How reliable are the real-world optical transport networks?

Due to the increase of bandwidth supported by optical transport networks, it becomes important to identify how reliable the actual real-world network topologies are, and the relationship of reliability with other key factors. Reliability is essential in this context, since the failure of elements such as fiber links and cross connects may cause the outage of several associated optical channels, leading to large data losses. Here we address reliability related to link failures. For a fixed number of nodes, the maximum network reliability could be obtained by maximizing the number of links, but it is prohibitive due to the technological and cost constraints. However, for a given number of nodes and links, there exist a number of topologies that may have different levels of reliability. How can we build the most reliable topologies satisfying these constraints? How reliable are the real-world topologies? In order to answer these questions, we explore a class of topologies that maximize reliability. Such topologies correspond to a subclass of Harary graphs. In this paper, we compare the reliability of the real-world topologies to the most reliable ones. Also, we compare both sets of topologies with respect to cost. Our results show that, in some cases, after maximizing reliability, the resulting topologies may also present lower capital expenditure.