[rhn-users] Low bandwith to localhost

[www.hi.is]

I have been trying to figure out some bandwidth problems, to one of our
servers, a quad CPU IBM X260 machine running RedHat Enterprise Linux
Adwanced Server, with ample ressources and a Gig ethernet. 
Backtracing the problem I narrowed it down to low bandwith to localhost.

The test was flood pinging one thousand 64kB packets to localhost and
time the command: This is the test:

time ping -f -c1000 -s65507 127.0.0.1 

On the IBM X260 the test takes about 12 to 15 seconds to complete. 
 That seems unlikely on this machine so I asked my collegaues to run the
test on their desktop machines, running Fedora Core 4 and 5
respectively, and they got approximately the same results. I ran the
test on my desktop, running Fedora Core 5 and got approximately 50 times
better bandwith. The result was about 0.2 to 0.3 seconds. Amazed at this
I tested also on an AMD 2000+ running Debian Sarge and the result was
about 0.9 seconds to complete. 

All the tests were performed on almost ilde machines, so it is not
caused by load. 

Pinging the external IP address gives approximately the same result.
Pinging those machines from another machines also gives approximately
the same result. 

So the net conclusion is that it takes 12 to 15 seconds to send about 64
MB of data. That is a poor performance for this powerful machine. 
Does anybody here know a trick or two to improve the bandwith? 

-- 
Kindest Regards, Anna Jonna Ármannsdóttir,
Unix System Aministration, Computing Services, 
University of Iceland.