The cost of the technology has already dropped considerably, but the watershed event will be the ratification later this year of standards supporting 10-gigabit speeds over unshielded twisted-pair copper cabling.
January 1, 2006
It will likely not happen because of a single killer application, but 10-gigabit networking is definitely going to happen.
Technology to push 10 gigabits a second over optical fiber in a local-area network has been on the market since 2001, and the standard was ratified in 2002. As with most bandwidth advances, 10-Gigabit Ethernet started off too expensive and too far ahead of immediate requirements to interest most network users.
“Not very many have done it yet,” says David Passmore, research director at Burton Group, a Midvale, Utah research firm.
The reason is simple. Except in a few vertical markets and applications, “there isn’t the benefit for the cost difference,” explains Roberta Fox, president and senior partner at Fox Group Consulting in Markham, Ont. The exceptions, Fox says, are medical and engineering applications requiring high-speed image transfers, and large server farms.
But the cost is gradually coming down and bandwidth requirements show signs of catching up.
“There’s no one specific application that needs 10-gigabit, so there’s no killer application,” says Paul Kish, director of systems and standards for IBDN cabling products at the Pointe Claire, Que., office of St. Louis-based Belden CDT Inc. “But it’s a killer enabler.”
With increased use of voice over Internet Protocol, video and other network applications, bandwidth needs will continue growing, contends John Schmidt, business development manager for TrueNet cabling at Minneapolis-based ADC Telecommunications, Inc. “You may not need 10-gigabit right away, but you’re certainly going to need more than one gigabit.”
And as organizations upgrade desktop connections from 100 megabits to one gigabit, says Jean-Francois Thietard, a U.S.-based product line manager with network equipment maker Compagnie Financire Alcatel of Paris, they will start looking at increasing the capacity in the backbone and from the wiring closet to the core from one to 10 gigabits.
Price of gear dropping
The price of 10-Gigabit Ethernet gear has dropped in the past couple of years, says John Yen, senior manager of the LAN switching group at Cisco Systems, Inc., in San Jose, Calif. He believes that will fundamentally change the type of customers adopting 10-gigabit technology.
The early adopters of 10-gigabit networks included universities and research institutions with very high bandwidth requirements. Hospitals have also shown interest, Kish notes, due to their increasing need to move large image files — such as x-rays – over internal networks.
Carleton University in Ottawa is in the process of upgrading its campus-wide network, providing 10-gigabit capacity throughout the core, says Ralph Michaelis, the university’s chief information officer.
Carleton doesn’t need the full 10 gigabits yet, but “for somebody going to build a new network today, you really have to take a good look at whether you’re going to do gig Ethernet or 10-gig,” Michaelis says. He calculated that it would be roughly as cost-effective to put in 10-gigabit as to run three one-gigabit links in parallel, a figure Yen corroborates.
So Carleton has not only built its campus backbone of single-mode fiber capable of 10-gigabit or higher speeds over long distances, but installed laser-optimized multimode fiber suitable for 10-gigabit right to the wiring closets, with dark single-mode alongside it for possible future use.
Already, Michaelis says, a few researchers are asking for 10-gigabit capacity to their labs. A virtual simulation laboratory now being built, for instance, will need high-bandwidth connections to the outside world. “We can do that with the infrastructure we’re building,” Michaelis says.
All that jazz
For another example of what 10 gigabits can do, look to a demonstration that a videoconferencing group from McGill University’s Centre for Interdisciplinary Research in Music Media and Technology (CIRMMT) staged at a supercomputing conference in Seattle in November.
Three 65-inch plasma displays were lined up side by side in Seattle, each connected to a separate camera, computer and high-speed network interface card in Montreal via CA*net 4, the 10-gigabit cross-Canada research network backbone operated by the Centre for Advanced Networks in Research, Industry and Education (CANARIE).
McGill music professor Gordon Foote was in Seattle for the demonstration, teaching a jazz ensemble of music students in a studio in Montreal. The highlight was when Foote picked up a saxophone and played along with the group in Montreal.
The large, high-resolution displays were not a first, explains Jeremy Cooperstock, associate professor of electronics and computer engineering at McGill, but the breakthrough that 10-gigabit bandwidth made possible was the very low latency, allowing participants to see what happened at the other end of the link when it happened, not a second later.
Only a high-speed landline connection could deliver that, he says. Delays on a satellite connection would have prevented Foote and the other musicians playing together.
McGill also set up a model train in its Montreal studio. Allowing conference-goers in Seattle to switch it among three loops of track by remote control to avoid hitting a mechanical cow gave them a feel for the network’s quick response.
The demonstrations earned McGill the conference’s Most Innovative Use of New Technology award.
While impressive, those applications won’t be useful in many businesses. But, Cooperstock says, high-resolution, low-latency videoconferencing would be in training and telemedicine. Imagine a specialist assisting with delicate surgery via a videoconference link, he suggests — “you don’t want them saying ‘don’t cut there’ and having that communicated a quarter of a second too late.”
Cooperstock also believes senior executives will be more receptive to videoconferencing for high-level meetings if it becomes more lifelike.
In addition to the education and health care sectors, Daniel Therrien, manager of product and technology marketing for Toronto-based Cisco Systems Canada Inc., says the financial services industry is a promising market. Schmidt at ADC agrees.
“Anyone who’s moving a large amount of financial data is a primary candidate,” he says. Kish adds that image and other high-volume file transfers, large storage-area networks and parallel and cluster computing will benefit from the added bandwidth.
Passmore, meanwhile, says that a move from Fiber Channel to iSCSI technology in storage-area networks will likely increase interest in 10-gigabit connections in that environment.
The cost of 10-gigabit networking has already declined from as high as $50,000 per port before the 10-Gigabit Ethernet standard was ratified in 2002 to a tenth of that or less, but the reduction so far is just the beginning.
The watershed event will be the ratification, in mid-2006, of standards supporting 10-gigabit speeds over unshielded twisted-pair copper cabling. Kish says using copper rather than fiber could lower the cost to $500 to $600 per port.
Augmented Cat 6
The Telecommunications Industry Association (TIA) is working on a version of the Category 6 cabling standard that is expected to support 10-gigabit speeds over twisted pair. Known as Augmented Category 6, it is meant to be complete by the summer of 2006. Meanwhile, a task force of the IEEE 802.3 working group — known as 802.3an — is developing a protocol for 10-Gigabit Ethernet over Augmented Category 6 and possibly other twisted-pair cabling.
The 10GBase-T standard should be ratified by mid-2006, says Bob Grow, a principal engineer at Intel Corp. and chair of the 802.3 working group. It would allow 10-gigabit speeds over distances up to 100 meters over new cabling, and possibly over shorter distances on older cable.
Without waiting for official ratification of the standards, several cable vendors have launched copper cable designed for 10-gigabit thr
And while copper was initially seen as playing only a minor role in 10-gigabit networking, it now looks more likely that the cabling status quo — fiber in the backbone, copper from there to the desktop — will remain largely unchanged as the world moves to 10-gigabit.
New copper cabling that can handle 10-gigabit speeds up to 100 meters will be the prevalent choice for horizontal distribution from the network core out to the wiring closet, and the link to the desktop itself won’t exceed one gigabit in the foreseeable future.
Andre Mouton, Montreal-based product line manager for Belden, says his company’s 10GX cable is catching on for general horizontal cabling rather than being limited to data centres and a few narrow vertical markets as the vendor first expected. Mouton credits buyers’ familiarity with unshielded twisted pair cable.
Those installing the higher-capacity cable may want to be cautious about counting on that familiarity too much, though. While it is still twisted-pair, cable designed for 10-gigabit speeds is a bit different. To combat the biggest problem with such high speeds — alien crosstalk — manufacturers have designed their cable to increase the space between wires within a cable and between adjacent cables.
This means that cables are thicker and have cross-sections that are not round. Peter Sharp, senior telecommunications consultant at Toronto-based consulting engineers Giffels Associates Ltd., says Category 6a cable will call for 1.25-inch conduit and in some cases bigger cable raceways will be necessary.
Even the weight of the cable may be an issue, says Schmidt.
Sharp also adds that the bending radiuses for the new cable will need to be wider — a fact that may cause substandard performance if inexperienced installers neglect it. While installation procedures for the newer cable won’t be dramatically different, “the tolerances on your workmanship are a lot slimmer,” says Rob Stevenson, communications division manager at cabling contractor Guild Electric Ltd. in Toronto, “so you have to take a lot more care and a lot more attention to detail.”
Oddly, that doesn’t necessarily mean making the cables look neat, Passmore observes. While cable installers often pride themselves on how neatly they can wire a building, he says, “with 10G-BaseT you want to be messy.” Cables bundled neatly together are more likely to suffer from crosstalk.
A further concern is that there are currently no defined testing criteria for copper cable designed to support 10-gigabit traffic, adds Mike Barnick, senior manager of solutions marketing at Systimax Solutions, a Richardson, Tex.-based unit of CommScope Inc.
Though newer twisted-pair cable is designed to carry 10-gigabit traffic up to 100 meters, existing Category 6 and even older cable may handle it over shorter distances.
Barnick says standards developers’ original goal was for 10-Gigabit Ethernet to work over Category 6 up to 55 meters. That goal may prove elusive. “The definitive answer has not come out,” says Barnick, “but I think it is becoming a consensus that 55 meters may not be a doable figure.”
The Telecommunications Industry Association (TIA) is working on a document called TSB-155 that will provide guidelines for running 10-gigabit traffic over existing Category 6 cabling, Kish says, but “there’s no guarantee that what’s in the field will meet those requirements.”
However well twisted-pair cable handles 10-Gigabit Ethernet, it isn’t doing so in the real world today. “The problem is, we don’t have the electronics yet,” Barnick points out. While cable manufacturers have forged ahead, aiming at customers who need to install cable now and want to prepare for future bandwidth needs, makers of network hardware are waiting for the standards to be complete. Barnick says designs are being finalized now and the products should be on the market this spring.
If you want 10-gigabit traffic to start flowing today, the only immediate answer is fiber. And while the idea of squeezing 10 gigabits a second through twisted-pair cabling raises eyebrows, it is worth remembering that not all existing optical fiber can handle such speeds over significant distances.
“There’s a whole lot of 62.5-micron out there,” Barnick says. While it handles one-gigabit traffic very well, he continues, 62.5-micron designed to work with light-emitting diodes can’t carry 10-gigabit speeds more than about 35 meters. “If I’ve got 100- to 125-meter runs of 62.5-micron,” Barnick says, “I’m capped.”
Fifty-micron, laser-optimized fiber, using laser light sources, can carry 10-gigabit signals up to about 300 meters — enough for the core of most single-building networks. For campus-wide and longer-haul links, single-mode fiber is the answer, offering distances of as much as 80 kilometers, says Yen.
The Optical Internetworking Forum (OIF), an industry group, is working on a technology called Electronic Dispersion Compensation (EDC) that could extend this to about 120 kilometers.
Even though relatively few enterprises need 10-Gigabit Ethernet today, it’s worth keeping an eye on the future as Carleton University has done. “Usually when you put in cabling you’re looking at a 15-20 year lifespan,” Kish points out, “so you want to put in something that will support this new technology.”
Grant Buckler is a Kingston, Ont. freelance writer who specializes in IT and telecommunications issues. He can be reached at firstname.lastname@example.org.