By Casimer DeCusatis, Ph.D., Distinguished Engineer, IBM System Networking Strategic Alliances, Member, IBM Academy of Technology IBM Corporation
Years ago, the distinction between telecommunications and data communications was pretty clear. Telecommunications referred to moving information (usually voice traffic) over long distances (hundreds or thousands of km), while everything else that stayed within a single data center was called data communication. Long distance voice traffic ran over ATM and SONET protocols, while many data centers contained a mixture of local area networks (Ethernet), storage area networks (Fibre Channel), clustering fabrics (InfiniBand), and other lesser used or proprietary protocols. Today, of course, the line between these two fields is blurry at best, and many people would argue that there really isn’t a meaningful distinction between them any longer. With the advent of new applications such as cloud computing and a renewed interest in long-standing applications such as disaster backup and business continuity, all types of traffic need to run well at extended distances. In this blog, we’ll consider the brave new world of distance-independent networking and protocol agnostic channel extenders using wavelength division multiplexing (WDM), a couple of topics that are sure to be widely discussed and debated in the many workshops and panels for OFC/NFOEC 2013.
Anyone interested in managing data traffic over distances of hundreds of km or more has to deal with the economics of long distance networking. It’s expensive to run fiber optic links over long distances, or to lease then from people who already own the right-of-way. This means that you need to have enough subscribers to your network service in order to share the cost of installing a long distance connection (this is why remote, rural areas often have trouble getting cable TV service, but often have wireless access). To help share the cost of a long distance optical network among many subscribers, we first turned to time division multiplexing (TDM) technologies. These were very efficient for voice traffic and concatenating many lower bandwidth signals into a single high bandwidth channel. But to more fully exploit the massive theoretical bandwidth capability of optical fibers, we must also use wavelength division multiplexing. This takes advantage of the rather elegant property that different wavelengths, or colors, of light propagating in the same optical fiber will not interfere with each other. As a result, we can currently pack as many as 80 different wavelengths over a metropolitan area network link, each running at speeds between 10 – 100 Gbit/second.
This technology will be the subject of many interesting papers at next year’s OFC/NFOEC meeting. For example, check out the tutorial on “Capacity constraints, carrier economics, and the limits of fiber and cable design” for some insight into the current tradeoffs between bandwidth and cost. There will also be tutorials on routing and wavelength assignment algorithms, and new research on ways to efficiently make the long distance network more dynamic and flexible in response to changing data center requirements. Years ago, when workloads were static, there was no need to provision a WDM link in less than a few days. Today, application aware networks need to respond in real time to changing bandwidth requirements from applications in retail, banking and finance, entertainment, education, medicine, and a host of other fields. Also, look for a keynote from the CEO of Adva Optical Networking on how companies can capitalize on the coming transformation of their networks. There is also a significant amount of new research in this field; for example, IBM recently announced a new silicon photonics breakthrough which could be used to create a WDM transceiver on a chip capable of data rates exceeding 25 Gbit/s. Using the same fabrication process, it is theoretically possible to develop a single chip capable of communicating at data rates of a terabit/second or more.
To understand the magnitude of the changes sweeping the WDM world, consider what’s happened to the telecommunication industry. There was a time not too long ago when provisioning a new voice service, such as an 800 phone number, was a lengthy and expensive process, reserved only for large, dedicated businesses. All that has changed with the introduction of technologies which allow service providers to provision new links in a fraction of the time it used to take, without requiring skilled human resources to spend days at the task. This made long distance service less expensive, and also helped drive new services (and new revenue streams) for telecom carriers. The data communication world is poised for a similar change, and the technical sessions at OFC/NFOEC will be filled with discussions on the best path forward for emerging industries such as cloud carrier exchanges, cloud service providers, and virtual server migration over WDM networks. The ability to implement dynamic wavelength provisioning, and wavelength tunable sources, has made it much easier to deploy wavelength services for data communication.
For the first time, many companies are starting to deploy 100 Gbit/second data rates over metropolitan distances, as short as a few hundred km or less. While 100 gigabit links remain relatively expensive for use within a data center, they can be economical when used to share a large number of lower bandwidth links using a combination of TDM and WDM. Since WDM systems can operate independent from the underlying network protocols, they are a particularly good choice for interconnecting multiple data centers which need to transport a wide array of different protocols. This same technology can be used to enable packet over optical networking for traditional carriers and cloud service providers. In fact, OFC will also include discussions on how to enable Ethernet over Optical Transport Networks (OTN) as an enabling technology for WDM systems. There are many examples of commercial products in this area, including those from industry leaders such as Adva, Ciena, Alcatel-Lucent, Huawei, and many more. You can expect all of these companies to be represented at OFC 2013, where over 550 companies will be showing off their latest advances on the trade floor and in technical session presentations. Furthermore, most WDM companies are beginning to adopt software-defined networking (SDN) technology to control their equipment, making it easier and faster to provision new links in response to applications which demand increasing bandwidth. SDN adds a whole new dimension to optical WDM communications, including the potential for end-to-end optimization of data communication traffic from a centralized network controller. Look for a keynote at OFC/NFOEC from one of the co-developers of software-defined networking.
Wavelength division multiplexing is just another reason to visit OFC/NFOEC 2013, and spend some time learning about the key technologies that you’ll need to design, deploy, and troubleshoot long distance optical networks. I’ll be touching on this subject as part of my tutorial on optical interconnection networks for datacom, and I hope to see you there. Or, feel free to continue the discussion online through my Twitter feed, @Dr_Casimer.
Disclaimer: Opinions, interpretations, conclusions, and recommendations are those of the author and are not necessarily endorsed by IBM.
Casimer DeCusatis, Ph.D. Distinguished Engineer IBM System Networking, CTO Strategic Alliances Member, IBM Academy of Technology IBM Corporation
Posted: 18 December 2012 by
Casimer DeCusatis, Ph.D., Distinguished Engineer, IBM System Networking Strategic Alliances, Member, IBM Academy of Technology IBM Corporation
| with 0 comments