Smart Algorithms for Hierarchical Clustering in Optical Network

Abstract

Network design process is a very important in order to balance between the investment in the network and the supervises offered to the network user, taking into consideration, both minimizing the network investment cost, on the other hand, maximizing the quality of service offered to the customers as well.Partitioning the network to smaller sub-networks called clusters is required during the design process inorder to ease studying the whole network and achieve the design process as a trade-off between several atrtributes such as quality of service, reliability,cost, and management. Under CANON, a large scale optical network is partitioned into a number of geographically limited areas taking into account many different criteria like administrative domains, topological characteristics, traffic patterns, legacy infrastructure etc. An important consideration is that each of these clusters is comprised of a group of nodes in geographical proximity. The clusters can coincide with administrative domains but there could be many cases where two or more clusters belong to the same administrative domain. Therefore, in the most general case the partitioning into specific clusters can be either a off-line or a on-line process. In this work only the off-line case is considered. In this Study, we look at the problem of designing efficient 2- level Hierarchical Optical Networks (HON), in the context of network costs optimization. 2-level HON paradigm only have local rings to connect disjoint sets of nodes and a global sub mesh to interconnect all the local rings. We present an Hierarchical algorithm that is based on two phases. We present results for scenarios containing a set of real optical topologies.