A couple of things.
1. If the servers are near each other, you might want to test with a direct patch between them. (Gigabit ports are auto-MDIX so straight-through or cross-over will do)
2. Use netstat -i -c and dstat to give you an idea if losses are due to CPU, physical errors, etc.
3. I use the "iperf" utility as nice way to generate and measure throughput
4. I really would have second thoughts on using a non-managed switch (i;m guessing the netgear is). A managed switch will enable you to see stats about each port and help you to isolate where the problem might be
Martin
On 8/29/06,
Raphael Kraus <[EMAIL PROTECTED]> wrote:
> Also, I've seen problems with the Realteks in 2.6.9 kernels, that
seemed to go away with later 2.6 kernels (around 2.6.13, from memory).
> Given the debian box is 2.6.8, that might be an issue, so you might
want to try a later kernel.
That may be a good point, however as the machine is off-site, rather
critical, uses some compiled software and eth0 is less important (i.e.
it's our internal network - I'm more worried about large backups not
taking forever to transfer), I'm not too keen to upgrade the kernel just
at the moment :^/
I'll discuss this with my colleage and we'll see if we can organise to
give it a crack...
Thanks heaps for all your help!
Regards,
Raphael Kraus
Software Developer
[EMAIL PROTECTED]
02 8306 0007 Direct Line
02 8306 0077 Sales | 02 8306 0099 Fax
02 8306 0088 Support
02 8306 0055 Administration
1300 13 WILD (9453) National | 1300 88 WILD (9453) Fax
--
SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/
Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
--
Regards, Martin
Martin Visser
-- SLUG - Sydney Linux User's Group Mailing List - http://slug.org.au/ Subscription info and FAQs: http://slug.org.au/faq/mailinglists.html
