Scalable, reliable and cost efficient backbone networks

Today the bulk of internet backbone traffic is carried by optical WDM transmission links offering increasingly lower carrying costs per bit of traffic. However, we still lack affordable solutions to switch packets and lower capacity circuits in the optical form. The result has been the growth of a backbone network architecture with electrical and optical switches in multiple layers. In this thesis we focus on several key problems in each layer hindering the scaling up of the backbone network capacity and we propose practical solutions to solve them.; In the optical layer usually pre-allocated backup paths are used to increase the reliability of optical circuits. In many networks the backup paths are shared by one or more circuits to reduce the network cost. In this thesis we propose two mechanisms to maximize the shareability of backup paths, thereby minimizing the cost of an optical network without sacrificing reliability. When traffic is split into primary and backup paths, the switch has to support bicast (one-to-two and two-to-one) connections. We derive new results on the blocking properties of Clos switching networks that takes into account the presence of bicast connections.; A grooming layer bridges the gap between the high capacity optical layer and the diverse connectivity requests of packet networks. We propose an efficient algorithm to combine a diverse mix of requested circuits of different capacities (51 Mbps to 2.4 Gbps) into 2.4 Gbps optical circuits. We also propose a more cost efficient alternative to current grooming architectures.; Lastly, we address the scalability bottleneck faced by the packet switching layer. We analyze the performance of a novel group based packet switching architecture that was recently proposed. We also design queuing algorithms for this architecture. Results indicate that it is highly scalable and a prime candidate for matching the capacity of optical circuits by aggregating broadband packet traffic from very large geographical segments.