Epidemic propagation of control plane failures in GMPLS controlled optical transport networks

In this paper, we investigate the behaviour of a dataplane-decoupled GMPLS control plane, when it is affected by failures that spread in the network in an epidemic manner. In particular, we consider network nodes to be either fully functional, or having a failed control plane, or having both a failed control and data plane. Through large-scale network simulation, we evaluate the effect of epidemically spreading control plane failures in terms of blocked connections requests and the amount of stranded capacity due to a dysfunctional control plane. Furthermore, we investigate the effect of the infection spreading rate and probability together with the infection start location in five network topologies. Our results indicate that the general availability of the network and the impact on its operation depend on a set of parameters, such as the network size and connectivity, the duration of the epidemic and the epidemic spreading intensity. In particular, networks with long epidemic durations do not necessarily result in worst performance in terms of blocked requests and capacity. Also epidemic scenarios, resulting in worst impact on the network availability does not necessarily result in worst network performance.