LAUSR

When pondering on the deployment and business cases of passive optical networks (PONs), dependability risk assessment is starting to play a key role. Not only entailing cost due to repairs or breaching service level agreements (SLAs), dependability risks also cover customer dissatisfaction and large outages affecting operators’ reputation. This paper proposes a risk assessment approach to evaluate the effect of hardware and software dependability in PONs with respect to these three aspects. First, a network geometric model (Manhattan model) captures the PON deployment, while software is modeled through a reliability growth model (Duane’s model) based on empirical data. Dynamic dependability behavior, combining both these models, is captured by a Markov cost model, solved by means of simulations. Thus, the probability distributions of the dependability-related costs and client dissatisfaction are shown, and also scatter plots of the clients affected by a failure versus the down time, for different time division multiplexedPON protection schemes. By considering the probability distribution, menaces that may go unnoticed with asymptotic analyses can be identified. Hence, risks threatening a successful PON in a business context with respect to costs, dissatisfaction, and large outages are pinpointed. Results show that, with stringent SLAs or high desired customer satisfaction, software failures are a threat to a successful PON deployment. Yet, hardware failures may lead to high costs and large outages sullying an operator’s reputation.