Connection management and recovery strategies under epidemic network failures in optical transport networks

The current trend in deploying automatic control plane solutions for increased flexibility in the optical transport layer leads to numerous advantages for both the operators and the customers, but also pose challenges related to the stability of the network and its ability to operate in a robust manner under attacks. This work proposes four policies for failure handling in a connection-oriented optical transport network, under Generalized MultiProtocol Label Switching control plane, and evaluates their performance under multiple correlated large-scale failures. We employ the Susceptible-Infected-Disabled epidemic failure spreading model and look into possible tradeoffs between resiliency and resource efficiency. Via extensive simulations we show that there exist a clear tradeoff between policy performance and network resource consumption, which must be addressed by network operators for improved robustness of their transport infrastructures. Applying proactive methods for avoiding areas where epidemic failures spread results in 50% less connections requiring recovery, which translates in improved quality of service to customers.