Designing VNT Candidates Robust Against Network Failures

Abstract

Future-generation networks are expected to be more robust to network failures. However, as each physical link carries multiple lightpaths when a VNT (Virtual Network Topology) is applied on an optical network, even the failure of a single link may tear down many links in the VNT, which can slow down the network or make it unusable. In this paper, we propose an algorithm called MFLDA (Minimum Flow Logical topology Design Algorithm) for designing VNT candidates that can accommodate a wide range of traffic patterns. Moreover, we show that the variant called MFLDA-FO (MFLDA with Failure Optimization) can design VNT candidates that have lower probability of congestion right after the failure of multiple nodes compared to HLDA, which is one of the best performing VNT design algorithms. Furthermore, we show that when these VNT candidates are used as attractors in an attractor selection algorithm, which was modeled on biological systems and proposed as a robust and self-adaptive control for future-generation networks, the average time to recover from difficult failure scenarios is less than the attractors designed by HLDA. Unlike HLDA, our VNT design algorithms and the attractor selection algorithm does not require the traffic matrix and the topology information after failure.