Asymetric network speed question

David G. Miller dave at
Wed Jul 5 17:21:36 UTC 2006

Tim <ignored_mailbox at> wrote:

> On Tue, 2006-07-04 at 22:46 -0600, David G. Miller wrote:
>>> Could a bad cable cause this or is it the skge module on bend (spindle
>>> and mutilate both run CentOS 4.3 with the sk98lin module and iperf
>>> says the speed is about the same in both directions between them)?  A
>>> bad cable seems odd given the speed is fine in one direction.
>It is possible to be a cabling issue.  You could have a problem with its
>wiring for data in one direction (there's two pairs for each direction).
>Obviously you don't want to pull more cable through, but you can test
>with loose wire through the room between the machines.  If that acts
>differently, try reterminating your other wire, one end at a time.  I'd
>still suggest testing with another cable rather than just reterminating
>it.  You eliminate the cable, itself, as a fault, that way.
>Swapping a NIC might be in order, too.
Did some patch cable swapping and the problem stays with the 
workstation.  I even put one of the "slow in one direction" patch cables 
between the server and the wall jack and I get symmetric speed between 
the server and the old PIII/733 box.  I also swapped which wall jack the 
workstation is plugged into to see if its the wire inside the wall.  No 
difference in behavior. 

I guess next thing is to try a different NIC in the workstation and see 
if its the NIC.  The NICs were surplussed to me from where I used to 
work so I can't complain too much if I have trouble with them.  They 
probably wouldn't have gotten rid of them unless there was a reason.

BTW, gigabit uses all four wire pairs in the cable.  Just something to 
keep in mind.  Lots of folks have gotten used to 100BASE-T only using 
two pairs.


Politics, n. Strife of interests masquerading as a contest of principles.
-- Ambrose Bierce

More information about the users mailing list