Design and protection algorithms for path level aggregation of traffic in WDM metro optical networks
Wavelength Division Multiplexing (WDM) promises to offer a cost effective and scalable solution to meet the emerging demands of the Internet. WDM splits the tremendous bandwidth latent in a fiber into multiple non-overlapping wavelength channels, each of which can be operated at the peak electronic rate. Commercial systems with 128 wavelengths and transmission rates of up to 40 Gbps per wavelength have been made possible using state of the art optical technologies to deal with physical impairments. Systems with higher capacities are likely to evolve in the future. The end user requirements for bandwidth, on the other hand, have been ranging from 155 Mbps to 2.5 Gbps. Dedicating a wavelength for each end user will lead to severe underutilization of WDM channels. This brings to forefront the requirement for sharing of bandwidth in a wavelength among multiple end users.;The concept of wavelength sharing among multiple clients is called grooming. Grooming can be done purely at the optical layer (optical grooming) or it can be done with support from the client layer (electronic grooming). The advantage of all optical grooming is the ease of scalability due to its transparency as opposed to electronic grooming which is constrained by electronic bottlenecks. Efforts towards enhancing optical grooming is pursued through increasing optical switching speeds. However, technologies to make optical switches with high speeds, large port counts and low insertion losses have been elusive and may continue to remain so in the near future.;Recently, there have been some research into designing new architectures and protocols focused on optical grooming without resorting to fast optical switching. Typically, this is achieved in three steps: (1) configure the circuit in the form of a path or a tree; (2) use optical devices like couplers or splitters to allow multiple transmitters and/or receivers to share the same circuit; and (3) provide an arbitration mechanism to avoid contention among end users of the circuit. This transparent sharing of the wavelength channel utilizes the network resources better than the conventional low-speed circuit switched approaches. Consequently, it becomes important to quantify the improvement in achieved performance and evaluate if the reaped benefits justify the cost of the required additional hardware and software.;The contribution of this thesis is two fold: (1) developing a new architecture called light-trails as an IP based solution for next generation WDM optical networks, and (2) designing a unified framework to model Path Level Aggregation of Traffic in metrO Optical Networks (PLATOONs). The algorithms suggested here have three features: (1) accounts for four different path level aggregation strategies---namely, point to point (for example, lightpaths), point to multi-point (for example, source based light-trails), multi-point to point (for example, destination based light-trails) and multi-point to multi-point (for example, light-trails); (2) incorporates heterogenous switching architectures; and (3) accommodates multi-rate traffic. Algorithms for network design and survivability are developed for PLATOONs in the presence of both static and dynamic traffic. Connection level dedicated/shared, segregated/mixed protection schemes are formulated for single link failures in the presence of static and dynamic traffic. A simple medium access control protocol that avoids collisions when the channel is shared by multiple clients is also proposed.;Based on extensive simulations, we conclude that, for the studied scenarios, (1) when client layer has no electronic grooming capabilities, light-trails (employing multi-point to multi-point aggregation strategy) perform several orders of magnitude better than lightpaths and (2) when client layer has full electronic grooming capabilities, source based light-trails (employing point to multi-point aggregation strategy) perform the best in wavelength limited scenarios and lightpaths perform the best in transceiver limited scenarios.;The algorithms that are developed here will be helpful in designing optical networks that deploy path level aggregation strategies. The proposed ideas will impact the design of transparent, high-speed all-optical networks.