Re: 100Mbs ether cards talking @ 10Mbs ...
> The only way to really measure the speed of the cards (and not of hard
> disk, nfs-subsystem etc) is to send _huge_ amounts of data really,
> really fast.
>
> Personally I like netcat (it´s debianized) for that sort of stuff, just
> dd if=/dev/null | netcat 1.2.3.4
> and, on the other side
> netcat -listen >/dev/null
> (I´m not sure about the exact synta, but it´s easy enough).
thanks, I used it, and it gave some new info about the problem. I run
the following tests:
BigFile: ~270Mb
Server$ cat BigFile > /dev/null
->Time: 24 seconds
Server$ netcat DiskLessStation < BigFile
Diskless Station$ netcat -l > /dev/null
->Time: 29 seconds (almost 10Mbytes/s =~ 100Mbits/s, great!)
Diskless Station $ cat BigFile > /dev/null
->Time: 227 seconds
(Diskless Station using NFS, v3 I guess -- I compiled it in in both
kernels. Options: wsize=8192,rsize=8192)
So clearly the problem is somewhere in/around the NFS. What is strange
is that NFS is not even using all bandwith (as I can see from eth0
bar in the gkrellm -> a monitor program). And the local disk is also
not the problem, it could go much faster ... CPU for procesing the
protocol seems not the be the problem either (around 10% usage while
copying). Maybe I should whip my computers so they stop being lazy and
start working harder ? :) Any other hints ?
:-)
jan
ps.: BTW, even with a slower network, compiling/running the project
I'm working with in parallel has made it much faster :) ("make --jobs 2"
is really cool)
>
> >am I forgetting to set something ? I think maybe this is a too
> >specific question for this list (sorry), so any pointers to other
> >lists/documents are welcome.
>
> If the cards are autosensing there´s a good possibility that they
> negotiated 10 mbit/s, most cards come with a DOS-driver where you can
> explicitly set spped&duplex (they memorize that, so you need only do
> it once).
>
> cheers,
> &rw
> --
> / Ing. Robert Waldner | Network Engineer | T: +43 1 89933 F: x533 \
> \ <Waldner@KPNQwest.at> | KPNQwest/AT | Diefenbachg. 35, A-1150 /
>
>
>
Reply to: