10GBase-T questions bring answers, which bring questions, which bring …
My assignment this month was to address two subjects: “When will we see 10GBase-T network equipment in the market? & ...
My assignment this month was to address two subjects: “When will we see 10GBase-T network equipment in the market?” and, “Given that there is not a Category 6A component standard, will there be an issue with the connector on the 10GBase-T network equipment?”
Down to ‘when’ ?
In June, the IEEE Standards Board ratified IEEE 802.3an 10GBase-T as an official IEEE standard. Based upon the IEEE 802.3 Ethernet frame format, this new high-speed data application will support a full-duplex transmission rate of 10 gigabits per second over 4-pair structured cabling (2.5 Gbits/sec throughput per pair). Developing 10GBase-T required huge advances in coding, signal processing and silicon integration technology.
During the course of its development, 10GBase-T PHY silicon industry companies participated in and closely followed the formulation of the 10GBase-T standard. Consequently, their major “network equipment manufacturer” customers should begin to see PHY layer chipsets as soon as late Q4 2006, which will allow network equipment systems manufacturers to assemble Ethernet switches and network interface cards (NICs) with 10GBase-T interfaces.
After several weeks of research, my best guess is that we will begin to see delivery of 10GBase-T network equipment starting in Q4 2007.
Equipment connector no issue
Today, September 2006, we still do not have a published cabling standard with even just a channel specification to support 10GBase-T based upon a worst-case 4-connector, 100-meter channel. But there are high hopes that will change after the TIA meeting in October. This is still just the channel, however, with no individual component specifications. My best guess is that we will not see a completed set of Category 6A component specifications before Q4 of 2008.
This will not be a problem; the connector on the network equipment is technically part of the equipment and, hence, part of the network equipment loss budget. You can rest assured that network equipment manufacturers will be using very high quality connectors regardless of what a Cat 6A component standard may someday require for the channel.
Where are we on the cabling?
Today, 80% of cabling in use inside buildings is Category 5e or better, but is this good enough? To date, the answer is a resounding maybe.
And there are two TIA documents in draft addressing the cabling performance requirements for 10GBase-T. The first is the proposed TIA TSB-155, which describes legacy Category 6 cabling as supporting the 10GBase-T application over 4-connector channel to 55 meters, and from 55 to 100 meters depending upon the field testing results. The second is the proposed Addendum 10 to the TIA/EIA-568-B.2 standard, which specifies Augmented Category 6 as supporting the 10GBase-T application over 4-connector channel to 100 meters.
While not addressed in a TIA document, the IEEE 802.3an 10GBase-T channel model defines individual measurements on the performance of cable in such a way 45 meters is achievable on Category 5e depending upon the field testing results…and we all know at least one customer that is going to try this, don’t we?
What about field testing?
The field certification of installed twisted-pair cabling for 10GBase-T includes all the test parameters currently specified in the TIA/EIA-568-B document for Category 6:
• Insertion loss;
• Return loss;
• Pair-to-pair NEXT;
• Power-sum NEXT;
• Pair-to-pair ELFEXT;
• Power-sum ELFEXT;
• Propagation delay;
• Delay skew.
The 10GBase-T test limits are identical to the limits for Category 6 up to 250 MHz, but the frequency range and performance requirements for these tests is extended to 500 MHz to support the much higher data rates.
The IEEE 802.3an helps a lot by substantially cancelling the effects of echo, near-end crosstalk (NEXT) and far-end crosstalk (FEXT). But now that we are dealing with frequencies to 500 MHz, we have a new set of transmission noise sources to address: Crosstalk between wire pairs in adjacent cables. Alien near-end crosstalk (ANEXT) and alien far-end crosstalk (AFEXT)-crosstalk from other cables-remain as the major impairments.
ANEXT can come from sources at the same end of a transmitting link or from sources that transmit at the other end, or somewhere along the length of the link, and a couple down the length of the link (AFEXT).
Alien NEXT sources are effectively mitigated through cabling design, qualification, and installation practices; on the other hand, alien FEXT proved to be a tougher trouble for cabling. To avoid the problem of a short-distance link far-end crosstalking into a link on a long line, the 10GBase-T requires transmitters to reduce their transmit power to only the level needed on their link. Short-link transmitters transmit with a power back off from the power used for 100-meter channels. Because the short link has much less signal attenuation than the long link, both links have plenty of margin.
Field testing of alien crosstalk is going to be very interesting! (See Hugo Draye’s article that begins on page 36 to get a feel for just how interesting.)
Déjà vu all over again?
Is it just me, or are you getting that “been there, done that” feeling-that eerie sensation of having already lived through a particular moment?
No, this is more like history repeating itself. A similar scenario unfolded during development of the 1000Base-T application standard. TIA TSB-95 defined the test parameters and limit values to certify existing Category 5 channels for use with 1000Base-T, while the standard for Category 5e was developed in parallel by the same committee.
While cabling channels that passed the requirements outlined in TSB-95 successfully supported 1000Base-T, Category 5e was chosen to become the recommended standard for 100 MHz channels with the release of the “B” revision of the TIA-568 standard.
The chicken or the egg?
Network equipment manufacturers design their equipment based on three factors: 1) capabilities of the underlying infrastructure, in this case cabling; 2) the percent of market share that the installed base represents; and 3) industry standards. In their view, it would be a hard sell for any network equipment that automatically required replacement of the cabling infrastructure.
Yet cabling manufacturers have been telling us that 10GBase-T would require a completely new cabling system: Theirs. They even offered cabling systems with “applications assurance” for 10GBase-T based upon a worst-case 4-connector, 100-meter channel since Q1 2003.
The next ‘big question’
So, while the cabling manufacturers are busy cost-justifying Category 6A and trying to stick a fork in Category 5e, and the PHY-layer chip manufacturers are busy trying to cancel the alien crosstalk monster and save the huge installed base of Category 5e and 6 cabling from the cable recyclers, what should you be doing?
In 2003, a highly respected member of our cabling industry said:
“Over the past 10 years, the demand for bandwidth has increased exponentially. Ethernet applications have evolved from 10Base-T, 100Base-T, 1000Base-T and now 10GBase-T. Memory and processing capacity at the chip level are known to double every 18 months. Networking equipment and the cabling infrastructure must be capable of keeping pace with this rate of change. For these reasons and others, including the growing number of IP-based applications, it is only a matter of time before the bandwidth of a 10G cabling system will be needed.”
- John Siemon, Chief Technology Officer, The Siemon Company
Following that logic, it is only a matter of a little more time before 10G also is exceeded.
My personal opinion:
Recable only if the need for 10GBase-T is present and the current cabling cannot support the application. For new builds, think beyond Category 6A to shielded copper systems and optical fiber.
DONNA BALLAST is BICSI’s standards representative, and a BICSI registered communications distribution designer (RCDD). Send your question to Donna at: firstname.lastname@example.org