Fiber-based Fast Ethernet tries to catch up with copper
As an industry alliance pushes for an 850-nm Fast Ethernet standard for less expensive fiber-to-the-desk, some worry that copper-based solutions may have solidified its domination of the market.
Robert Pease
The current Fast Ethernet standard for fiber uses 1300-nm light-emitting diodes (LEDs), increasing the price tag on the migration from Ethernet (which uses 850-nm sources) to Fast Ethernet. The costs associated with this wavelength disparity make copper the more economically attractive medium for applications associated with Fast Ethernet...and that just doesn`t sit well with proponents of fiber-to-the-desktop architectures. With a "strength in numbers" initiative, the horizontal fiber marketeers are throwing their weight behind a new Fast Ethernet fiber standard in an attempt to penetrate the copper-dominated local area network market.
As standards go, the process appears to be moving rapidly and alliance members are thrilled with the progress. When compared to the three-year initiative to get Gigabit Ethernet standards accepted, the alliance certainly appears to be on the fast track. But could it still be too little too late? Has the lack of this standard caused fiber to lose headway in its effort to oust copper as the medium of choice in the horizontal?
Fast work for Fast Ethernet
The drive for a new standard began early last year, when 22 companies formed an alliance to draft a new standard for short-wavelength (850-nm) Fast Ethernet (100-Mbit/sec) transmission over optical fiber (see Lightwave, July 1998, "Push for short-wavelength Fast Ethernet standard," page 17). The primary focus of the Short Wavelength Fast Ethernet Alliance was to offer network designers an easy, sensible, and less-expensive fiber stepping stone from 10 to 100 Mbits/sec in fiber-to-the-desktop architectures.
The alliance has made significant progress since its inception and has attracted at least six new members (see page 58 for a list of alliance members). It also has attracted the attention of the Telecommunications Industry Association (TIA), which now sponsors the group under its Fiber-Optic LAN Section (FOLS). The group is in the process of releasing the ballot for comment within TIA on its draft proposal for the short-wavelength Fast Ethernet fiber standard and has invited IEEE, the home of Ethernet standards, to make comments on the draft standard.
At Networld+Interop last October, the draft standard was demonstrated using electronic and passive equipment from at least nine companies. The goal was to run Fast Ethernet at 850 nm over a distance of 300 m. After the successful demonstration, other companies joined the alliance to push for the standard.
A critical element to the draft standard--missing from the Networld+ Interop demonstration, but extremely important to the alliance--is the 10/100-Mbit/sec auto-negotiation issue. A smooth migration between 10 and 100 Mbits/sec without switching network interface cards (NICs), similar in nature and functionality to what copper has available today, is expected to further level the playing field between copper and fiber.
Who benefits from the standard?
Proponents of fiber-to-the-desk are quick to point out the benefits this standard will provide to end-users. The first and most obvious is lower cost. The existing Fast Ethernet standards for fiber were written for 2000-m distance specifications and require 1300-nm LEDs. The alliance feels that 2000 m isn`t typical among users, and there`s money to be saved for distances less than 300 m.
"We shortened the length distance to 300 m, which we believe covers the majority of intra-building installations, both vertical and horizontal," says Allen Dixon, chairman of TIA`s FOLS and senior product specialist in Siecor`s (Hickory, NC) premises marketing group. "The second cost reduction comes out of the first. Once you`ve reduced the scope of what you`re trying to do by shortening the distance from 2000 to 300 m, less expensive light sources can now be used because you don`t need all the power to drive it 300 m."
The FOLS climbed aboard because the alliance`s goals went hand in hand with the TIA group`s mission to promote the use of fiber and to educate customers on the advantages of fiber in the horizontal. Dixon says customers have needed and asked for a more cost-effective fiber solution in their LAN systems, and the short-wavelength option gives them what they want without the additional expense typically associated with fiber networks.
"This is a great idea for the end-user because it not only lowers the cost of fiber-to-the-desk, but also virtually guarantees upgradeability of his plant through the next couple of generations, including Gigabit Ethernet if it becomes necessary," Dixon surmises.
Of course, the customer isn`t the only one who stands to reap the benefits of the 850-nm, 100-Mbit/sec fiber standard. Manufacturers are working hard to get fiber electronics on an even keel with current copper offerings. From a manufacturing perspective, Dixon says the standard was deliberately drafted to allow use of off-the-shelf parts to build the architecture. As the draft stands, the standard can be implemented without the need for anything new to make it work.
"Even though the subcomponent manufacturers may have to do something differently with their parts," says Dixon, "they don`t have to develop new parts to make this system work. That also makes them a little more comfortable with the standard from that standpoint."
Is the standard too late?
While the progress of the alliance has been impressive, will it prove to be too little, too late for Fast Ethernet? For example, with a fiber standard for Gigabit Ethernet that includes both 850- and 1300-nm LEDs, coupled with the explosion of data-driven bandwidth demands, the casual observer may wonder why a market would even exist for Fast Ethernet. Why not grow an Ethernet network directly to Gigabit Ethernet?
Migration from 100-Mbit/sec Ethernet directly to Gigabit Ethernet at 850 nm is not a viable option for most network managers, mainly because of the cost. The electronics involved would be expensive to the point of being prohibitive for most LAN requirements, say industry sources. With most desktop networks only beginning to migrate to 100 Mbits/sec today, Gigabit Ethernet appears to be limited, at least for the near future, to backbone applications.
But the real question isn`t whether Fast Ethernet will compete with Gigabit Ethernet--it`s whether fiber-based solutions of any kind can compete with copper. "From a market perspective in comparison to copper, fiber solutions at either 850 nm or 1300 nm do offer an easier migration path to higher-bandwidth protocols such as Gigabit Ethernet," says Mark Thomas, an industry analyst with Ryan, Hankin, Kent (RHK). "It`s not a question of this standard being too late. It`s a question of whether companies want to spend money today to upgrade on a fiber network for more accessible bandwidth in the future."
If fiber expects to compete with copper at the desktop, it must provide comparable electronics at comparable prices. Thus, the alliance is fiercely promoting the issue of auto-negotiation for fiber electronics.
"We need to at least mirror some of the interoperability and auto-negotiation that you see for copper," says Dan Silver, product market development manager at 3M Telecom Systems Div. (Austin, TX). "If fiber is going to be used at 100 Mbits/sec, there needs to be some compatibility. With Gigabit Ethernet in the backbone and 100 Mbits/sec in the horizontal, vendors need to be producing NICs and media conversion if the end-user is to have some confidence in the ability of fiber in his network in terms of compatibility and functionality."
The 10/100 auto-negotiation capability would put to rest at least one negative to transitioning to an all-fiber system. In the copper world, 10/100 NIC cards have been on the market for a few years, a significant disadvantage for fiber.
"There are only a few suppliers involved in the development of products for supporting both 10- and 100-Mbit/sec operations," says Tony Beam, director of global systems manufacturing at AMP netconnect (Winston-Salem, NC). "At this particular point in time, I don`t know of anyone who is actually shipping any 10/100 fiber products to market. Although these cards, when developed, will be more expensive than the straight 10-Mbit/sec products, our hope is that they will be as cost-effective, if not more cost-effective, than the current 100-Mbit/sec product for long-wavelength."
Another factor to consider in migrating from Ethernet to Fast Ethernet is whether new builds would view fiber as a viable alternative to traditional copper if the alliance is successful in driving prices down. Many fiber-to-the-desk proponents and industry analysts agree that any time cabling is being replaced in the horizontal, a fiber plant makes good sense (see Lightwave, December 1998, page 45).
"For those who installed a copper solution using Category 5 or Category 6 cable and bought their cards and switches, the chances of them pulling out cable and recabling with fiber are slim to none," says 3M`s Silver. "If we`d had this standard a few years ago, it would definitely have given the end-user another option to confidently consider. But it`s still not too late. Yes, we lost some ground for fiber, but once we have this standard we`ll be able to give customers a fiber option they can have confidence in."
Dixon agrees that a new build should definitely look to fiber as an attractive option inside a building and between buildings. He points out that the same fiber plant may be used if the buildings are close enough together.
"For example, if the distance between your buildings is 500 m, a fairly substantial distance, you could conceivably put in the same fiber and use it for today`s 100 Mbits/sec, then turn around later and use it for Gigabit Ethernet, should it become necessary," says Dixon. "And I think almost everybody thinks it will become necessary."
RHK`s Thomas doesn`t see a clear advantage in performance or cost for 850-nm fiber Fast Ethernet versus copper. "In the absence of such an advantage," says Thomas, "it is unlikely that upgrades on installed systems will be done. For new installations, it may prove advantageous to deploy fiber, and this is where the standardization of the 850-nm standard for fiber is important."
Satisfied with progress?
So with the goal of the alliance to promote fiber to the desktop by offering a low-cost, interoperable and scalable system, are they satisfied with the progress they`ve made to date? The vision began almost a year ago and until the standard is passed, it may be difficult to win the battle against copper for Fast Ethernet applications.
"We`ve seen a tremendous amount of cooperation and support from all the vendors involved in this issue," says Silver. "The vote has been overwhelming to get this standard passed. Gigabit Ethernet took about three years and was considered a fast-track program, so to be where we are in less than one year seems like pretty good progress."
Dixon hopes to see the standard passed sometime this year. He says all feedback from the end-users indicates they like the short-wavelength Fast Ethernet option, and will use it to put fiber to the desktop.
"In my most wildly successful dreams, five years from now we`ll see all of the desktops connected via fiber," says Dixon. "That`s not realistic, but it sure would be great to see a significant ingress of fiber into the horizontal marketplace."
There are companies with allegiance to both copper and fiber, offering products for both mediums. One such company, Lucent Technologies (Murray Hill, NJ), views fiber as more of an option to copper, rather than a threat. Although he agrees with the alliance`s new standard because it offers another option for end-users, Lucent`s John Struhar, who chairs the membership committee for FOLS, doesn`t see copper falling victim to fiber anytime soon.
"I`m not willing to make a prediction as to when everything will be all fiber," says Struhar. "We think copper will be around for some time. People are buying so much copper right now that it would be impossible to guess its longevity. We`ve `killed` copper so many times and it keeps bouncing back. "
Dixon agrees to a point that copper continues to be alive and well in LAN applications. But he also provides some interesting food for thought. He points out that through the development of both mediums, fiber hasn`t changed much. It`s still basically the same multimode fiber plant that the industry recommended putting in back around 1989.
"On the other hand, if you look at copper, you`ll see some astonishing leaps in technology," says Dixon. "and each has required the end-user, in order to do the things required, to re-cable--every time."
And the fiber-versus-copper battle wages on. The short-wavelength Fast Ethernet standard proposed by this 28-company alliance will certainly help push fiber to the desktop. It is sure to save money for those LAN managers who view fiber as the best alternative for futureproofing their architecture for years to come. The standard will likely succeed in the development of new and less-expensive electronics, such as 10/100 NICs for fiber applications, as well as better interoperability equipment.
But will it bring about the demise of copper? Back in 1989, much of the industry (many writing in the pages of Lightwave) heralded the death of copper because there was just no way it could go to 100 Mbits/sec. Since copper proved them wrong then, there`s reason to use a little caution in any predictions of copper running out of steam. A more prudent response might be that fiber, with all its advantages, still has work to do before it can effectively compete within copper`s horizontal domain. Most believe the acceptance of this standard will be a great start, but then it`s back to the production labs. u