Until recently, the Institute for Electronics and Electrical Engineers (IEEE) expected to approve the 802.3z specification for Gigabit Ethernet in March 1998. Last fall, however, the 802.3z Task Force discovered a problem, now identified as differential mode delay (DMD). At first, the task force thought that DMD would pose only a minor setback and would not delay the March deadline, but by February, the task force knew that DMD required further attention. As a result, the task force was forced to delay the deadline until June 1998, at the earliest.
THE PLAN
The 802.3z Task Force might never have encountered the DMD problem had the task force not insisted on creating a Gigabit Ethernet standard that works with existing cable. Existing cable meets the ANSI/TIA/EIA-568-A Commercial Building Telecommunications Cabling standard, more commonly known as TIA-568.
The TIA-568 standard was developed jointly by the Telecommunications Industry Association (TIA) and the Electronic Industries Association (EIA) and was adopted by the American National Standards Institute (ANSI). TIA-568 is a cabling standard that provides specifications for a vendor-independent cabling system that supports both voice and data requirements.
THE PROBLEM
To ensure that Gigabit Ethernet transmissions would work over existing cable that meets the TIA-568 standard, the 802.3z Task Force conducted dozens of tests. For example, the task force commissioned the FO2.2 subcommittee of the TIA/EIA group to conduct a series of tests using Gigabit Ethernet over multimode fiber from a number of vendors. The FO2.2 subcommittee distributed the multimode fiber to four laboratories, ensuring that testers would be unable to tell which fiber came from which vendor. The laboratories conducted several tests, traded the multimode fiber, and conducted these tests again. During the second set of tests, one of the laboratories discovered the DMD problem.
DMD occurs with particular combinations of lasers and multimode fiber. The DMD problem stems from the fact that multimode fiber was designed for light-emitting devices (LEDs) that spread a light signal evenly across all of the fiber's modes. In contrast, a laser emits a concentrated signal over only one or a few modes of fiber. As a result, when the signal reaches the receiver, this signal is too weak for the receiver to process.
DMD is a potential problem for any high-speed networking technology that uses lasers to transmit signals at gigabit speeds over multimode fiber. The 802.3z Task Force discovered this problem only because of extensive testing and because the task force was the first high-speed networking group to use gigabit speeds over such a long distance.
THE SOLUTION
DMD could have been resolved (or never even encountered) with tighter specifications for multimode fiber. According to Brian MacLeod, director of Marketing for Packet Engines Inc., TIA-568 specifications for multimode fiber are "kind of sloppy." It's not the developers' fault, MacLeod is quick to add, "it's just that they were not thinking of gigabit speeds when they were doing their work in the early 1990s." A tighter specification would have helped but would have defeated the goal of supporting existing cable.
Instead of revamping the existing cable specifications, the 802.3z Task Force assembled a team of optical experts, forming the Modal Bandwidth Investigation (MBI) subgroup, to find a solution to the DMD problem. MBI's solution, called a conditioned launch, entails conditioning the laser to spread the signal evenly throughout nearly all of the fiber's modes.
The 802.3z specification makes the conditioned launch solution mandatory for 1000Base-SX and 1000Base-LX transceivers on multimode fiber. Most 1000Base-SX transceivers are equipped with internal lenses that seem to sufficiently distribute the signal. However, 1000Base-LX transceivers must use an external device to evenly spread the signal.
This external device is a special kind of jumper, which is called a mode conditioning jumper, which goes between the 1000Base-LX transceiver and the multimode fiber. The mode conditioning jumper, MacLeod explains, "scrambles up that coherent laser signal, exciting enough modes in the fiber to guarantee that the signal successfully traverses the cable."
THE UNFORTUNATE HITCH
Since November 1997, the 802.3z Task Force has worked hard to characterize DMD and to confirm that the conditioned launch solution resolves the DMD problem. Unfortunately, at the beginning of February 1998, the task force concluded that more tests were necessary to ensure that the conditioned launch solution eradicates the DMD problem in all circumstances.
The 802.3z Task Force is satisfied with the tests conducted on 1000Base-LX transceivers. However, because these same tests are insufficient for 1000Base-SX transceivers, the task force must create new tests. Unlike the predefined tests the task force had been using, the new tests must be created from scratch--a time-consuming process that casts doubt on the revised deadline of June 1998. If the task force cannot meet this deadline, the Gigabit Ethernet standard will not be approved until September 1998.
THE DELAY'S IMPACT ON EARLY ADOPTERS
The 802.3z Task Force doesn't expect this delay to affect customers' plans to adopt Gigabit Ethernet. At this stage in the standards process, the only companies using Gigabit Ethernet are early adopters that are either evaluating Gigabit Ethernet or using Gigabit Ethernet to solve network performance problems.
If you are evaluating Gigabit Ethernet, you are not currently making a major investment. Instead, you are simply conducting tests to determine whether, in the future, Gigabit Ethernet might serve your company's needs. The delay should not affect your plans to evaluate Gigabit Ethernet at all.
If you are using Gigabit Ethernet to solve network performance problems, you are clearly using Gigabit Ethernet products that successfully solve these problems. Because the products are already working for you, you are probably less interested than other users in when the Gigabit Ethernet standard will be approved.
The real issue is what to do when you have an installed base of prestandard equipment that you can supplement only with post-standard equipment. "The message," MacLeod says, "is to have a plan in place with your vendor to carry you through the transition period."