Leadership

Implement 10 GB Ethernet over copper with planning

One clear theme from Interop Las Vegas 2008 is that 10 GB Ethernet is upon us. A demonstration that caught my eye in particular was related to crosstalk when using 10 GB Ethernet on copper cabling.

I am again coming to you live from the Mandalay Bay convention center in Las Vegas at the Interop exposition. One clear theme is that 10 GB Ethernet is upon us. One demonstration that caught my eye in particular was related to crosstalk when using 10 GB Ethernet on copper cabling. In particular, alien crosstalk from other copper cables in the same bundle or run can cause severe noise interference depending on the cabling media selected.

At the show, I had an opportunity to see a demonstration from CommScope with their new GigaSPEED X10D (pronounced extend) cabling. In this demonstration, it was pretty clear that the cabling did make a difference in throughput over the line. One of the visible differences with the GigaSPEED X10D cabling was a small increase in space within the housing cable. The additional room makes a large difference in the amount of crosstalk, and more importantly, the effect of that crosstalk on other lines. In the demonstration, the GigaSPEEDX10D cable was compared against other CAT 6 and 6A for a 90-meter run for performance and crosstalk on the line. This was the network performance test performed to show the performance results:

Dispaly showing wire performance

Using copper cabling for 10 GB Ethernet will be a tempting topic to the network administrator. Expanding the copper infrastructure that may be in place to include 10 GB Ethernet performance is attractive from many perspectives. This includes cable management, patch panels, termination tools, cost, and ease of installation -- all making 10 GB Ethernet an attractive migration.

What kind of planning is needed?

We are only going to go to 10 GB Ethernet because we need the speed. Therefore, it would be a good idea to ensure that the best performance is met with your selected media. In the demonstration I was shown, the alien crosstalk was significant when compared to the other cabling that underwent the same test. But what does that feel like in overall network speed or overall functionality?

To plan your implementation, I would recommend a test that would have a number (at least six) of your candidate cable bundled together in the same fashion that you would run them up to your maximum length. Then run a series of repeatable network traffic performance tests that can give you a benchmark for the performance of that media. Then, repeat the test with as many other candidate cables as would be required to identify a clear winner.

About

Rick Vanover is a software strategy specialist for Veeam Software, based in Columbus, Ohio. Rick has years of IT experience and focuses on virtualization, Windows-based server administration, and system hardware.

9 comments
teckk
teckk

When it is impossible to get 1 GB speeds over CAT6. Radio wave interference, power cables close by, cable terminations - connections, hum from AC lighting. You can twist those pairs however you want. I don't believe it, getting 10 GB on copper. I think it must be in theory only.

RussInRedding
RussInRedding

With the cost difference so low, running fiber plant for your 10Gb links makes much more sense to me. Why worry about bad terminations, cross-talk, or other interference when fiber does not suffer from any of these? Not to mention, MUCH further distance capabilities!

walter
walter

Remeber that Cat7 was ratified before Cat6. Cat7 is a shielded solution. AMP and Siemon do a PiMF cable (Pair in metal foil) which has 4 twisted pairs with each pair enclosed in a metal foil and and overall screen around the cable. If I recall correctly the cable is rated at 700Mhz (vs Cat5e at 100Mhz and Cat6 at 250Mhz but tested to 350Mhz - Please note Mhz not Mbps). 100Meg runs at about 93Mhz (over 2-pairs) and 1Gig at about 60Mhz (over 4-pairs). PiMF cable costs more than a 4-core fibre cable but should work out cheaper on a per link basis once the ports have been purchased. PiMF will run 10GB, probably better than the X10D cable tested which I think is a UTP cable (American manufaturers prefer UTP solutions but I think UTP has reached it's limits, for now. SFTP is more of a European specification.)

PeggiD
PeggiD

We seriously considered running CAT6A in a new facility but nixed it when we started evaluating the additional space required for cable mgt trays, conduit size, raceway size in modular furniture etc. While CAT6A may be a good idea for the data center the additional cost of CAT6A vs. CAT6 is only the beginning of the cost differential, not to mention the other headaches we ran into just trying to ensure that the infrastructure would support it.

walter
walter

The cost of impleting 10Gb over copper is higher than the cost of installing fibre, but be careful what fibre is used as poor quality can seriously affect your networks performance. Consider blown fiber as an alternative, especially in a campus type environment where digging up gardens and roads to re-lay cables can be prohibitive. The real cost in implementing 10Gb is not in the cabling but in the equipment installed. 10Gb equipment also tends to have low density ports and therefore requires more rack space. As far as I'm concerned 10Gb is, for the forseeable future, purely for use in a few backbones that have extremely high bandwidth requirements. Most of us aren't even putting any strain on our 100Meg links.

Michael Kassner
Michael Kassner

I tend to agree that copper is nearing its limits, unless there is a major technology breakthrough in the not-to-distant future. I would be curious to learn your thoughts about the new fiber optic switches by Cisco's co-founder Mr. Bosack. It looks like this switch technology could really help justify the decision (RoI) to use fiber for other than backbone applications. http://blogs.techrepublic.com.com/networking/?p=487

johnmckay
johnmckay

1) New build; Fibre as much as you can. 2) Upgrade; of existing flood wiring???? Big savings in kit, implementation time, upheaval, and costs, so long as it works. I guess the cross talk will be the killer but all those guys that installed quality cabling from the outset might just reap the rewards here. As for me; I work for a multinational company where we beg for 100 meg never mind 10 Gb. What a contrast.

Editor's Picks