Hi, On Sun, Jun 27, 2010 at 02:03:41AM -0400, Paul wrote: > All the switches/routers are Cisco based, but I seriously doubt that has > anything to do with this. :P
Oh, if it's 2960 or 3750 switches, you could run into the "these switches
have too tiny buffers to be useful" problem.
Your server sends out a burst of packets to the client, with the full
interface speed it can handle. Now if the link switch->internet router
is slower (100m only) or has "other server" traffic on it, the switch
needs to buffer these micro-bursts for a few ms - and that needs some
amount of RAM, which the 2960/3750 just don't have... -> packet loss,
bad throughput.
(Unfortunately, design goals for the 2960S/3750X were different than "get
this fixed", so the buffer size is the same)
gert
--
USENET is *not* the non-clickable part of WWW!
//www.muc.de/~gert/
Gert Doering - Munich, Germany g...@greenie.muc.de
fax: +49-89-35655025 g...@net.informatik.tu-muenchen.de
pgpPYfzxvpCBD.pgp
Description: PGP signature
_______________________________________________ cisco-nsp mailing list cisco-nsp@puck.nether.net https://puck.nether.net/mailman/listinfo/cisco-nsp archive at http://puck.nether.net/pipermail/cisco-nsp/
