The differences in CPU and memory should not make any difference for single
ftp connections. Much more important is the I/O paths. He does have the
most recent and best driver for the 3C905C, which is good. He says he is
using quality fast SCSI drives (but did not specify HA or drivers).
The things I would look at first are connection parameters. He says all
connections are 100/FD, but does everybody agree they are 100/FD? It is
easy on the switch ports to see what they are set at with most modern
switches. Some NICs do negotiate FD, and the switch responds but the card
then goes to HD. A couple of years ago the Intel EE/100 Pro's were
notorious for this. The workaround is to set options in the config to
force FD. So the first thing I would try in this scenario is to use the
driver options to ensure that the NIC is going into FD and then retry the
test. Next step is to see what phase is slow. What about
routing? traceroute output for each test would be nice.
That's a start, but CPU and memory will not be the issue; I can get an
order of magnitude better performance out of the 75MHz Pentium with 64 Mb
RAM under my desk than he reports.
- rick warner
On Thu, 2 Nov 2000, Scott Sharkey wrote:
> Are you kidding? It appears that you have a) 2-3 times as much
> memory in W2K machines, b) Much faster CPU's in W2K machines, and
> c) much more recent software in W2K machine.
>
> So, yes, I would expect different performance among the machines
> <grin>
>
> 40Mbps may in fact be the maximum output for that machine. Don't know.
> The only way to compare would be to now load W2K on that hardware, and
> see if it also get's about 40 Mbps.
>
> There are numerous things you can do to tune Linux, but it's not
> clear that they are necessary at this point.
>
> You're comparing apples and elephants, and trying to make judgements.
> Won't work.
>
> -Scott
_______________________________________________
Redhat-list mailing list
[EMAIL PROTECTED]
https://listman.redhat.com/mailman/listinfo/redhat-list