Sometimes, technology adoption is driven by the simple things like cabling and compatibility and that can be the case with the migration from 1Gb Ethernet to 10Gb Ethernet.
Sometimes, technology adoption is driven by the simple things like cabling and compatibility and that can be the case with the migration from 1Gb Ethernet (1GbE) to 10Gb Ethernet (10GbE). Before the advent of widespread 10GBase-T, most 10GbE adapters and switches support optical fiber or direct attach copper (DAC) connectivity via expensive SPFs that are not compatible with twisted pair cabling systems typically used in data centers. The result of this was using a switch to provide backwards compatibility with 1GbE. However, the next wave of 10GbE solutions are leveraging low-cost, backwards compatible 10GBase-T that is changing the cost complexity and cabling issues around the migration to 10GbE in the data center.
Analysts such as Crehan Research predict a greater acceleration of 10GbE adoption with a significant factor being emerging support for the 10GBase-T standard (based on the chart below, which indicates 10GBase-T is expected to grow exponentially). 10GBase-T enables 10GbE connections with unshielded or shielded twisted pair cables over distances up to 100 meters. This article explores how data centers can approach the migration to 10GbE using 10GBase-T adapters that are backward compatible with 1GbE and offer support for converged network environments where local area network (LAN) and storage area network (SAN) traffic move over one cable.
Data center wiring today
Today’s data centers have cables covering thousands of miles with a mix of optical and twisted pair cables—optical for Fibre Channel storage networks and twisted pair for IP networks, primarily 1GbE. Optical fiber provides the high-performance bandwidth required for Fibre Channel and twisted pair is the pervasive, cost-effective solution for Ethernet connections.
The biggest challenges with 10GBase-T have been performance characteristics of the cable and the power required to support 10Gb bandwidth with server adapters and switch ports. Part of the solution is Cat 6 and Cat 6a cabling that has improved insulation which minimizes alien cross-talk. The other part of the solution is reduced power requirements for newer 10GbE controllers that are used for server adapters and switches. Reducing the power consumption for the controller allows more power to be used for the physical connection (PHY) that sends signals over the twisted pair cable.
With these improvements, 10GbE adapter and switch providers are introducing 10GBase-T support that is becoming widely available at low prices. This will drive the accelerated adoption of 10GbE. The next step will be next-generation rack servers that ship using 10GBase-T for the LAN on motherboard (LOM) controller that is integrated into the server. In this phase, 10GBase-T will become the default network connection.
Need for backward compatibility
A high percentage of data centers are currently using 1GbE over twisted pair cable, particularly at the server access layer. Deploying 10GbE with optical fiber or DAC connectivity was an all or nothing proposition. 10GbE switches were required at the server access layer since there was no backward compatibility with 1GbE networks. 10GBase-T solves this problem. Servers with 10GBase-T adapters will auto-negotiate to support 1GbE or 10GbE. That enables you to deploy servers today in a 1GbE environment, and then migrate to 10GbE tomorrow, thereby future-proofing your investment.
As I know, the "alien crosstalk" was one of the biggest challenges for using the Cat6 cable for Gbit Ethernet. What is the change that is made in the insulation to reduce the alien crosstalk?
In the second para in this article, it is mentioned about using "unshielded cable" for 10GbaseT. Again, I still see a major risk in using an unshielded cable for 10Gbit/sec due to EMC concerns.
Sanjib – The Cat6A cables I have seen reduce alien crosstalk due to plastic ribs in their core which specially separate the twisted pairs. Since electromagnetic radiation intensity falls off as the square of the emission radius, it’s a very effective technique.
As for Electromagnetic interference, I would agree with you if it wasn’t for the recent advances in the DSP algorithms inside of the 10GBase-T transceivers which have been shown to provide immunity to very strong EMI events of up to 10V/m. Those DSP engines are able to place a notch filter at the interferer’s emission frequency and prevent it from causing havoc with the received data.
Hi Ron, I got your points on crosstalk & EMC. Yes, that's how it could be done. Thanks a lot!
Regarding the EMC (or rather EMI?) I was more concerned with the "unshielded cable" emitting RF energy creating trouble for others. Any thoughts?
Sanjib – Yes, radiated emissions also need to be contemplated. But the good news is that the IEEE Standards body anticipated the potential for radiated emissions and has taken corrective measures. First, recognize that, like in 1000Base-T, 10GBase-T signaling is differential in nature and transmitted on twisted pairs that cancel emissions to first order. Nevertheless, there is residual radiation and even some common mode components due to non-idealities. To cope with those, IEEE was careful to architect the output energy of a 10GBase-T transmitter to be more evenly distributed over the transmission band and, furthermore, in Clause 55.9.5, requires compliant equipment to meet FCC Part 15 or CISPR 22 (depending on geography) limitations for radiated emissions. Bottom line: it’s purpose built to live and work in a data center environment without causing problems.
Good point! I am not much concerned with only one cable. But I think, usually there would be many...isn't it? I thought it would be challenging to meet EMC norms as they would emit more when the cables are unshielded?
But again I saw many have passed the EMC tests. One such example could be found below: