating the VNE problem into node and link mapping sub-problems without coordination might cause a high embedding cost. This dissertation presents two independent approaches to solve the aforementioned challenges. First, we study two-stage coordinated survivable VNE (SVNE) problem and propose an adaptive path splitting based SVNE (APSS) scheme. We first develop a concise anchor node strategy to restrict the solution space of the candidate substrate nodes, which coordinates node mapping with link mapping to limit the distance spans of the virtual links. Then, we employ an adaptive path splitting policy to provide full protection against single-link failures with partial backup resource, and design an agile frequency slot windows choosing mechanism to mitigate the spectrum fragmentation for link resource efficiency. Simulation results demonstrate that the proposed APSS scheme can achieve satisfactory performance in terms of spectrum utilization and blocking ratio. Second, we propose a synchronous evacuation strategy for VNs with dual virtual machines (VMs) inside a disaster risk zone (DRZ), which suffer higher risks than the VNs with single. The evacuation strategy exploits post-copy technique to sustain the online service alive and enhances synchronous VM migrations to shorten the dual-VM evacuation time. Numerical results show that the proposed strategy can outperform the best-effort scheme in terms of average and total evacuation times of dual-VMs.