Path based p-cycle for resilient MPLS network design : a thesis presented in partial fulfilment of the requirements for the degree of Master of Engineering in Telecommunications and Networking, Massey University, Wellington, New Zealand

Abstract

Resilient networks are those that are capable of continuously offering telecommunication services in the presence of network failures. New paradigms for reliable network design have been emerging and constantly improving network survivability. Failure-Independent Path-Protecting (FIPP) p-cycles are a path based extension of the well-known p-cycles and inherit the attractive properties of ring-like recovery speed and mesh-like capacity efficiency. They are suitable for application to the MultiProtocol Label Switching (MPLS) protocol that is widely used in Next Generation Networks (NGNs), in that it provides shared, failure independent, end-to-end protection to whole working paths. This thesis contributes to advance in the state of the art for FIPP p-cycle based resilient networking. We firstly examine the two basic models: known as FIPP-SCP and FIPP-DRS. This is followed by an introduction to a Joint Capacity Allocation (JCA) design based on the FIPP-SCP model, which is more favourable to be used in MPLS networks. The network design is referred to as the MFIPP-JCA model and involves three specific cases: i) The BR model allows for bifurcated normal routing and imposes no restriction on the use of FIPP p-cycles; ii) NBR or non-bifurcated routing, focuses on single path routing, while it retains the flexibility of a protection domain; iii) An SNBR model where the main difference from the NBR model is that only a single FIPP p-cycle can be used to protect a working path. Case studies investigated the performance of the MFIPP-JCA models and, for a comparison with the basic FIPP-SCP model. The main areas of interest include the total cost of capacity, the number of FIPP p-cycles, and the solution time. The studies are also performed regarding changes in performance with regard to the number of eligible cycles. Those candidate cycles are the N-shortest cycles that are selected on either a circumference-based or hop-based criterion. The final contribution of this thesis is an in-depth discussion on the implementation issues for FIPP p-cycles in MPLS networks. We propose two operation modes both for bidirectional FIPP p-cycles, and make a judgment on the potential of unidirectional FIPP p-cycles.