(This article is sponsored by The Boston Group)

Why IP over WDM?

Typically, the network consists of multiple layers, from the IP to the ATM to the SDH/SONET to the optical to the fiber duct. Each one of those layers offers their own flexibility, advantages, and features. The goal is to optimize the network in order to take advantage of the capabilities of a particular layer in a certain part of the network. For example, it does not make sense to do IP layer forwarding with traffic from New York to San Francisco at two or three routers in between. It would make much more sense to carry the traffic that is originating in New York to San Francisco in one single hop within the optical domain completely. That will provide higher performance, protocol transparency, and cost optimization.

Currently, there are too many layers in the protocol stack. IP is running over a frame relay and/or ATM layers. SONET/SDH, which is quite dominant, is also included here. Now, the emergence of WDM results in IP-over-ATM-over-WDM or IP-over-SDH-over-WDM.

Multiplicity of these layers is really a historical coincidence, and is the result of the need to be backward compatible. If we were to devise the protocol layers from scratch today then many of the layers (e.g., ATM, SONET/SDH) would probably not be there. Let's briefly discuss the reasons for running IP directly over WDM or optical.

WDM exists because the fiber is in place. WDM provides a lot of bandwidth, and IP offers great convergence, which is a winning combination. IP will be in the network because it is a revenue-generating layer, whereas WDM is a cost-reducing and performance-enhancing layer that also provides tons of capacity. ATM and SONET/ SDH do not add much value for data transport because SONET/SDH was designed for the circuit switched world where the basic bandwidth granularity was 64 kbps, when data was not as predominant as it is today. The value of ATM and SONET/SDH will also be minimized because IP will potentially deliver the quality of service. WDM will support the rich features of SONET/SDH, and nodes will be able to process information at line rates. Having said this, it must be noted that ATM, SONET/SDH and WDM continue to get deployed in the Internet core and metro networks for the reasons of various levels of granularity, protection/reliability, and capacity, respectively, that they offer.

The question is not whether IP over WDM is a long-term solution, but how long would it take for IP over WDM to get deployed at large scale. While this evolution takes place, SONET/SDH and high-capacity IP traffic will coexist in the optical WDM backbone. The SDH or SONET traffic could be multiplexed in the WDM network in individual wavelengths, so wavelengths could be added and/or dropped in the optical nodes. Similarly, an IP router with a WDM interface with tunable laser(s) could be connected directly with an optical switch or an optical router in the backbone.

Conceptually, it is quite possible to build a network node that will switch at different layers of the protocol stack. For example, at the lowest layer, one could have fiber cross connect. At a higher layer, it would be possible to cross connect the wavelengths from one incoming fiber to an outgoing fiber. At an even higher layer, one could do ATM switching with the SONET/SDH interfaces. Some of the wavelengths could carry the SONET/SDH traffic, which will have ATM in the payloads. Another layer up could have an IP switch. This type of platform, which could switch at multiple layers of the protocol stack, can indeed be built today.

IP over WDM involves higher layer IP protocol integration issues. For example, one challenge is to find techniques to do wavelength routing and wavelength assignment in order to build an optical lightpath based backbone networks. This may require a measurement-based approach that will relate with the open shortest path first (OSPF) protocols. Another consideration is lightpath assignment and routing. That is, can there be cut-through based on the lightpaths? Network reliability and protection switching are also major issues for the WDM networks. If SDH/SONET will be replaced, then WDM will have to provide network reliability and protection switching that the carriers are accustomed to in the SONET/SDH world, which has worked very well so far.

Another issue is how to access the channels. SONET/SDH can be used for channel access, but the overall goal is to minimize electrical-to-optical (E-O) and optical-to-electrical (O-E) conversion. There are a number of proposals for a generic WDM MAC, which brings about issues of framing at the WDM MAC layer.

The extent of optical transparency is also still unclear. That is, the extent to which one should do O-E or E-O conversion depends on the amount of the traffic load and the ability to carry optically transparent traffic as much as possible. One thing is clear that the O<=>E conversions and signal regeneration or amplification due to loss of power add significantly to the network costs and reduction in overall performance.

Another issue is managed versus unmanaged network bandwidth. Would there be a need to manage the bandwidth, the optical layer, and the WDM layer? Or could it remain unmanaged because of the tremendous capacity that will be available at negligible cost per bit?

Conclusion

The key challenges to IP over WDM are to build multi-layer, multi-service modular network nodes and to develop the supporting optical infrastructure. Inter-layer communication and management are also key challenges. For example, the WDM may have a client layer that will access some of the services of the WDM layer and provide the requirements from the client layer to set up the lightpaths. The industry is already developing techniques for topology design, routing and wavelength assignment, wavelength conversion, wavelength cut-through, channel access, fast packet switching, network survivability, scheduling and buffering, and quality of service in the optical network. Another issue is managed versus unmanaged network bandwidth. Would there be a need to manage the bandwidth, the optical layer, and the WDM layer? Or could it remain unmanaged because of the tremendous capacity that will be available at a very low cost per bit? For the new vision of the optical Internet to become a reality, the cost of optical components also must decrease, which has fortunately been declining by a factor of approximately fifty to seventy-five percent every year. In addition, the standards organizations and industry fora must standardize the solutions, define the interoperability requirements, and set the vision of the emerging end-to-end Optical Internet.

To find answers to some of these intriguing questions the reader is encouraged to read the book (http://www.wiley.com/WileyCDA/WileyTitle/productCd-0471212482.html).

Book Overview

The major focus of my book is why, how, and when of the issues associated with the support of IP over optical/WDM layer. To lay the foundation to effectively address the topic of IP over WDM, chapters 1 through 9 have been devoted to reviewing the background material on the technologies involved in enabling IP over WDM. These cover topics such as the protocol design concepts (TCP/IP), enabling optical device technologies, electro-optic and wavelength conversion, contention resolution in optical packet switching, wavelength switching and routing, access and metropolitan networks, and Ethernet passive optical networks (EPONs), The remaining chapters (10 through 16) specifically target the issues, requirements, solutions, and standards efforts pertaining to IP over WDM. These cover topics such as optical network engineering, traffic management for IP over WDM networks, optical burst switching, control and signaling, protection and survivability, and finally optical internetworking models and standards directions.

|