Class of service based protection in multi-layered networks with SRLG constraints

Today's optical mesh networks are multi-layered networks, powered by bandwidth-hungry applications that can encounter multiple concurrent failures due to a single fault. Class of Service (CoS) is a level of availability required by the user and provided by the network. It is defined as the end-to-end availability of a light path. Guaranteed or 100% service is the highest type of service class ensuring that a light path will always be available. To provide the required CoS-based protection, a routing algorithm which is efficient, robust and able to assess all possible failure points is required. However, a protection mechanism in one layer may not be sufficient to provide guaranteed service. To tackle this problem, we employed the concept of Shared Risk Link groups (SRLG) as physical fiber links that share the risk of failure at the same time. SRLG diversity and link diversity together provide survivability in both the logical and physical layers. Despite extensive research in the field of network survivability, there has been little work done to achieve optimality with SRLG diversity. In this research, we strive to c9mbine SRLG diversity with its failure analysis and look to achieve optimality through bandwidth efficiency.;The ILP is formulated with an objective function to minimize the failure cost and backup bandwidth capacity. It proceeds to find all possible route pairs simultaneously under given constraints. The constraints include: Flow, link diversity, SRLG diversity for guaranteed service and SRLG failure probability for partial service. The heuristic algorithms were implemented in C ;The heuristic algorithms were compared with a link-only protection algorithm (without SRLG consideration). Analysis of the results indicates that fewer demands are dropped in an SRLG-protected network compared to link-only protection. ILP and heuristic algorithms were compared and it was found that the ILP algorithm always found the optimal path in the network.;In this research we tackle three major problems: (1) network survivability (2) optimization of spare bandwidth resources, and (3) offering differentiated classes of service. We present two classes of service: Guaranteed Protection and Partial Protection based on the level of SRLG diversity, failure probability and bandwidth availability. We also develop two novel heuristic algorithms called: 'One-Step Guaranteed Protection Algorithm' and 'Two-Step Partial Protection Risk Algorithm'. With both algorithms, the working path cost function is based on the number of SRLGs and their failure rates. The backup cost function is based on complete SRLG diversity for 'One-Step Guaranteed Protection Algorithm'. However the backup cost function in the 'Two-Step Partial Protection Risk Algorithm' provides partial SRLG diversity for 'high-risk' common SRLGs that fail the user-specified availability criteria. These heuristic algorithms are fast and efficient for large networks, however are sub-optimal in terms of computing the backup path based on an acceptable failure rate for the working path and backup path together. To overcome this problem, we use another method called an Integer Linear Program (ILP).