10gig over copper

[AFB]

When do you think we will see this ? If we see it

 

[spidey07]

5 years.

 

[AFB]

Originally posted by: spidey07
5 years.

Rubs hands together.

 

[spidey07]

I believe there are alread proprietary 10 gig copper solutions out there. but it is severely distance limited (like 5 meters, max)

I could be way off base, but I remember reading something about it. Specifically how 10 gig was going to require 8 pairs or 2 full cat7 cables.

 

[choby]

I think the first place you will see 10GB over copper will be router interconnects which are in close proximity. For example, a lot of big companies will run tandum routers in the net pop for redundancy. It is extremely inefficient to spend all the money on 10GB optics when you are only interconnecting two devices which are in adjacent racks. I dont know when we will start seeing these roll out, but im betting that is where they will appear first.

IMHO

 

[spidey07]

Originally posted by: choby
I think the first place you will see 10GB over copper will be router interconnects which are in close proximity. For example, a lot of big companies will run tandum routers in the net pop for redundancy. It is extremely inefficient to spend all the money on 10GB optics when you are only interconnecting two devices which are in adjacent racks. I dont know when we will start seeing these roll out, but im betting that is where they will appear first.

IMHO

I thought a few vendors were either working on such a system or already had it.

I don't know the interfaces of cisco's new big router, as its a whole new architecture.

 

[AFB]

LOL, I was talking about when you would see it in your local BestBuy. I think I remember reading where some company in Japan had gotten it to go for 300ft or so. Maybe it was here. Also, the environment was well controlled.

 

[spidey07]

Originally posted by: amdfanboy
LOL, I was talking about when you would see it in your local BestBuy. I think I remember reading where some company in Japan had gotten it to go for 300ft or so. Maybe it was here. Also, the environment was well controlled.

well if you look at history
1997 - switched 10/100 meg affordable for business nets (500 bucks a port, we thought that was a steal!)
2004 - gig ethernet ports affordable for business nets (nics built in, price premium between 10/100 is only about 25%)

So the transition from 10 to 100 to 1000 has followed a pretty sucessful recipe and time line. So I guess 100 meg switched has been cheap for what...3 years?

So looking into my crystal ball 10 gig copper will be ratified with production arond 2009.

I'm really guessing here, just following what has been done before. There are physical limits to the cable that are making 10 gig copper challenging. But then again, they said that about 1000 Base-T.

 

[ToeJam13]

I'm really guessing here, just following what has been done before. There are physical limits to the cable that are making 10 gig copper challenging. But then again, they said that about 1000 Base-T.

Agreed.

There are ten Gigabit copper solutions right now, however, they are limited to proprietary backplanes for corporate switches. Since distances are less than 1m, signaling is not an issue.

One of the largest problems with Fast Ethernet and Gigabit Ethernet is noise. External RF interference, in-cable echoing, and signal drop are killers. Besides advanced noise/error correction, I believe one of the changes between Fast and Gigabit Ethernet was an increase in the transmission voltage.

Remember that when Fast Ethernet was originally released, several vendors came together to make a legacy specification called 100Base-VG (Voice Grade). It was compatable with existing 10Base-T (CAT 3/4/5 UTP) wiring and LocalTalk (CAT2 STP) wiring.

The later is what impresses me the most: Fast Ethernet over category two cable. I imagine that ten Gigabit could be done using shielded category five or six cable. People aren't going to like replacing their existing CAT5 cable, but then, they didn't like upgrading their CAT3 and coax cable, either. They did it, though.

 

[spidey07]

Originally posted by: ToeJam13

I'm really guessing here, just following what has been done before. There are physical limits to the cable that are making 10 gig copper challenging. But then again, they said that about 1000 Base-T.

Agreed.

There are ten Gigabit copper solutions right now, however, they are limited to proprietary backplanes for corporate switches. Since distances are less than 1m, signaling is not an issue.

One of the largest problems with Fast Ethernet and Gigabit Ethernet is noise. External RF interference, in-cable echoing, and signal drop are killers. Besides advanced noise/error correction, I believe one of the changes between Fast and Gigabit Ethernet was an increase in the transmission voltage.

Remember that when Fast Ethernet was originally released, several vendors came together to make a legacy specification called 100Base-VG (Voice Grade). It was compatable with existing 10Base-T (CAT 3/4/5 UTP) wiring and LocalTalk (CAT2 STP) wiring.

The later is what impresses me the most: Fast Ethernet over category two cable. I imagine that ten Gigabit could be done using shielded category five or six cable. People aren't going to like replacing their existing CAT5 cable, but then, they didn't like upgrading their CAT3 and coax cable, either. They did it, though.

meh, standard life of cable plant is 10 years. which means many folks are starting to get to the end of life on their cat5 installations. At least they will in the next 5 years.

 

[nightowl]

The current spec for 10g over copper is for 50ft over infiniband cabling. This is for use mainly in data centers as choby mentioned where everything is in close proximity. It is also fairly cheap as Cisco has a $600 module for its 6500. Granted it looks like this only 1 port which is kind of a waste for a entire slot in a 6500 but it is still cheap none-the-less.

 

[ToeJam13]

The current spec for 10g over copper is for 50ft over infiniband cabling.

How true you are.

10GBase-CX4 (IEEE 802.3ak Standard) specifies 10Gbps Ethernet over XAUI and InfiniBand 4x. It uses four 100 ohm twinax cable with an Infiniband connector. Cable lengths up to 15m are supported.

10GBase-T (IEEE 802.3an Draft) seems to be a working model for 10Gbps Ethernet. Several manufacturers, including Broadcom, NEC, SolarFlare and Teranetics, have submitted draft proposals.

Manufacturers agree that 10Gbps over twisted pair is achievable using CAT 7 or CAT 6 STP. The noise floor of CAT 6 UTP may be too high, and the bandwidth of CAT 5e too low.

SolarFlare has demonstrated a working 10Gbps Ethernet model over 50 meters of CAT 5e. The problem is, the cable they used exceeded CAT 5e specifications (claiming 350MHz), even though it was sold as CAT 5e (speced to 100MHz). They claim that much of the cable on the market is the same. This may be true for the cable, but what of the jacks and patch panels? This is still often rated for only 100MHz, and will cause problems with 10Gbps.

A few other companies have shown test models using CAT 6 UTP. However, it is becoming clear that this cable type will only be good for short runs under 50 meters due to cross-talk and noise floor issues.

A majority of companies seem to agree that for 100 meter lengths, CAT 7 or CAT 6 with either screen twisted pair (ScTP) or double screened twisted pair (SSTP) cabling will be required. As such, we may soon see a new CAT 6e standard produced.

 

[Genx87]

I have rarely had a need for 100Mbit much less 10Gbit

But I do plan on getting a 1Gbit network setup in my house within the next 18 months.

 

[lexxmac]

Copper? You don't need no stinkin' copper! Use fiber!

Just kidding of course, but you could of course overclock gigabit ethernet signaling and using parallelism, say, four or five full cables each with four pairs of wire, get 10 gigabit ethernet. 100Mbit uses 2 (one send, one receive) pair @ 100MHz each, gigabit over copper uses 250MHz over four pairs in PAM (pulse amplitude modulation) to get the full one gigabit. Going much faster than 250MHz is going to be difficult becasue of the cable length. Oh, by the way, 350MHz over 4 pair cable is only going to give you 1.4Gbit. It is just so easy with fiber...

 

[Goosemaster]

Originally posted by: ToeJam13

The current spec for 10g over copper is for 50ft over infiniband cabling.

How true you are.

10GBase-CX4 (IEEE 802.3ak Standard) specifies 10Gbps Ethernet over XAUI and InfiniBand 4x. It uses four 100 ohm twinax cable with an Infiniband connector. Cable lengths up to 15m are supported.

10GBase-T (IEEE 802.3an Draft) seems to be a working model for 10Gbps Ethernet. Several manufacturers, including Broadcom, NEC, SolarFlare and Teranetics, have submitted draft proposals.

Manufacturers agree that 10Gbps over twisted pair is achievable using CAT 7 or CAT 6 STP. The noise floor of CAT 6 UTP may be too high, and the bandwidth of CAT 5e too low.

SolarFlare has demonstrated a working 10Gbps Ethernet model over 50 meters of CAT 5e. The problem is, the cable they used exceeded CAT 5e specifications (claiming 350MHz), even though it was sold as CAT 5e (speced to 100MHz). They claim that much of the cable on the market is the same. This may be true for the cable, but what of the jacks and patch panels? This is still often rated for only 100MHz, and will cause problems with 10Gbps.

A few other companies have shown test models using CAT 6 UTP. However, it is becoming clear that this cable type will only be good for short runs under 50 meters due to cross-talk and noise floor issues.

A majority of companies seem to agree that for 100 meter lengths, CAT 7 or CAT 6 with either screen twisted pair (ScTP) or double screened twisted pair (SSTP) cabling will be required. As such, we may soon see a new CAT 6e standard produced.

damn..I guess I made a repost in the other thread

 

[Goosemaster]

One QUestion:

I realize that they modify the frequencies to further prevent attentuation in the high and low frequencies( 10G over copper is like 3.xx Ghz) Just how do they do it and how do they compute or discover how to modify/filter the frequencies? In addition, how much of the 10G speed is due to the issue of ateenuation and how much is due to processing etc?

ANy links?

 

[nightowl]

Yes, it is easy to do 10Gbit over fiber or even faster. The problem is that the optics are friggin expensive. 7-10 grand for one optical interface versus under a grand for a copper interface is a lot better deal when dealing with shorter distances common to data centers.

 

[ToeJam13]

Yes, it is easy to do 10Gbit over fiber or even faster. The problem is that the optics are friggin expensive. 7-10 grand for one optical interface versus under a grand for a copper interface is a lot better deal when dealing with shorter distances common to data centers.

Which doesn't make sence. 10Gbps over single-mode fiber, yeah, I could see being expensive. You're using a real laser for the transmitter. Those kinds of prices are not justified for 10Gbps over multi-mode fiber. The transmitter costs for those have dropped so far...

If I were an Oliver Stone fan, I'd just say that it was one great conspiracy. They're out to get us! Then, it could also be attributed to price targetig - datacenter folk are willing to pay a lot for long distance equipment, no matter the costs.

 

[spidey07]

toejam,

supply and demand. R&D

sure transmitting at those speeds are no problem, being able to actually make sense of it are a whole new ballgame.

 

[SgtBuddy]

10Ge is available now.

I remember when I said I would probably never see GiGE in my house. It really makes a difference.