Hey, > What kind of switch is being used for the storage back end? Does each > server have a GB nic used exclusively for storage? Are jumbo frames > enabled on the nics and the switch? Did you do any sysctl tweaks to > adjust rx/tx buffers to be more in line with jumbo frames? > > Finally, what are you using on the storage node for a target? vblade, > kvblade .. or one of the other multi threading user space > implementations?
I'm using newest version of vblade on target. Currently I have only direct link between target and client machine, so there is no switch inbetween. The link is only dedicated for AoE traffic, although IP address is set for both nics if that may cause any issues. Jumbo frames are also enabled on both ends (MTU set to 9000). I have tried changing value of /proc/sys/net/core/rmem_max and rmem_default according to what's said in README file in vblade package. I have also tried changing RX buffer with ethtool -G command. Also I'd like to underline that it works like a charm when I have Opensuse on client side and I don't have to do any other tweaks than use jumbo frames. Just ask if you need more information, I'm still fairly new with aoe :) -samuli ------------------------------------------------------------------------------ Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day trial. Simplify your report design, integration and deployment - and focus on what you do best, core application coding. Discover what's new with Crystal Reports now. http://p.sf.net/sfu/bobj-july _______________________________________________ Aoetools-discuss mailing list Aoetools-discuss@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/aoetools-discuss
