10GBASE-T is expected to become reality in July. A new short-reach operating mode will be defined in the standard.
March 1, 2006
It is time to stop for a moment and look at where we are and where we are going in networking technology.
10GBASE-T is a fourth-generation technology that provides 10 times the speed for three times the price of the current generation of Ethernet. It is expected to become reality in July of this year, almost on schedule in the timeline of events, which ushers in a new generation of Ethernet every five years.
The technology to make this happen for distances of up to 100 meters on twisted pair copper requires a lot of horsepower under the hood — more transistors, higher clock speeds and more power. In fact, the amount of power that is consumed to perform all the Digital Signal Processing (DSP) functions is a critical factor for the developers of the next- generation switching equipment.
It is desirable to place all this functionality and performance into a small package, and at the same time, do it more efficiently. But is it achievable and what are the tradeoffs?
At the last meeting of IEEE 802.3an task force this topic was very much under consideration in the resolution of the latest 802.3an-D3-0-sponsor ballot.
A new short-reach operating mode will be defined in the forthcoming standard. Although it is not explicitly stated as such, the rationale for the short reach operating mode will be to conserve power and to reduce complexity.
Longer channel lengths and less performing cabling require more signal power and more complexity for signal processing and noise cancellation circuitry.
A distance of 30 meters was proposed as a starting point for the short reach operating mode. Both augmented Category 6 (Class EA) and Category 7 (Class F) cabling will be considered in determining what distances that can be achieved for the short reach operating mode.
A presentation from Cisco Systems indicated that early 10GBASE-T silicon may be as high as 15 Watts for 100m, 7 Watts for 55m or 4 Watts for 30m. The presentation showed that the expected distribution of cable lengths would be centered around 30 meters and most links wouldn’t need full length/full power capability.
Systems could use power management on short links to conserve power. As an example, a 48-port blade switch could be implemented using 200 Watts of power, consuming about 50 Watts for the fabric, 25 Watts for sundries and 125 Watts / 48 ports for the transceivers.
This would give about 2.5 watts per port average. The detailed power management and control is left to the implementation, and could incorporate such features as link characterization by DSP during training, automatically selecting reduced power mode depending on length, port prioritization, and allowing fallback operation at 1Gbps.
All power save modes would be optional and the PHY/system may implement some, none or all modes in any combination.
X2 small form factor
Another consideration is the selection of the package size that is used to house the pluggable 10 Gigabit-per-second (Gb/s) transceivers. X2, is a small form factor specification that has been adopted by network equipment manufacturers (NEMs) in the optical fiber transceiver market for 10 Gb/s.
The X2 multi-source agreement (MSA) specifies a module that is physically smaller than the popular XENPAK transceiver format. It mounts on the topside of the host printed circuit board (PCB) and uses the established electrical I/O specification defined by the XENPAK MSA.
X2 is targeted at enterprise, storage and telecom applications that do not require the thermal capacity provided by XENPAK.
The thermal capacity limitations of the X2 format is about 4 Watts, and although desirable, may not be practically realizable in a full featured 10GBASE-T transceiver implementation.
So what is the result of all these deliberations? The standard is moving forward on schedule as expected.
It will provide something for everyone, although the end user will need to be savvy to understand the capabilities of the 10GBASE-T equipment that will be selected for the application.
It could be that some vendors will implement only the short reach mode option as a quick entry to market for limited distance data center applications.
Other vendors will choose to implement transceivers with full length capability (up to 100 meters), and provide power management functionality (scaled back power) for shorter distances. In all cases, the equipment will need to interoperate with one another.
What does it mean for the cabling vendors? The penalty to compensate for transmission impairments in the cabling is greater complexity and higher power utilization in the electronics.
Therefore, for new installations, augmented Category 6 or Category 7 cabling will provide greater efficiencies and make a lot of sense, even for short reach applications, where they are not strictly required.
Paul Kish is Director, Systems & Standards at Belden CDT. He is also vice chair of the TR-42 engineering committee.
Disclaimer: The information presented is the author’s view and is not official TIA correspondence.