Slashdot Mirror
Category 6 UTP Standard is (finally) Here
An anonymous reader writes "This is only important for the networkphiles out there, but the Category 6 UTP specification is finally here. The standard is the TIA/EIA-568-B.2-1. The significance of this is that now you can transmit at 250Mhz frequencies (vs 100Mhz of Cat 5/5e). So 1Gbps is easily achievable. Of course ther's still Category 7 (600Mhz) in development, but I guess we should eventually move to fiber." Who
hasn't crimped cat-5 before?
...When you're wiring about 500 workstations and servers over a reasonably sized office. You run into having to buy literally *miles* of cable when you wire even a medium-sized IT office. At that volume, buying Cat6 or Cat5 is non-trivially less expensive than fiber.
The next Slashdot story will be ready soon, but subscribers can beat the rush and slashdot the links early!
but I guess we should eventually move to fiber
Usually for me it's the other way around, Fiber gets me movin'.
"Oh no, 3 horny women and only 2 condoms...Thank god I read slashdot"
Unless you put your fiber cables in an unbendable channel, it isn't worth the hassle of having to replace a faulty cable because some bozo decided to fold the cable up and break the fiber. I have seen this happen many times.
For the forseable future, gigabit to the desktop is more than 95% of users will need unless computing environments move to server-side VR operating systems that are fully streamed to a user with full motion and sound.
Server back planes and clusters are two of the biggest bandwidth hogs that might possibly need something faster than gigabit ethernet.
A friend of mine was running cat7 in building a few years ago.. odd? I never payed much attention to that at the time, since I assumed it was for real. what I want to know is why wer are still using TP when we have fiber available? we should be driving the price of fiber down, not TP. lets face it, TP can be prone to interference from strong sources, ect. Just my opinion..... If we drive the price of fiber down, maybee we can all get 3l33t fiber links in our homes :)
I was under the impression that the currently available solutions for 1000Mbit over copper worked just fine on Cat5 installations as long as the distance was kept short.
If that is the case, what benefit does Cat6 bring to the table? More distance? Lower bit-error-rates? Something else?
Free your mind!
I get redirected to http://www.tiaonline.org/browser_error.cfm.
Browser Requirement Error
To view this site you need a browser capable of suppporting HTML 4 or higher.
Download Microsoft Internet Explorer
(recommended)
OR
Download Netscape Navigator
I worked for a medium-sized IT consulting firm. When we moved into a larger office space, they saved money by making everyone in the office make patch cables. Office Admin., everybody. Glad I was billable :)
My sig sucks.
So we plan to save money on Cat-6 by using two Cat-3 cables in parallel.
CEE5210S The signal SIGHUP was received.
Why does everyone keep talking like fiber is new? It was very available 20 years ago and didn't replace copper (which was 10Mbit/sec thicknet - try pulling a few hundred feet of *that*) then. It won't replace copper now, especially since copper is even more convenient, durable, and cross-compatible. They both will continue to be used where appropriate.
And you're sooo sexy with that coil of ethernet slung over your shoulder like you're Tarzan or something...<giggle>
Although, I suppose it should be theoretically possible to create a standard that uses just one universal ground wire for a 600x7 = 4.2Gbps rate...
Uh, last I checked, Ethernet is a balanced signal, there is no ground. This eliminates problems with ground potentials between two distantly seperated devices.
It's basically like this
Pair:
TX+
TX-
Pair:
RX+
RX-
High signal might be +5 and -5 on the other, in relation to some certain ground. There is no single point of reference per se, it's just the difference between the voltages. The same signal may appear to be +7 and -3 at the other side, but it doesn't matter that the ground potential is different, since the difference is the same.
I think GB ethernet does something slightly more complex, but I believe that is a balanced system too. Coax is unbalanced, there is a ground on the sheath, hence you use a Bal-Un (Balun) (balanced to unbalanced) to convert between the two.
Also your post is ignorant in other ways, you think we can only encode one bit per cycle? This is analog we are talking about here, things like QAM let you get several bits per cycle.
I've had enough abrasive sigs. Kittens are cute and fuzzy.
I actually SAW CAT6 on sale at Best Buy yesterday. $25 for 25ft.
It's already commercially available and overpriced!
CompUSA is already advertising the cable in their circular for $25 in 7 feet quantities.
Of course, it's probably going to be cheaper to crimp it yourself, but at over $3 per foot, it's quite expensive.
-- If you try to fail and succeed, which have you done? - Uli's moose
With Gigabit over Cat-5, and 10-Gbit in the works, that's as fast as I'll need to my workstations for a long long time. I certainly wouldn't recomend fiber to the workstation, unless you've got a hard shell on it, and your workstations are anchored to the desks! Fiber is far too fragile.
For the backbone? Go fiber, of course!
And servers? If you don't need the extremely high bandwidth, distances and reliability of fiber to your servers (or don't like the price tag), don't hesitate about going with copper.
One thing I like to note about fiber v. copper...
You need to get new copper cabling almost everytime the speed of the network increases. With fiber, the fiber doesn't change, just the lasers/LEDs at the ends. So, fiber is nice in that regard, but the fiber NICs/modules are still quite a bit more expensive than copper equivalent... Then again, more hubs and switches are needed with copper than fiber, so you save a little money that way, if you go with fiber.
The advantages and disadvantages of each can even out. I'd say go with copper anywhere fiber is too fragile, and, if you aren't scared by the initial costs, go with fiber everywhere else.
I had this company trying to sell me a Cat6 or Cat7 install. When I busted him by reminding him that there was no such spec yet, He fumbled, insisting they'd comply to the standard which would be out "any day now." I told him if he ever tried to sell me a line of shit like that again, I'd thow him and his installers out immediately. This was 1996. I should track that dork down.
I can't imagine fiber being too much worse
Until you get a piece of glass in your heart and die.
"Plastic".
Sure, plastic fiber isn't optically as good as glass, but it's good enough for some things. The new 1394B spec, as I recall, goes to 3.2 Gbit over up to 50 meters of plastic fiber. And it's a lot less fragile than glass fiber.
Plastic fiber to the workstation seems eminently practical.
-- Alastair
I figure there's no reason why we can't figure out something more complex, like a 1-wire reference.
I think more pairs would be an interesting way to go, and still preserve the EFI immunity that TP offers. Just imagine a 50Gbit link made up of 5 cables running at 10Gbit, all transparent from software, implemented on the hardware level.
Of course we have other hurdles to go before we need to worry about that, the PCI bus is running out of steam even for regular Gbit ethernet. 64bit PCI is backward compatible, but 64bit 66Mhz is not, and even then you are still only getting a 4X speed increase.
Maybe someone will come out with an AGP network card, since that port is probably the highest bandwidth expansion interface a lot of computers have these days, and there is no need for a server to have a 3D card in it.
I've had enough abrasive sigs. Kittens are cute and fuzzy.
>Higher frequency signals/noises travel easier and farther.
No the opposite actually. The paralell capacitance and the serial impedace together for a low pass filter. So it's harder to send a high frequency signal down a wire.(this is why thay must improve the cable to do so) The bandwidth(efectivly the highest frequency that can be sent) however is nessesery for sending at a high bitrate. Imagine sending 101010101010101010 down the line. That whould basically create a square wave. The fundamental(the lowest frequency in a signal) whould be half the bitrate, the first(3rd. actually) harmonic whould be 3 times the fundamental and 1.5 times the bitrate with one third the voltage of the fundamental (such is the tao of the square wave) and so on. But not all of the harmonics are needed to get the data across. So in the end you get approx. the bitrate of the bandwidth(often bw. is used to mean bitrate) but it depends on the encoding method. This is why cables that can handle higher frequencies are needed.
FRA: STFU GTFO
ummm... no. On ethernet, there is no carrier signal, and the higher the frequency, the worse the propagation though cables, at 100 MHz the primary culprit is skin-effect, as you go higher, the primary source of loss is dielectric losses (like above 1GHz).
Twisted pair is pretty good for transmitting high frequencies, but not great- I've measured cat-5 cable and about 130m of cable gives you about 10dB of loss at 16 MHz. The advantage of cat-5 over coax is that it is much cheaper and easier to terminate. You want to use higher frequency cables because it lets you get faster rise-times on your signal, so you can stuff more signal changes/second.
For 100 Mbit ethernet, the *baud* (symbol) rate is 125 MBaud/s (and hence 125 MHz bandwidth)- it uses an encoding called 4B/5B to encode the clock into the data and only waste about 1/5th of the bandwidth. The *bit* rate is just 100 Mbits/s. 10 Mbit uses a much lossier way of encoding the clock and data, which sends about 20MHz of bandwidth down the wire. Gigabit Ethernet is also 125 Mbaud/s, but each symbol encodes more than one bit (it has 5 level signalling) and 4 channels (pairs) to transmit on. Plus they use 8B/10B encoding. Since it is still 125Mbaud/s, it still is within the same bandwidth as what 100 Mbit uses. It just uses it more efficiently. Of course, you don't get something for nothing, so you lose sensitivity in your reciever. The major gigabit ethernet PHY manufacturers all use DSP cores in their parts to achieve gigabit speeds.
I remember having to wire something up when the power went out (no, not network cabling, more mundane stuff). Well, when the soldering iron got too cold to work anymore (no, I didn't have a battery powered one -- they weren't decent in those days), you start stripping the cables as usual, twist them, wrap them in solder, and use a match to secure the connection. A temporary hack, to be sure, but it worked for as long as it had to.
I will say, that if you plan to do a lot of this, (and "a lot" can be "as little" as retrofitting structured wiring in a house"), get the proper tools: a Greenlee punch down tool for jacks and headend (usually comes with either a 66 or 110 blade -- you want the 110 but it's worth paying the US$15 or so for the other) at about US$45, a hand crimper for RJ45/RJ11/RJ14 (usually comes with a bunch of plugs) at about US$20, a coax wire stripper with RG6 and R59 settings at under US$10, and a decent RG6/RG59 coax crimper: around US$20. Surprisingly. Home Depot has all this stuff, including plugs, structured wallplates and jacks, Cat5e cable, etc. (Having the coax stuff is, less surprising). BTW, crimping cables, particularly RG6 coax connectors is hard on the hands -- do get a good tool.
I retrofitted structured wiring to a house I bought a year ago. (You don't want to do this: putzing around in the attic, drilling through non-load bearing top-plates is double plus not fun -- I hired a guy who had network experience and did residential "cable" and "phone" cabling, but only had him help tie-wrap and pull cable -- it was stilla lot of work and definately a two-person job.)
I pulled two Cat5e ant two RG6 cables to six drops, plus an attic "subdistribution area" (existing cable and telco drops terminated up there) from a headend which received the DSL line, POTS, dual LNBs pointed at two satellites, and a terrestrial SD/HD/analog TV antenna in the attic. There are breakout panels in the headend. So, that's 14 Cat5e jack terminations (headend side is punched down to 110 blocks), and 28 coax terminations, just for primary cabling. Then there's end-cables to crimp, terminating satellite lead-in (8 more coax connectors: one each end of four cables), satellite cross-connect cables (8 more!), and break-out panel to multiswitch cables (yet another 8). 7 cables (14 more coax connectors!) go from the multiswitch to the coax breakout panels. 7 Cat5E jumpers (14 RJ45 crimps) run from the firewall/router to the Cat5e breakout panel, and 7 punched down jacks on that panel to the 110 blocks. There are some odds and ends (line power inserters for the attic-located terrestrial antenna amp) as well. Oh, and if you do this, you will be making jack extention cables (two coax, two Cat5e, around 100 feet long), with four coax and four Cat5e crimps, for testing back to the headend when you suspect the cabling to a jack.
The bottom line is that if you wire, retrofit structured wiring in a home, you will crimp and punch down so much, by the time you're done, you will be an expert. One upside is that you will almost never buy pre-made cables again: you'll just make your own, to length, as required. Oh, and if you run two cables, do get two spools, or you will go crazy running a cable, going back, running another, and so on. Yes, this means you will have two spools of leftover. Save it to make patch cables.
In my case, I bought 2000 feet of Cat5e and 2000 feet of RG6 (the guys at Home Depot thought I was nuts, and BTW, RG6 on the spool gets heavy fast), and ended up using around 1500 feet of each in a 3200 square foot house. I got headend enclosures, patch panels, a multiswitch, diplexers, and misc. stuff from Home Tech and satellite gear from American Satellite.
You could've hired me.
A few years ago I purchased 1000' of CAT 5e rated at 350MHz for around $50 to install in my house. I guess the cable is at least as good as CAT6. It's a real pain in the butt for crimping connectors, though, since each twisted pair is bonded together (like a 2 conductor ribbon cable) and must be separated.
-Aaron
This post is encrypted twice with ROT-13. Documenting or attempting to crack this encryption is illegal.
Except this is several orders of magnitude different. Most officeworkers today don't even really need a 10Mbps connection, let alone an 100Mbps or 1Gbps (a very large percentage just browse the web and send email). So saying that 1Gbps will be enough for the forseeable future would be like saying in 1980 that 8 megs of RAM would be enough for the forseeable future -- and it was.
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
Even without the super fancy compression,
45Mbps is still sufficient for a Hi-def video stream. It's pretty hard to watch more than one
video stream at a time, so most people won't even need over 100 Mbps.
But although 100 Mbps may be all people need, they will still want more.
Once bandwidth is cheap enough, we'll all keep everything on our personal servers.
P2P piracy will take 30 seconds at Gig-e speeds, for a single film.
Sharing my entire library would take hours.
-- this is not a
I think 4x and 12x Infiniband use multi-fiber cables. In practice, everyone uses copper for Infiniband, though.
Aside from the problem of exposure to elements, this might be fine for a single Cat5 cable dropped from an upstairs room to a downstairs one, but running two Cat5 and two RG6 cables makes for a messy bundle. Even the combined cable you can get is 3/4" thick. A single RG6 cable is about 1/4" thick. To do this right, you really have to run a steel conduit on the outside (one inch diameter to accomodate the cable -- this gets visible and ugly), and drill through the wall.
Why the, err, fuss?
Neatness counts. Maybe not for you, but it does for me.
If you are worried about looks, go wireless, has been around for quite some time, though I can understand you not going that route if it was not accessible when you where doing things.
Wireless is still expensive, and puts everyone on the same (or a few) contention-based networks. I like my 10/100 Mb/s switch. Still, it is a good idea for mobile devices and the odd room that you can't reach by pulling a cable.
Finally, you want something that adds value to the home. I figure my $1000 investment and sweat equity added $5k to $10k to the resale value of the home.
You could've hired me.
I just had my new house done in Cat 5e. Now I'll have to do it all again to stay 1337
If my call is important, why am I talking to a recording?
I have not seen a 48-port GigE switch either, though Extreme does make a 32-port GigE switch (total of 32 10/100/1000BaseTX copper or 1000BaseSX fiber ports).
s /summit7i.asp
http://www.extremenetworks.com/products/datasheet
64 Gbps switching fabric and 48 Mpps switching performance... you're not going to find this at CompUSA and I think the list price would scare me!
I hope someone with gobs of real ethernet experience and know-how can post some meaty tech specs and info for the 99% of those of us here that only know the basics. I've run Cat5, I have a couple switches and an old hub. I know that Cat5e and Cat6 are better than Cat 5. But that's about all I know, aside from a few brief searches on Google for some gigabit info. Perhaps someone can clear up some of the many misconceptions, myths, and marketing BS by explaining the current state of ethernet in terms the average Slashdotter can understand and respect!?
You make patch cables out of the same cable that you run through your house? While you might have gotten lots of experience crimping cables in doing your house, you obviously didn't do very much research. If you are going to lay semi-permanent cable through your house, you should be using solid-cored cable. First of all, it is less expensive than stranded cable, which is important since you're using it for long lengths. Secondly, the keystone jacks and punch-down blocks are designed to cut through the insulation into SOLID cable. On the other hand, the knives in the crimps for patch cable are designed to cut down into STRANDED cable. If you try to use one type of cable for the other's purpose, you can end up with poor connections at the crimps and the punch-down blocks.
When I wire offices, I always make sure I have a spool of each type of wire.
Though, I have found that using solid core cable with RJ45 plugs designed for stranded cable works fine: the biggest issue is flexibility of the cable vs. that of stranded.
You could've hired me.
So, if that's what a homeowner wants, it's either gonna cost for a custom job, or you'll suffer with a retrofit. (No, the builder generally won't let you on the site after the framing is complete, but before the sheetrock goes up -- though I've known of, he he, "exceptions").
In general, when home prices dip, things like that prevent them from dipping as much as other comparables, though in a rising market they neither add or subtract anything.
You could've hired me.
Right now, only two: my Athlon XP1600+ and an old Pentium 200 Mhz acting as a bulk data server (1/4 terabyte of uncompressed music). But there are plans for computers in the kids rooms, an STB-type system in the family room (so I can route music there), and possibly systems in the game room and spare bedroom.
There is an HDTV terrestrial/satellite receiver in the family room, and a plain Jane satellite receiver in one kid's bedroom. Another one is slated for the other kid's bedroom when he gets old enough.
The point is, if you're going to wire, you may as well get it all done and over with at one shot.
Wow, just did a search for RG6, never knew there where so many types of Coax. . . . LOL. I've had such crappy experience with Coax I've pretty much given up on it, you know what the friggin quality loss on that stuff is after the first 50 feet?
Loss varies with cable, distance, and frequency: higher frequencies are attenuated faster. That's what RF amps and tilt compensators are for. RG6 is sctually pretty good compared to run-of-the-mill RG59.
I said, "Neatness counts. Maybe not for you, but it does for me."
I didn't have much of an option, either that or routing it through a story of drywall, which was, err, put up by my friggin drunken grand father (gee thanks ga'pops) Yeesh. Boards spaced pretty much randomly throughout. . . . ick.
Any kind of retrofit, particularly between stories (as opposed to attic to top floor and basement/crawl to bottom floor) is going to be a pain... and older homes have fire breaks between the studs (what fun!).
Oh, and to refer back to the first line of your reply;
exposure to the elements? Well besides perhaps acting as a bit of a mini-lighting rod, not much has happened actually. Kind of surprising considering the wind and rain storms that happen around here, but hey, it keeps on working!
That's pretty good, then.
Ok so granted at one point I was pinging 200 to a computer on my own LAN but. . . . hehe. ^_^ (now it is down to 25 or so)
I see ping times of around 0.367 ms.
You could've hired me.
I see. There's a practical reason why it's unoptimal to do this, though.
If you look at the knives in the ends of the RJ45, the ones that cut through the insulation, you'll see that for each conductor, it's a single knife that cuts through the middle of the wire.
With stranded cable, this works perfectly, since the wire is soft and it cuts right into it.
With solid cable, the knives rarely cut directly INTO the table, but instead just slice into the insulation and fall to the SIDE of the conductor. This leads to a relatively fragile electrical connection. The ends also fall off much more frequently.
If you see how the knives are designed in 110 patches, you'll see that there are TWO of them, forming a v-shape that the conductor slides down into, cutting in on two sides, and wedging it there. Quite clever. Stranded cable has a tendency to be sliced completely in two by these knives, though.
Ethernet does NOT saturate at 60%. With switched full duplex I regularly acheive >95% of theoretical limit over udp.
Maybe back in the day of co-ax or half duplex hubs, but modern ethernet is a differnet story...
Price fiber vs copper installations with connectors and labor. The defacto standard Siecor Unicam connectors used by most fiber installers run around $9/each. For a large install, you're looking at a huge cost increase. For desktop installs with two ports, that's $72 just for connectors for one station! Panduit Cat 5E jacks on the other hand are about $3.50 each, so $14 for two ports in this case. Labor costs are also higher since it takes time to strip, clean, and terminate .. even with the Unicams. That isn't counting fiber trays and the higher cable costs for fiber vs copper. It just isn't practical for most companies except between closets or for special applications.
Jason
"FORMAT C:" - Kills bugs dead!
You're right about using up the bandwidth though - 10Mbps, 100Mbps, 1Gigabit... no matter how much I have, I want more.
-- this is not a
According to this website, Cat5e is 350MHz and designed for gigabit ethernet. What am I missing? Is Cat6 or 7 needed for gigabit?
Er, the standard ping that comes with Red Hat Linux 7.2.
You could've hired me.
Your point is noted, but I haven't had trouble. Fortunately, bad patch cords are easily replaced.
You could've hired me.
Given the wattanted lifetime of a good quality structured wiring installation is 15 years, and the bulk of the cost is the labour, rather than the components, it's a reasonable gamble to pay a small premium now to reduce the risk that you may need to prematurely rewire the entire building...
And the context was a home network. With one switch. It's going to be awhile before that thing feels slow.
"We returned the General to El Salvador, or maybe Guatemala, it's difficult to tell from 10,000 feet"
everyone tring to download a map off of someone will kill the bw for a little while, getting patches from someone, as patches is getting over 100 megs and 30-40 people getting it takes forever
You can handle this with a switch that's got a Gig-E port. Nobody else needs >100 Mbps.
"We returned the General to El Salvador, or maybe Guatemala, it's difficult to tell from 10,000 feet"
ethernet however saturates at 60% so you can only get real transfer of about 60 Mbps
At 10Mbps with a 3C509B (nice card if 10Mb ISA is your bag), I get actual transfer rates of ~1.14MB/s (1,200,000bytes/sec (96%)). I guess the rest is protocol overhead.
On my 100Mbps connections (iBook OS X - PII300 OpenBSD) I get 9.8MB/s with ftp transfers.
Where did you get this 60% number?
War crimes, torture, lies, illegal spying... Would someone give Bush a blowjob, already, so he can be impeached?
What doy you need >100Mbps for, anyway? Until you have an answer, just leave it alone.
My old seagate drives sustain 17MB/s on their own, however they're in RAID-0 so they do a fair bit more than that. New IDE drives typicaly do around 25MB/s+.
So if you want to copy data from one machine to another, why not get rid of the 10MB/s bottleneck if the price is worthwhile? 1000Mbit is getting pretty cheap now.
Until I have an answer? A full 650MB CD takes 66 seconds over 100Mb and about 7 seconds over gig. There's an answer. ; )
War crimes, torture, lies, illegal spying... Would someone give Bush a blowjob, already, so he can be impeached?
Er, sounds like you're describing Etherchannel or a Multi-Link Trunk with those 5 cables. Why would you want to make a new cable standard with 40 pairs in it instead of just running five separate runs using standard 8 pair cables?
No, I think you're wrong. 100Mbit is sent over a copper cable at 125MHz, not 200MHz. I think the original poster was right.
The 60% number is based on the wait times after a collision on a non-switched ethernet segment. After a certain usage you collide so often that the waits slow your usage down.
*Not a Sermon, Just a Thought
*/
Hmm, FreeBSD4.0 gives me 3 decimal places too. Of course in both the FreeBSD and GNU/Linux case, I see three decimal places when the ping time is under one milisecond.
You could've hired me.
No, I think you're wrong. 100Mbit is sent over a copper cable at 125MHz, not 200MHz. I think the original poster was right.
I'm not sure what you are disagreeing with, but I'm right. I don't know where you got the 200MHz number from. 100 Mbit has an 8ns symbol clock- or 125 MHz. What I'm absolutely sure about is that it is not modulated, there is no carrier- 100Mbit ethernet is sent as baseband, +1, 0, and -1V signal levels. I look at 100 Mbit and gigabit every day- if you want to know for who, it should be pretty obvious if you look at my web-page.