This may not be right forum to ask this question but I am hoping some
Fedora buffs will know the right place for me to go with this.
I have a network built around some Cisco switches. There is a 3550-12T
and a pair of 2950-48's connected via gigabit links to this switch. I
have 7 server boxes connected onto the 3550 via Intel E1000 cards. These
servers also have 100 Mbps links to the 2950's over e100 on board nic's.
I have the 3 nic's on each server set up under a bonding driver in
active backup mode with the gigbit nic as the active master and the 100
Mbps nic as the passive backups. Everything functions as expected with
fail over on link failure being 100% transparent ( 8-) ) EXCEPT the
bandwidth available down the gig links server to server is at best 288
Mbps! The bond runs a native (untagged VLAN) and a tagged VLAN offering
on 2 separate networks. The machines are Intel SRMK4 running dual
1GigaHz Xeon's vintage 2000/2001 with 4GB of memory and the latest stock
Fedora 9 PAE kernels.
I have set up iperf tests and thumped as hard as I can on the E1000
tuning parameters and this is the best I can achieve. Now what I want to
know is should I be able to do better with this set up. I am wondering
if I am hitting limits on the PCI bus, the switching capability of the
3550 or just that the E1000's can't really provide full bandwidth
without using jumbo frames (which are not available as the bond driver
stops this).
Does anybody have any ideas, or pointers to likely solutions?
Regards, Howard
--
fedora-list mailing list
fedora-list@xxxxxxxxxx
To unsubscribe: https://www.redhat.com/mailman/listinfo/fedora-list
Guidelines: http://fedoraproject.org/wiki/Communicate/MailingListGuidelines
