A bad length of Cat5 leading to a testing query

ED E

Hi all,

One of our machines in the home office was reportedly often slow. I took a quick a look at it a few times and while it was a little slow to resolve pages some times it would speedtest fine using speedtest.net. Today I finally got to take a look properly and did some more testing. This time locally. A 100 meg link between a switch/AP in the home office and the machine in question("COMPUTERONE") was only throughputting 13-15Mbps. Replacing that link with a new cable to test brought it to a healthy 75Mbps, grand, but the next bit is whats got me thinking. On the old link, right after it tested at 11Mb locally speedtest.net read the full 52Mb, with 0ms ping. Obviously wrong with no latency, but Im wondering could error correction or caching con speetest.nets testing payload? Or what would do it?

The problem is solved but I wont be able to let it go unless I know why it was giving those results.

Table of related data attached as a PDF here.

ED-E.

L1011

The speedtest.net server in Enniskillen reports 0ms ping most of the time, if that's what you were using.

Caching should have zero impact on speedtest.net but a cable can be very intermittent in what speed devices can connect act.

srsly78

Cat5 cable uses 4 pairs of wires, this gives some redundancy and helps against interference. So there are 8 wires, but only 4 "data channels" sent over them.

If one of the wires is damaged the fault tolerance is enough to have 10mbps work ok. But not much more than that. I am pretty sure this is what you are seeing. Many years ago I encountered a very similar problem that caused much scratching of heads.

Use a cable-tester to confirm. Completely ignore speed-test tbh.

eddhorse

Agreed, speed test to the internet is hardly a good way to test your LAN.

You should be basic pinging really to check for any dropouts, and you can increase the packet size if its a file transfer type issue you are looking at.

smee again

Speedtest.net is for internet speeds, your device interfaces may be many times faster than your link to the internet. File transfer isn't always a good indicator either, Gigabit speeds are much faster than the read speeds of hard drives

Use this: http://www.totusoft.com/lanspeed.html

ED E

eddhorse wrote: »

Agreed, speed test to the internet is hardly a good way to test your LAN.

You should be basic pinging really to check for any dropouts, and you can increase the packet size if its a file transfer type issue you are looking at.

Maybe I wasnt clear with the question, I wasnt using the speedtest to test the link, I was wondering why the result was good in spite of my testing results. Testing was done with Iperf. Regarding packet size, see below.

smee again wrote: »

Speedtest.net is for internet speeds, your device interfaces may be many times faster than your link to the internet. File transfer isn't always a good indicator either, Gigabit speeds are much faster than the read speeds of hard drives

Use this: http://www.totusoft.com/lanspeed.html

As above, obviously a 50 meg UPC connection is useless for a Lan testing, wasnt the question. Also as above, iperf.


So, image below seems to explain it. The link performs terribly with small windows, but at ~50% with larger ones(64k to 512k to 1M). The other similarly at all sizes. Might try re-crimp the original before ripping it from the trunking. Thanks all.

srsly78

Useless results. Use a cable tester!