Still pondering Gigabit Ethernet? It's not too early to think about the proper fiber for 10-Gigabit Ethernet.
In my October column, "The future of multimode fiber," I touched on some of the issues that are driving the development of new generations of multimode fiber. This month, I plan to expand on this theme, concentrating on the impact that gigabit laser-based systems will have on multimode fiber. Although this diverges from the Fiber Optics LAN Section focus on the horizon for the short term, there are some long-term implications that are important for network managers to consider.
After the standard for Gigabit Ethernet (GbE) over multimode fiber was ratified in June 1998, there was tremendous activity as companies rushed to develop and market products to meet the demand for GbE networks. And the standard itself continued to evolve with the development of GbE over copper and new discussions about 10-Gigabit Ethernet. While contemplating 10-GbE within the local area network may still shock some network managers, there are valid reasons to pursue the 10-GbE standard.
The main reason for developing a 10-GbE standard is that Gigabit Ethernet will move to the desktop and, even before that happens, will influence the choices that network managers make as they plan and deploy their infrastructure. Currently, GbE is used mostly to handle traffic in the backbone and riser, aggregating desktop traffic between hubs. As the bit rates at the desktop climb from 10 Mbits/sec to 100 Mbits/sec and eventually to 1 Gbit/sec, the bit rate in the riser and backbone will have to increase. Even though there are not yet applications that require GbE to the desktop, the products are available. Several GbE network interface cards exist, as well as GbE hubs for the desktop. More products continue to be introduced regularly, and as they become more prevalent, the infrastructure will develop to support them. Most important is that GbE in the horizontal translates to the necessity of having 10 Gbits/sec in the building riser.
While many companies are still contemplating the move to GbE, the IEEE 802.3 committee already is working on developing a standard for 10-GbE. They formed the High Speed Study Group (HSSG) to begin work on a 10-GbE standard and propose having a draft standard in the spring of 2001. Until then, several issues need to be addressed, including:
- Should 10-Gbit/sec transmission be achieved using serial transmission with a single wavelength or wavelength-division multiplexing with four wavelengths of 2.5 Gbits/sec each?
- Should multilevel encoding (MLE) be used on the bit stream?
- What media can be used at which lengths?
- What should the loss and jitter budgets be?
- What sources and wavelengths can be used?
- What should the baud rate be?
- What laser bandwidth is needed?
- Can the installed base of fiber be used at a shorter distance to achieve 10-GbE?
Over time, as the various proponents offer and argue their position on these issues, proposals will be modified and consensus will be reached.
There are some areas where the members of the standards group are already partially in agreement. For long 10-Gbit/sec links outside of buildings, standard singlemode fiber (SMF) is the media of choice. There are, however, ongoing discussions about SMF transmission distances as well as some investigations on SMF connector return loss.
Within the building, the standards group is looking to develop a system solution that provides the end user with the lowest cost and lowest complexity. It should also support legacy applications and be able to meet the needs of in-building networks with a single universal medium.
The current proposal is for a specially enhanced 50-micron fiber with a 0.2 numerical aperture (NA) and an over-filled launch band width of 500/500 MHz-km at 850/1300 nm. These are the standard specifications used today. The "enhanced" part comes from the laser-launch bandwidth requirement that (depending on the encoding scheme) will be around 2200/500 MHz-km at 850/1300 nm for serial transmission.
This enhancement is different from current fiber in two ways. First, the laser-bandwidth measurement spe cifies the performance of the fiber with laser sources. This measurement is not currently a standard. The standards bodies have completed much of the necessary research to develop a standard for the measurement using 1-Gbit/sec sources, but final publication of the standard is still pending. Additional work to characterize 10-Gbit/sec sources is necessary. When the standard is published, the 2200/500-MHz-km laser-bandwidth specification may change to a number that more accurately reflects the performance requirements using the new measurement.
The second difference is that this fiber is optimized for 850-nm short-wavelength operation. Current multimode fibers are manufactured to perform in both the 850-nm and 1300-nm operating windows. Thus, the same fiber can be used in an Ethernet network with 850-nm light-emitting diodes (LEDs) or a Fast Ethernet network with 1300-nm LEDs. The majority of GbE systems will operate using vertical-cavity surface-emitting lasers (VCSELs) because VCSELs are less expensive when compared to 1300-nm singlemode lasers and offer faster signaling rates than LEDs. VCSEL technology is rapidly improving, and VCSEL sources for 10-GbE are anticipated to be commercially available in the very near future.
Using VCSELs offers a low-cost and low-complexity system solution but requires improved performance from the fiber. To provide this level of performance, fiber manufacturers will need to modify their manufacturing process to optimize the fiber for superior performance at the chosen wavelength while ensuring adequate performance at 1300 nm with LEDs for legacy systems. This is not an easy task, but initial testing and early success has proven it can be accomplished.
Although GbE and 10-GbE are still a long way from dominating the desktop, bandwidth demand is growing exponentially. This means that network managers need to consider upgradability to faster speeds, like 10-GbE, when they are planning their networks today. Even though there are several technical issues to be resolved, work is progressing on the 10-GbE standard. The work to date clearly shows that multimode fiber is not only a viable solution for today's networks, but is also poised for growth for the future.
Steve Stange is chair of the Fiber Optics LAN Section (FOLS) of the Telecommunications Industry Association (TIA) and senior product manager for Transition Networks (Minneapolis, MN). He can be reached via e-mail at firstname.lastname@example.org.