A new DWDM core switch design

Dai, Minjie
Major Professor
Committee Member
Journal Title
Journal ISSN
Volume Title
Research Projects
Organizational Units
Journal Issue
Electrical and Computer Engineering

WDM (wavelength division multiplexing) network has been envisioned as the next generation network carrier technology. IP (internet, protocol) is currently the dominant network layer technology and is believed to continue to be in the foreseen future. Although ATM and other protocols challenged furiously with IP in the past, IP networking has managed to make intelligent modification to accommodate various demands yet stay flexible and relative simple to implement. It has been strengthening its dominance in most area of competition. The seamless combination of upper level protocols with underlying WDM facility is the goal of future network design. My focus of research was on the core switch design for backbone networks. Its unique combination of cost structure and performance demand is the focus of design for real world applications. Trade offs were balanced among various options; discussion of choices of selections was also presented. To accommodate real time packet level switching, fast control and switching algorithms have to be deployed and hardware has to be able to reconfigure fast enough to service requests from control module. Therefore, fixed filters and wavelength converters were used instead of tunable ones to make it possible for faster reconfiguration for packet switching. There was no O/E/O conversion for the data; only control headers and routing information need to be converted to electronic form for analysis and computation. A new space division optical switch was also proposed to facilitate the exploitation of those components and to support desired functionalities. Multicast was supported since I believe it will be the vehicle for majority of the network traffic in the future. Synchronized network is also assumed to support more efficient use of bandwidth. The core router has to use control and switching algorithms that can overcome the bottleneck of computation to achieve high throughput, low delay and less storage, thus it can be cost effective yet support desired features. Decisions were made based on my knowledge of present and future network applications, and they intended to solve believed more important aspects of the problems. The success of this design can be tested by simulation in future effort.