1+N Protection in Polynomial Time: A Heuristic Approach
Mirzad Mohandespour, Ahmed E. Kamal · 2010
The generalized 1+N protection, protects N link disjointly provisioned unicast connections by a single Steiner tree connecting all end points of the connections. By sending network coded packets on the protection Steiner tree in parallel with the working traffic, 1+N is able to recover from any single link failure without enduring the delay from switching to the backup path. This makes 1+N a better protection mechanism in terms of recovery time compared to traditional 1:N and in terms of protection capacity compared to 1+1 protection. Optimal cost provisioning and 1+N protection of a given set of connections is an NP-hard problem comprising of three NP-hard subproblems: partitioning of the connections, finding edge disjoint primary paths and Steiner tree protection circuit for the subset of connections in each partition. In this paper a polynomial time heuristic algorithm for 1+N protection is proposed which combines heuristic steps to address the three NP-hard components of the problem. Our simulations show that the heuristic algorithm provides average cost reduction of 29.2% and 18.5% compared to 1+1 protection in COST239 and NSFNET networks. An asymptotic bound is also derived for the case of complete graph networks which shows that 1+N can achieve maximum of 66.6% cost improvement compared to 1+1. When compared to the optimal 1+N solution from ILP formulation, the heuristic algorithm increases the cost no more than 13%.