While i can use the full speed of our connection (100MBit) between my two Gentoo machines it falls down to 0.1% of the speed if i try to connect to a Debian or Mandrake system. Debian and Mandrake are able to speak with each other at normal speed. This problem occured in smbfs & ssh connections. I checked it as well with netperf: Recv Send Send Socket Socket Message Elapsed Size Size Size Time Throughput bytes bytes bytes secs. 10^6bits/sec 87380 16384 16384 11.27 0.06 And to the other gentoo machine: 87380 16384 16384 10.01 93.94 I'd be glad to help out finding the solution to this bug, but it is quite strange. Even shutting down the firewall did not help. Reproducible: Always Steps to Reproduce: 1.scp a file to another machine 2.try to copy files over smbfs mounted filesystem
Have you try to change physical things like: changing tp-cabel, port in switch and nic? Just personal experience where you get some traffic through, but not bursts.
Are these machines on the same subnet? Same switch/hub? Are they all using good quality Cat5 cables? Any of these factors could be affecting your network throughput, not just the operating systems. For example, if your Gentoo boxes are on a different switch than the Debian and Mandrake ones, there could very well be some slowdown related to poor/old wiring. Also, if some are behind a NAT, slowdown will occur as well. You should also check to be sure you don't have any odd policies set up on one or the other set of machines (some sorta proxy/firewall may be causing trouble).
are you doing any tweaking of TCP parameters? in some cases the parameters can cause very bad performance when they are communicating with other machines that can't match the params.
I tried changing physical things: no result The machines are on the same switch. We checked the TCP parameters. This was a Gentoo installation without any modifications. The weird thing is that Debian and Mandrake cause the problems, if it was only one of them i'd say its their fault. This way it seems like if something goes wrong with the Gentoo machine i have. And the problem occurs on both of my Gentoo machines (one amd64, one x86)
Closing NEEDINFO, until/if there's enough information to reproduce the problem.