Optimal multicasting of multiple light-trees of different bandwidth granularities in a WDM mesh network with sparse splitting capabilities

• Authors: Narendra K. Singhal;Laxman H. Sahasrabuddhe;Biswanath Mukherjee
• Venue: IEEE/ACM Transactions on Networking (TON)
• Year: 2006

With the advent of next-generation, bandwidth-intensive multimedia applications such as HDTV, interactive distance learning, and movie broadcasts from studios, it is becoming imperative to exploit the enormous bandwidth promised by the rapidly growing wavelength-division-multiplexing (WDM) technology. These applications require multicasting of information from a source to several destination nodes which should be performed judiciously to conserve expensive network resources. In this study, we investigate two switch architectures to support multicasting in a WDM network: one using an opaque (optical-electronic-optical approach and the other using a transparent (all-optical) approach. For both these switch architectures, we present mathematical formulations for routing and wavelength assignment of several light-tree-based multicast sessions on a given network topology at a globally optimal cost. We expand our work to also accommodate: 1) fractional-capacity sessions (where a session's capacity is a fraction of a wavelength channel's bandwidth, thereby leading to "traffic-groomed" multicast sessions) and 2) sparse splitting constraints, i.e., limited fanout of optical splitters and limited number of such splitters at each node. We illustrate the solutions obtained on different networks by solving these optimization problems, which turn out to be mixed integer linear programs (MILPs). Because the MILP is computationally intensive and does not scale well for large problem sizes, we also propose fast heuristics for establishing a set of multicast sessions in a network with or without wavelength converters and with fractional-capacity sessions. We find that, for all scenarios, the heuristics which arrange the sessions in ascending order with respect to destination set size and/or cost perform better in terms of network resource usage than the heuristics which arrange the sessions in descending order.