Technology - IP Over WDM: Road To Optical Internet / Background
(This article is sponsored by The Boston Group)(Sudhir Dixit is a Research Fellow at Nokia Research Center in Burlington, MA. He has co-edited a book on Wireless IP and Building the Wireless Internet (Artech House, December 2002), and edited a book on IP over WDM (Wiley, March 2003). His third book on Content Networking in the Mobile Internet is due to be published by John Wiley in June 2004. )
IP over WDM might sound like an alphabet soup, but in simple terms it refers to running the Internet (i.e., the Internet Protocol (IP)) directly over the fiber (using wavelength division multiplexing (WDM)). For the past several years the telecommunications industry has been abuzz with megabits per second (Mbps), gigabits per second (Gbps), terabits per second (Tbps), and so on. We better get used to the power of the exponent, as major advances are happening in silicon, routing, switching, wireless access, and the fiber transport. Whereas the bandwidth, or the computing power, has been doubling per year and price has been decreasing by half every eighteen months or so, that threshold has been exceeded in the optical arena. WDM (wavelength division multiplexing, where a wavelength refers to a color of light) technology has revolutionized the data transmission rate in optical fiber, and forms the backbone on the global Internet today. A fiber, about the diameter of a human hair, can carry lights of different color, and each color can carry gigabits (thousands of a billion of bits) per second. When the bits of all colors are combined together you can send data at the rate of terabits (equivalent to thousands of gigabits) per second. The best thing about WDM is that once you lay a fiber you can turn on many colors of light in it. As you need more bandwidth you simply add more colors of light, each color of light supporting Gbps of capacity. Thus, you get a multiplier effect. Of course, this requires additional transmitters and receivers at each end. Each transmitter consists of a laser emitting a certain color of light, which is driven or modulated by the incoming digital bitstream. The opposite is done at the receiver by a detector to convert the modulated light back to the digital bitstream for each wavelength of light.
Nowadays, 40 wavelengths, each wavelength running at 2.5 Gbps, are common over a pair of fibers, one for each direction. Much newer systems being deployed are capable of carrying 160 10 Gbps channels (i.e., 1.6 Tbps) over a single pair of fibers. (In real terms, this means that a fiber can support 25B simultaneous phone conversations when voice is uncompressed or 200B connections when it is compressed to 8 Kbps/channel). Systems capable of running at 80 Gbps on each channel have already been demonstrated in the laboratory environment. 160 channels at this rate would equal 12.8 Tbps of capacity, which is much more than the entire bandwidth of the phone network in the world today. The rate of growth in fiber capacity is simply phenomenal, with plenty of room still left for growth (the theoretical limit being about 1015 bps or 1 petabit per second (Pbps) from ideal Physics).
A critical question arises as to what impact this capacity would have on the network infrastructure of the future, e.g., Quality of Service (QoS), digital loop carriers, protocol stacks, core network, metro network, and the access network. It is important to look at why wavelength division multiplexing (WDM) is so important and why is it important to run Internet Protocol (IP) directly over WDM. It is interesting to look at the historical reasons why there are currently so many layers in-between the IP and WDM layers. The industry is examining whether all of those layers will be needed in the future. If a protocol stack were built from scratch today, would all those layers still be included? If not, the industry must consider how to evolve toward establishing this infrastructure and what challenges lie ahead. Issues such as QoS (including reliability), time to provision (or to set up connections on demand), bandwidth granularity, cost, and management issues must also be considered before this vision becomes a reality. It should be noted that, often times, the terms WDM and DWDM (dense WDM) are used interchangeably and refer to the same thing. Recently, coarse WDM (CWDM) is receiving a lot of attention. The main difference between DWDM and CDWDM lies in how the wavelengths are packed together in the fiber.
Before we delve more into IP over WDM, it is worthwhile to review today’s infrastructure. It is usually built with three tiers of rings: local, regional or metro, and core or backbone (long haul). The rings connect with each other via either optical cross-connects (OXCs) for high throughput junction points or add-drop multiplexers (ADMs) for low throughput junction points. The access network typically consists of copper drops using analog channels, xDSL, ISDN, fiber drops, or hybrid fiber coax (HFC). The OXCs and ADMs in today’s architecture rely on optical-electrical-optical (O/E/O) conversion since only the interfaces are optical, but all processing in the node is done in the electrical domain. The metro and local rings mostly use SONET/SDH in the transport. The backbone rings or meshes use DWDM with multiple wavelengths in each fiber pair. The signal regeneration is also done in the electrical domain after every 200-500 kilometers (depending on the type of the fiber). This architecture dates back to the time when telephony was supreme and the network was optimized for the basic granularity of 64 kbps voice channels. The present day architecture suffers from the problems of high costs due to O/E/O that needs to happen in the regenerators, OXCs, and ADMs. It also makes it difficult for the operators to upgrade the network and it takes too long to provision the circuits through multiple rings, which by-and-large is still a manual process.
For these reasons there is a lot of push to use mesh networks as much as possible with IP/optical routers (with interconnecting DWDM links) setting up the lightpaths dynamically on-demand where all signal regeneration/amplification and routing will be done in the optical domain without any O/E/O conversion. Dynamic connection setups, which may encompass LSPs (labeled switched paths) from the IP layer all the way down to fibers, would slash provisioning times to milliseconds or seconds from days or months. Eliminating O/E/O conversions and reducing signal regenerations except in long haul networks will slash the costs.
Why IP and Why WDM?
A definite transition from circuit switched to packet switched networks has unquestionably occurred, and the packet mind-set has already set in. In volume terms, the data traffic has already surpassed the voice traffic today, and in about five years time, it will increase by about ten folds. Investment in IP solutions is natural since it is the common revenue-generating integration layer that is transparent to different link and physical layer technologies. This is despite the lack of IP layer support for real time services, or the services that require quality of service guarantees. While many QoS solutions for IP are still awaiting maturity and commercial success, fortunately this problem has been alleviated to some extent due to the rapid explosion in the availability of bandwidth that the optical networks provide. Because of this, real time multimedia services are already becoming possible even with the best effort Internet. In the near future we will see more deployment of fiber in the metro and access networks that will transform the Internet into a real end-to-end optical Internet with copper, coax, or wireless in the final drop to the customer premise still playing a major role. Of course, broadband wireless will also play a key role in connecting the user with the fixed infrastructure.
Packet-based public networks require massive, reliable, and scalable routers, add-drop multiplexers, cross-connects, etc. If the bandwidth capability continues to increase over fiber, the above network elements must be able to process the data at the line rate, preferably all optically. To realize this vision, the network architecture must be changed. The public switched network must also be scalable and quality of service enabled. It must support network-based servers, applications, and web farms. All this needs to happen at as least a cost as possible. Fortunately, the optical networks provide much lower cost per bit than the other non-fiber media that were used in the past. Broadband access is critical for the optical backbone capacity (which is at present overbuilt) to be fully utilized and economic benefits to be derived.
Increasing the efficiency of the network will require fewer network layers and integrated support for IP. That requires developing an adaptation layer (or an IP-aware MAC layer) that goes (as a shim) between the IP and the WDM layer. 2.5/10 Gbps will be the natural switching granularity in the backbone. If the rate reaches 2.5 or 10 Gbps on each channel in the optical fiber, then the optical transport layer should be in that range as well, since switching a packet at a lower rate is much more expensive in the optical domain than in the electrical domain. Finally, concatenated 155Mbps, 622 Mbps and 2.5 Gbps interfaces with light weight Synchronous Optical Network (SONET)/Synchronous Digital Hierarchy (SDH) framing will make performance monitoring easy with standard chipsets.
What Does WDM Offer?
WDM offers massive bandwidth multiplication in existing fibers. It solves the problem of best effort quality of service in the short term. While QoS standards (e.g., MPLS, DiffServ) mature and are implemented in production networks, enormous bandwidth capacity already exists that could potentially solve the problem of quality of service by overprovisioning. The upgrade is much less expensive than a forklift upgrade that is necessary with the SONET/SDH solution. WDM offers a potentially low-cost solution; three times the cost will amount to thirty times the capacity. It offers a secure network, needs less power, and requires low maintenance.
WDM also provides the ability to establish lightpaths. This could lead to minimizing the number of hops in the network. Numerous applications requiring QoS would benefit from having a single hop from one edge to the other edge of the network or from one major city to another major city just by having an all-optical (without E/O/E conversion) lightpath. WDM also offers improved rerouting and protection switching and transparency in signals. It can carry a mix of analog and digital signals and protocols. It is protocol transparent; it does not matter what a wavelength is carrying in its payload. Although initial deployments of WDM were point-to-point for transport, mesh topologies are now preferred with more and more intelligence residing in the optical layer. Thus, the WDM is evolving from a dumb tactical transport to an intelligent on-demand reconfigurable strategic network layer technology.
You may also access this article through our web-site http://www.lokvani.com/