From: Steve Biggs <[EMAIL PROTECTED]>
> > But, I am using eepro 10MB on both ws and server.
That's ok, but you will see much better performance at 100MB.
> > You really don't want to know what I am using for
> > some
> > of these wires, just let me say that I have had
> > heavy
> > traffic over them with no problems. They are all
> > less
> > than fifty feet each.
The fact that you can run reliable TCP connections over
the cables is irrelevant to whether the cables are ok
because of the massive amounts of error correction and
retransmit that the TCP protocol will transparently do.
If you look at the ifconfig output, it reports how many
hardware errors and collisions were detected by the card.
If the numbers are high, they are the cause of the problem.
NFS and DHCP both use UDP, which does not have the automatic
error correction so higher layer timeouts have to occur.
These take a lot longer to kick in (but are configurable).
> > There are three systems with a StarLan HUB between.
> > All three run Windows and two of them dual boot
> > RedHat
> > (7.0 and 7.1).
Make sure the network cards all agree that they should
run in half duplex mode because of the hub. If one of them
thinks it can do full duplex, the network is hopeless for UDP.
_____________________________________________________________________
Ltsp-discuss mailing list. To un-subscribe, or change prefs, goto:
https://lists.sourceforge.net/lists/listinfo/ltsp-discuss
For additional LTSP help, try #ltsp channel on irc.openprojects.net
