Interesting about 10 gigabit ethernet

[spidey07]

So Siemon did a big spiel about 10 Gig ethernet over copper. They covered some interesteing aspecs of cable design...

Cat 6 will not support 10G for more than 55 meters, the spec will state this.
Cat 6e is uselss
Augmented Cat 6 will support 100 meters of 10G, spec will state this
Cat 7 will support 100 meters and not have any difficulties with alien crosstalk (pairs in other cables causing crosstalk...each pair is individually shiled. Nice cable actually)

augmented cat 6 will use different approaches to mitigate Alien XT:
None - uses different fill ratios and installation methods
Foil UTP - similar to shieled...has foil shield on outer jacket
Dielectric - use larger jacket to keep cable pairs in one cable away from pairs in another

So its all very interesting but I couldn't help but feel like this was nothing but a sales pitch. They did have a very knowedible person from Siemon who is editor of the standard in draft. But as soon as she said "nobody is doing cat 6 today...we sell primarily augmented Cat 6 and Cat 7" my BS meter went into overdrive.

With all that being said it was interesting to say the least. I still don't see an application for 10G over copper. But it is coming none the less.

Good whitepaper if you can weed through the spin.
http://www.siemon.com/us/white_papers/0...net_Over_Structured_Copper_Cabling.asp

 

[Genx87]

I must be out of the loop, Cat 7?

10Gb ethernet may make interoffice\floor backbones cheaper? No more having to use fiber?

At 100 meters however that is very limited in scope imo.

 

[spidey07]

Originally posted by: Genx87
I must be out of the loop, Cat 7?

10Gb ethernet may make interoffice\floor backbones cheaper? No more having to use fiber?

At 100 meters however that is very limited in scope imo.

Category 7 cable.

As far as backbone, those will probably always be fiber. And the 100 meters follows 10, 100, 1000 Base-T - they are max 100 meters as well.

This is more about "to the desktop" and the horizontal cable needed to support it. Horizontal cable being the stuff that runs from the wiring closet to the desk/outlet.

 

[randal]

Honestly, that cat7 must be up there pretty close to fiber when it comes to cost ... so ... why not use fiber again? And for when it comes to the desktop ... I don't really see 10gbps being necessary at the desktop in anything approximating the near future. I mean, gigE is still rolling out, with 100mbps being very plentiful & generally `fast enough.`

I appreciate the innovation, but the ubiquituous deployment of 10Gb is far away. GG on the spin & hype, though.

 

[spidey07]

Originally posted by: randal
Honestly, that cat7 must be up there pretty close to fiber when it comes to cost ... so ... why not use fiber again? And for when it comes to the desktop ... I don't really see 10gbps being necessary at the desktop in anything approximating the near future. I mean, gigE is still rolling out, with 100mbps being very plentiful & generally `fast enough.`

I appreciate the innovation, but the ubiquituous deployment of 10Gb is far away. GG on the spin & hype, though.

Yeah. I'm not buying it either. I just don't see it happening even in 5 years, probably not even 10.

As far as fiber the big reason why that idea is dead is "power over ethernet" and a lack of affordable client cards.

 

[cmetz]

spidey07, my read of this document is that there will be an actual cat6e sometime in the future, which covers the deltas to get 10GBaseT over cat6 in the same way that cat5e covered the deltas to get 1000BaseT performance over cat5. It's just that some vendors are calling their supercalifraglistic cable "cat6e" now because marketing people thought that would be k00l. Those vendors will get smacked, as they should be.

I believe that 10Gb is a lot more desirable than you seem to believe, because it will be used in the near term to interconnect gigabit-class switches. Take the old 10/100 uplinked by gigabit architecture, and shift it up a power of ten. There are now $3500 switches from SMC with 48 10/100/1000 ports and a couple of 10Gig ports - but the optics modules for 10Gig are sold separately, and probably a lot more than the whole rest of the box. Thus, the drive to get this working over some sort of copper: $$$. As soon as this can move from an optics/photoelectronics quality issue to a DSP quality issue, they can begin ruthlessly driving down the cost in the same way as happened with 1Gig. Once 1000BaseT hit the market, the price dropped, FAST, and the availability of gigabit gear exploded. Now that 1Gig is getting commoditized and taken over by the Chinese/Taiwanese manufacturers, the higher end players need 10Gig ports to be a profitable place in the market to play.

The other issue with fiber is that fiber requires skilled installers. Twisted pair cable really does, too, but in practice it's a lot more forgiving. Well, cat5e/cat6 is. My suspicion is that the cat6e and cat7 that will be needed for 10GBaseT over twisted pair is going to be a lot more sensitive to installer stupidity than stuff we use today, including termination (terminating cat7 is a whole topic unto itself), bend radius, and keeping it away from noise sources. But time will tell.

And don't worry, 40Gig Ethernet is coming real soon now, so once this 10Gig stuff starts becoming more mainstream we'll have another high end to work on. Beyond 40Gig, the curve's going to flatten out some.... that darn physics stuff is going to become more and more of a nuisance 😉

 

[Sunner]

I doubt 10GigE will be very common outside the enterprises for another 10 years or so.
Heck, it was like 10 years since 100 Mbit equipment started becoming affordable for average Joe, and now GigE is becoming affordable.

 

[spidey07]

Originally posted by: Sunner
I doubt 10GigE will be very common outside the enterprises for another 10 years or so.
Heck, it was like 10 years since 100 Mbit equipment started becoming affordable for average Joe, and now GigE is becoming affordable.

True. But the lions share of network equipment is the enterprise/provider market.

Home and small business is almost nothing revenue wise.

 

[Sunner]

Originally posted by: spidey07

Originally posted by: Sunner
I doubt 10GigE will be very common outside the enterprises for another 10 years or so.
Heck, it was like 10 years since 100 Mbit equipment started becoming affordable for average Joe, and now GigE is becoming affordable.

True. But the lions share of network equipment is the enterprise/provider market.

Home and small business is almost nothing revenue wise.

True dat, was thinking more of the fact that articles such as this are more likely to affect home/SOHO users rather than enterprises who will be getting 10GigE out of need rather than greed(yeah, I play WoW).

 

[spidey07]

Well that's the thing.

I don't really see the need. cmetz makes an excellant point about its application however.

But the manufacturers are chomping at the bit to sell the new "latest and greatest" when for the most part 10/100 does just fine. With 1000 Base-T now in full force you'll see more 10 G in the backbone arena/interswitch links. I do agree with cmetz that the cost of the optics for 10G are hampering its deployment.

But these guys were pushing it for horizontal/workstation cabling. Only the most demanding applications would even use it (healthcare, imaging, engineering, etc)

 

[n0cmonkey]

Is fibre that much more expensive to produce? You'd think by now it'd be a bit more reasonable...

 

[spidey07]

Originally posted by: n0cmonkey
Is fibre that much more expensive to produce? You'd think by now it'd be a bit more reasonable...

The cable is relatively cheap. Its the optics (interfaces to send the light) that are expensive. For example I believe 10G pacs run 5-10 thousand depending on how far you need to send it. But its been a while since I've prices any.

 

[cmetz]

spidey07, I see a lot of interest in putting things like fileserver on gigabit links, not because they need a full gigabit, but because they need more than 100Mb/s. They basically are sites with 100Mb/s to the desktop (now pretty common) and figure that if a lot of desktops are sharing one big fast file server, that file server needs more bandwidth. But once you've got gigabit ports in your network, it's just a matter of time before that pool grows 😉

 

[Sunner]

Originally posted by: spidey07

Originally posted by: n0cmonkey
Is fibre that much more expensive to produce? You'd think by now it'd be a bit more reasonable...

The cable is relatively cheap. Its the optics (interfaces to send the light) that are expensive. For example I believe 10G pacs run 5-10 thousand depending on how far you need to send it. But its been a while since I've prices any.

Yeah, the equipment for FC is damn expensive.
We're looking at getting our own fiber between two sites, 1 Gbps speed, and cost of the equipment itself eclipses the cost of the physical fibre itself by a LONG shot...

 

[cmetz]

Sunner, I'm puzzled by your statement. A pair of SMC or Dell switches and long-reach single mode mini-GBICs should run you something like $4k. Unless you've got a pretty ideal situation (e.g., pre-existing conduit) I'd expect the fiber run between buildings to cost more than that. Now, if you need SONET connectivity, that's at least another digit.

 

[spidey07]

Originally posted by: cmetz
spidey07, I see a lot of interest in putting things like fileserver on gigabit links, not because they need a full gigabit, but because they need more than 100Mb/s. They basically are sites with 100Mb/s to the desktop (now pretty common) and figure that if a lot of desktops are sharing one big fast file server, that file server needs more bandwidth. But once you've got gigabit ports in your network, it's just a matter of time before that pool grows 😉

yeah. we've got thousands of servers and they pretty much on 1000 Base-t (moving server replacements to 10/100/1000 blades)

I also agree, servers pretty much need more than 100 base-t if they have any kind of load on them.

 

[p0lar]

Uses?? Plenty!

1) Switch trunks
2) metro-ethernet for ptp office/site connectivity
3) Extremely high performance, commodity-grade component iSCSI
4) large, live database redundancy

I could sit and spin a few more off, but these are things that even gigabit can't do well in an enterprise right now. I can't speak for home-based applications, but overkill is a lifestyle. 😉

 

[Sunner]

Originally posted by: cmetz
Sunner, I'm puzzled by your statement. A pair of SMC or Dell switches and long-reach single mode mini-GBICs should run you something like $4k. Unless you've got a pretty ideal situation (e.g., pre-existing conduit) I'd expect the fiber run between buildings to cost more than that. Now, if you need SONET connectivity, that's at least another digit.

Looking at Cisco routers for this, and for gig speeds, Cisco charges an arm and a leg for that.
I'm not heavily involved in that, but I do believe they wanted ~$16.000 for the endpoint equipment, while the actual fiber will cost us more like $5.000.