2

I've just cabled my house (Yay!). When using iperf I notice that I don't ever get a speed over about 500 Mbits/s. I know that lots of older consumer grade gigabit ethernet cards won't actually do full gigabit ethernet, I'd just like to check that this drop in speed couldn't be caused by a wiring fault. Could it?

Mike Pennington
  • 29,989
  • 11
  • 79
  • 152
JoeS
  • 131
  • 2
  • It's very hard to tell, expecially if you don't explain your network setup. Anyway, if you run linux, you should check layer 2 errors using ethtool ( ethtool -S eth0 ). – Marco Marzetti Jun 30 '13 at 02:00
  • I am sorry, but home networking questions are off-topic... have you seen [su]? – Mike Pennington Jun 30 '13 at 02:06
  • 1
    iperf is very dependent on the machines at both ends. Look at the cpu usage on both machines while running the tests. – Ricky Jul 01 '13 at 02:13
  • 1
    @MikePennington, while normally I tend to agree with you on these things, the question/answer would easily be a valid question with my edit and there was no need for the rollback. The accepted answer of check interface stats also indicates non-home typical gear. Don't like the answers provided, then add your own or explain why you feel the rollback needs to occur, but I don't see it. – YLearn Jul 08 '13 at 03:12
  • 2
    @YLearn, your edit is invalid since it substantially changes the context of both the question and answer – Mike Pennington Jul 08 '13 at 03:15
  • 1
    @MikePennington, I disagree. My edit removes the setting and leaves a valid question and should be answered as such. If a home user does not have the tools/resources to utilize the answer, then so be it as that is often the case here. As for the answer, I would use the interface stats as my first indicator on a live connection, and again, consumer gear typically does not provide this info. The acceptance of this answer puts this in a context that indicates enterprise or at least small business gear. – YLearn Jul 08 '13 at 13:40
  • @MikePennington YLearn's edit is perfectly valid. It removes some of the grey area which caused you to interpret this question as a home networking question, while this question is undoubtably relevant in an enterprise network. Removing the context of the question makes it applicable in the environment which is on-topic on this site, and does not alter the question at all. – Libbux Jul 08 '13 at 20:38
  • 2
    @TheLibbster, I'm sorry but that's complete nonsense... if the answer doesn't match the question, you don't modify the question unless the OP changed their mind; however, the OP didn't suddenly go to work and run cabling himself. The point is that in an Enterprise network the accepted answer is still not the right way to find the problem. The right answer is to put real cable testing equipment on the cable. – Mike Pennington Jul 08 '13 at 21:02
  • 2
    @MikePennington, so your assertion is that looking at the interface counters for CRC errors is not a valid means of determining if you may have a cable issue? Again, if you don't like the answer, provide your own (and yes, I do agree with you that a cable tester should be part of the answer). Shall I point out other questions here (or other SE sites) where the chosen correct answer is indeed not the best answer? For Example – YLearn Jul 08 '13 at 23:21
  • I created a meta discussion about this question - I'm not sure if it's 100% clear, I'm a little tired - but this comment thread should continue over there. @YLearn – Libbux Jul 09 '13 at 05:08
  • CC above @MikePennington – Libbux Jul 09 '13 at 05:09
  • 1
    @YLearn, my assertion is that you are making invalid edits in an attempt to make home-networking question fit the site. I will not answer the question because it's a home networking issue and this is not a home-networking site. – Mike Pennington Jul 09 '13 at 11:38

2 Answers2

4

The way to determine this is to examine interface statistics for FEC failures.

Generally speaking, 1000GBASE-T is relatively tolerant of wiring that's a little noisy, but you will see CRC32 errors - if you're getting these, especially when transferring at however close to full-whack you can get, then yes, it's likely the wiring. Otherwise, nope.

Olipro
  • 2,202
  • 12
  • 23
  • 1
    I would argue that the best way he could check for cable faults is to get a real cable quality tester... then again, it is a home user and that is extremely unlikely... thus this whole discussion is off topic – Mike Pennington Jun 30 '13 at 10:17
  • 2
    I disagree that it's offtopic - a business might also have legacy (read: crap) equipment. Using a cable tester obviously has the prerequisite of unplugging the connection, which for most production environments, is generally going to be unacceptable as a first-line thing to do. Checking interface statistics is non-invasive. – Olipro Jun 30 '13 at 10:51
  • 1
    please read the OP's question... I've just network cabled my house (Yay!)... home networking is off-topic – Mike Pennington Jun 30 '13 at 11:09
  • 3
    So basically, if he deletes that part and makes up some bull about it actually being in a business office, it's acceptable.

    The suitability of a question should have nothing to do with whether or not the user is doing it at home or in a business. Someone setting up BGP in their home across VPN tunnels is just as legitimate as someone doing it for a Fortune 500.

    This commentary is now going offtopic, start a meta question if you want to take this further.

     – Olipro Jun 30 '13 at 12:17
  • 2
    if the OP deletes the home networking portion, the answer to his question is "ask your cabling vendor to test the cabling", or "get your fluke cable tester"... the question and answers are completely different. Your assertion that nothing changes about home networking questions is quite empty – Mike Pennington Jun 30 '13 at 12:44
  • 1
    @MikePennington "ask your cabling vendor to test the cabling" or "get your fluke cable tester" would not be terribly helpful answers. Whether this is in a home network or in a datacenter is irrelevant, you could give the same answer to a home networking question: "ask the electrician to check the wiring" - I don't know about the OP, or anyone else for that matter, but I don't see that as an acceptable, helpful, or re-usable answer. It would be legitimate grounds for question deletion:

    "not even a partial answer to the actual question" (from help/deleted-answers)

     – Libbux Jul 09 '13 at 04:32
  • 1
    @TheLibbster, obviously the content of the answer is expanded in a real answer... the problem with this question is a) it's an trivial question to resolve in a professional setting, and b) most professionals know what to do. We can disagree about the helpfulness of that answer. If you guys what a different question, then you should ask a different question instead of wishfully rewriting this one. – Mike Pennington Jul 09 '13 at 10:08
0

Check that the link is auto negotiated to the right speeds (gbit and full), and at both ends. I had a vender install cables recently and they Fluke validated all of the runs. I am not sure what happened, but one of the links auto negotiated at 10mbit full. Asked the vender to check it out, and he was surprised that it had any link at all. Not sure how that slipped by the Fluke, but he punched it down again and it has been fine since.

WMIF
  • 151
  • 2