From: Stefano Lassi (stefano.lassi@sysma.it)
Date: Tue Dec 05 2006 - 01:00:44 ART
Hi
I'm trying to understand Cisco Etherchannel Hash load balancing
algorithm
I have got 2 6509 (CatOS 8.5(3)) connected via a 2 port port-channel,
load balacing method is "ip both"
testing on both sides I found:
SWCORE1> (enable) sh channel hash 1666 192.168.100.35 192.168.100.23
Selected channel port: 1/2
SWCORE1> (enable) sh channel hash 1666 192.168.100.23 192.168.100.35
Selected channel port: 1/2
SWCORE1> (enable) sh channel hash 1666 192.168.101.23 192.168.240.89
Selected channel port: 1/1
SWCORE1> (enable) sh channel hash 1666 192.168.240.89 192.168.101.23
Selected channel port: 1/1
---------------------
SWCORE2> (enable) sh channel hash 1666 192.168.100.23 192.168.100.35
Selected channel port: 1/1
SWCORE2> (enable) sh channel hash 1666 192.168.100.35 192.168.100.23
Selected channel port: 1/1
SWCORE2> (enable) sh channel hash 1666 192.168.101.23 192.168.240.89
Selected channel port: 1/2
SWCORE2> (enable) sh channel hash 1666 192.168.240.89 192.168.101.23
Selected channel port: 1/2
so, hash algorithm on different side seems produce different
results!!!!!
Is it not XOR hash algorithm using only source and dest IPs?
It seems there is something more involved inside calculation ...
1) hostname?
2) management interface IP?
3) some channel/port specific attribute? ("Ch-if Index" or "Oper-group"
or "Admin-group" ... displayed by [sh channel <id> info] commands ...)
4) something else?
My goal is reach situation where traffic between same hosts couple going
forward and back on same link.
Of course I could exchange cables on one end ... but how I could be sure
this situation (XOR load-balancing outcome) will stay the same after
next reboots?
Thank you very much
Stefano Lassi
CCNP/CCDP/CCSP
ITALY
This archive was generated by hypermail 2.1.4 : Tue Jan 02 2007 - 07:50:36 ART