Abstract

The explosive growth of the Internet saw a corresponding expansion of information exchange dwarfing the very modest designs of early networking elements. With increasing dependence on this exciting medium the number of users has increased manifold. The exponential growth in the services that needed to be supported has spurred traffic volumes, resulting in the need for huge bandwidth and faster communication. Network infrastructures had to keep pace with these rapid developments but were unfortunately saddled with severe bottlenecks that the only solution was costly upgrades. Routers, the mainstay of the Internet had to cope with a wide range of protocols. They have undergone several design modifications over the decades. A router essentially must perform two fundamental tasks-compute best routes and forward data packets. The evolution of routers is often described in terms of three generations of architectures. From the first generation with single CPUs and multiple interface cards with a shared bus, the routers evolved to the third generation with a switch fabric. Reservation signaling protocols have become an indispensable part of the Internet service. Resource reservation protocols were originally designed to signal end hosts and network routers to provide quality of service (QoS) to individual real-time flows. Recently, Internet Service Providers (ISPs) have been using the same signaling mechanisms to set up provider-level Virtual Private Networks (VPNs) in the form of MPLS Label Switched Path (LSP). Traditional IP router architectures cannot scale to meet these demands, forcing architects to explore alternative designs. However, due to rapid growth of the Internet, the architecture of the third generation routers are not able to meet the expected amount of traffic (i.e., multiple terabits or petabits per second). This development results in emergence of next generation routers with large switching capacity and high speed interfaces. The traditional centralized software model, where the control card is solely responsible for all routing and management operations is not able to efficiently utilize the new hardware platform architecture of line cards. Distributed architecture is one of the promising trends allowing routers to improve their robustness, scalability and resiliency. In this thesis path calculation and memory resources are proposed to be available on both control and line card in order to perform routing and forwarding tasks. We investigate the ability to reallocate components of the current MPLS/RSVP-TE architecture on to the line cards in order to share the load between the control card and line card. This allows significant improvement in scalability, resiliency and robustness of the system. New mechanisms for message exchange, synchronization, table and session management, and storage are developed. Performance evaluations in terms of CPU consumption, memory requirements, load balancing and bandwidth utilization for both centralized and the proposed distributed software architectures indicate that the processing time and memory utilization are significantly reduced