Slashdot Mirror
Should Network Cables Be Replaced?
Jyms writes "As technology changes, so hubs routers and switches are upgraded, but does the cabling need replacing, and if so, how often? Coax gave way to CAT 5 and CAT 5e replaced that. If you are running a 100Mbit/s network on old CAT 5, can that affect performance? Do CAT 5(e) cables get old?"
Just like any cable, cables will break. So, yes, they do get old.
Also, there is cat6 cables out with better specs and can handle at least up 10gb/sec.
I'm not sure what would "get old" exactly. It's insulated copper, so I think it should be good so long as they aren't damaged. If anyone knows better, feel free to correct me.
If you want to be sure, though, test them. Transfer files over your network. If the connection is bad, you can try replacing the cable and see if that works. But the fact that Cat6 is out doesn't mean you have to rush out and replace all your CAT5e cables, especially if you're only dealing with normal 100mbps connections. But I use CAT5e for 1gbps connections, and that seems to work fine.
Cables don't get "old" by themselves, but they might have been installed incorrectly from the start (too tight bending, swapped pairs/cables, twisted pairs separated for a longer distance, shields not connected properly, grounding done wrong). Furthermore mechanical stress (too much work being done on a patch panel over the course of several years, cables pulled hard while moving racks, ...) might have damaged parts of the cabling.
To cut a long story short: Properly done CAT5 should be good enough for Gigabit, but often what's called CAT5 works well for 100 Mbit networks even though it doesn't meet the specs.
Get a decent LAN tester (not just two computers, using "ping") that prints out attenuation, crosstalk and all the other things... and preferably tells you what "category" your cabling still is compatible with. Replace all the stuff that's out of spec. Then you have hard numbers you can rely on should you ever ponder if your local network infrastructure can handle 100M/1G/10G bit/s. Everything else is guesswork.
While you can link at those speeds with Cat5 you cannot actually get those speeds. Usually it tops out about 200-400mbit for me when I've tried. For most uses that's perfectly fine but in some cases it's not like my entire graphics and video editing departments. Servers are all connected with Cat6 if they use a lot of bandwidth.
I ran into this problem in Vegas as the place only had Cat5 connecting all the rooms to their closets so I had to use LACP trunking to get my bandwidth up.
Are you sure it's not your NICS/Servers/Switches? I remember seeing 250-400Mbit over Gigabit Ethernet a few years ago on older machines. Newer stuff particularly servers seems to be able to get close to a full Gigabit over the same cabling.
I suspect the bottleneck is actually the ability to deliver data and not the cabling.
Jason.
5 and 5e are only rated for 100MHz per pair, so although you can get link speed aggregated at 1000bT, your max throughput due to crosstalk, signal reflection, and EMI is going to limit your switch to a speed closer to 300-400Mbit. Many switches will detect Cat-5 issues and downgrade your link connection automatically on problematic runs.
Connector quality has more to do with the connection quality than the cable itself. 5e simply has tighter specs to maintain. Really, there's not much of a difference, especially is you're using good patch panels.
Cat 6 runs 250MHz per pair, tru gigabit speeds are supported.
10G over copper is most commonly limited to 15M, and requires special 4 lane copper cabling, not Cat6 cabling. It's similar to Infiniband in design. A Cat 6 option was later offered, though few companies support this format. It's limited to 66m, and suffers similar bandwidth issues due to signal quality that running Gig-e over Cat-5 exhibits. Cat6a cabling can be used for 100m 10G deployments. Note this requires 650MHz Cat 6 cable ends, not 250MHz cat 6 cable ends as are normally deployed, for which there is a difference, and also requires 10G rated patch panels. Cat 6 cable can come in one of 3 thicknesses (guage). only one of these is commonly reccomended for 10G speeds.
Cables do go bad over time, due either to environmental factors or movement. Exposure to direct sunlight is bad fort cabling. Non-constant temperatures is also a cause of degredation.(cables in plenem space or inside walls tend not to remain at constant temperatures). Oxidation of the copper connector is the most common failure. higher quality cables and patch panels use silver, gold, or other corrosion resistant metals for this reason. Many cables are also made with lower quality plastics that simply fail over time (some are practically designed that way I sometimes feel). When the plastic fails, the cables corrode quickly.
More often I find a switch port fails before a cable (usually because someone plugged something in they should not have, or a charge makes it way into the cable due to being too close to a power cord, or long term exposue to magnetic fields causes elecrical resistance and damages the switch over time.
typically, I'd leave cables in place until a hardware upgrade or data bottleneck justifies the change. ALLWAYS use high quality cables rated for the installation location. lower guage (thicker copper) are generally better, but they should ALLWAYS be within spec. Buy cables from companies that offer 20 year lifetime warranty. (Hitachi, Mowhawk, etc) Have them installed by professionals who back that warranty and use properly rated panels and punch downs and you should have no issues. Anytime you;re running cables, allways run a class of cable 2-3 tiers better than your current needs, and for workstation drops or other complicated runs, allways run spares (the labor typically costs more than the cable, and running 2 or 3 at once costs less than 1 now and 1 later). Use cable trays or hooks EVERYWHERE, never let cables lie on ceiling tiles or underneath floors in channles.
This sounds like overkill, and probaly is for a small business, but when you have 14,000 desks in your copmpany (most with 2 netowrk and 2 phone drops) and over 3,500 servers, labor to replace cabling tallies in the millions of dollars...
There is no contest in life for which the unprepared have the advantage.
Although shielding is nice to have, it's not necessary for network cable because network cable is balanced twisted pair. Indeed, most high-performance network cables are not available with shielding, they can't maintain the spec with a shield in the jacket. It might be that your cable is older.
If you do have shielded cable, don't ground both ends!!! Bring all cables at one end to a common ground, and let the other end float. Otherwise, you will create a ground loop and actually make the noise worse.
Bruce Perens.
Well, oddly enough I did IT work for a plumbing company. One of the things they did was "repipes" of residences. Usually they were done due to recurrence of leaks. Copper pipe corrodes from the inside because of chemicals in the water (like chlorine). Sometimes it was done where there was minimal water flow, when corrosion or sediment didn't wash away, but built up.
So, if you got a new sink because you were changing the rusty faucets and then you found your water flow wasn't improved, you may be a candidate for repiping part or all of your home. It could just be buildup in the shutoff valve too.
But to stay on topic......
I've seen people run network cables over or under their carpet. They start getting degraded service as people walk on it. Yes, those need to be changed. Properly run network cables (in the walls, with no rat infestations to eat the cables) don't generally need to be changed, unless something else happened. And I have seen network cards that handle regular CAT5 at speeds over 100Mb/s with 1000baseTX interfaces. Sometimes you can just reterminate the cables because the "spare" 2 pairs were never wired right, assuming those extra wires are ok (i.e., the installer didn't yank the cable through and bugger it up.
But, realize (to the OP) that there are limitations to GigE on a computer. Most of the time in the real world, you won't see 1000Mb/s. It's dependent on the switching hardware, PCI/PCIe bus, hard drives being read from/to, etc, etc, etc....
Serious? Seriousness is well above my pay grade.
Grounding one end makes the shielding at the ungrounded end an efficient antenna, and can actually increase the radiation from the cables.
Cable shielding isn't a good way to avoid inteference with the signal on the cable, and isn't a good way to avoid radiation by the signal on the cable, unless properly terminated for one specific frequency (or narrow frequency range). One or both ends of the shielding would need to be coupled to case ground by a capacitor chosen for the frequency that you want to shield.
Shielded cable at this frequency is likely to cause more harm than good. There's usually little need to shield against common-mode interference, and I've never heard of shielded cable being used in a TEMPEST set-up (carefully matching impedances is the best way to avoid emissions, creating giant gorund loop antennas is not).
Socialism: a lie told by totalitarians and believed by fools.
Hate to nit-pick, but UTP means UNSHIELDED twisted pair.
So, yeah, how did that shielded unshielded twisted pair work out for ya?
My sig can beat up your sig.
He said FTP, which is also known as S/UTP, screened unshielded twisted pair, or "fully shielded" twisted pair.
The difference between FTP and STP is that in STP, each cable pair is shielded.
FTP has only an overall shield that covers all the pairs (each pair isn't individually shielded).
So FTP _is_ shielded UTP.
5 and 5e are only rated for 100MHz per pair, so although you can get link speed aggregated at 1000bT, your max throughput due to crosstalk, signal reflection, and EMI is going to limit your switch to a speed closer to 300-400Mbit. [...] Cat 6 runs 250MHz per pair, tru gigabit speeds are supported.
That is incorrect. Gigabit Ethernet (1000BASE-T) uses the same bandwidth as Fast Ethernet. The higher speed is achieved by using
1) all four pairs (Fast Ethernet uses two),
2) each pair in both directions (Fast Ethernet uses one pair to send, one to receive), and
3) a more efficient encoding (more bits per baud).
The full Gigabit Ethernet speed is specified for Cat5 cables.
Sure, it's a little like an antenna, insofar as a Faraday cage is.
But grounding the shield at both ends creates ground loops. You might not notice them right away, but you sure will the first time the MOVs in the surge suppressors at one end or the other shunt a spike to ground, and some of that current decides that its preferred path to ground is over your STP Cat5.
Eventually, after you blow up enough switch ports, you'll stop doing it that way.
It's generally pretty bad form to ground both ends of any shielded wire that traverses any real length.
Kid-proof tablet..
Not only can your gear affect your speeds, but conflicts with speed auto-sensing can wreak havoc.
We recently swapped out some older edge switches in one site for a beautiful new CISCO core switch. Within days, the help desk had reports that some users' network performance had gone from fine to terrible. (Not our intent.)
As it turns out, a boatload of older NICs were mishandled by the new switch which downgraded speeds, communicated in half-duplex, and even then continually reset the connection.
Had the users not complained, we never would have known there was a problem.
As it turns out, each port on the core switch can be manually set to a fixed 100mb full duplex (and ignore auto-sensing) which then operates just fine.
So much for plug-n-pray.
Live Long and Prosper - Thanks Leonard. You are missed.
You talk about grounding like folks used to talk about SCSI termination[1].
The reality is far simpler, and there's no goat's blood required: Use shielded wire where it makes a difference. Ground it at one end (or in the middle if that's what the situation dictates, but under no circumstances at more than one point).
And then, have a beer. While imbibing, rejoice in the fact that one's cell phone has ceased ringing with voices proclaiming connection issues.
[1]: SCSI buses should be terminated exactly twice; once at each end of the bus. If you have more than two terminators, they are in the middle of the bus, or your "bus" has more than two ends, you've done it all wrong and need to start over. Same with TV coax, radio coax, 10base2, or any other transmission line medium.
Kid-proof tablet..
Bullshit - you just hired stupid electricians. Probably lowest bid on the job, right?
We let our contracted electricians run our coax, cat5e, 110v, 208v, and 480v - and they did a great job. Even labeled all the ports on the patch panels, each end of the cables, and even used the cable management we installed.
Data wiring companies are just over priced electricians.
I'm sure you meant to use this link for the best price - the comments here will fill you in on the AWESOMENESS of these cables.
Denon AKDL1 Dedicated Link Cable:
http://www.amazon.com/Denon-AKDL1-Dedicated-Link-Cable/dp/B000I1X6PM/
~hylas