On Mon, May 4, 2009 at 3:43 PM, Alan Schilla (OET) <alan.schi...@state.mn.us> wrote: > Interesting. Do you have any numbers on how much overhead? Is it 10% or 80%? > I plan to use LACP too but my intent is more for high availability that high > bandwidth. If I dual path OSA's and Cisco 3750 switch ports into zVM VSWITCH > I would rather use both ports than have one sitting in standby. Personally I > am not impressed that you must dedicate OSAs to a Link Aggregation VSWITCH > since I would just as soon EMIF and share this to 2 zVM LPARS. I am still > interested in the overhead of LACP.
Have a look at http://www.rvdheij.nl/Presentations/zLX44.pdf (pg 29-) Roughly 50% of an IFL to receive 100 MB/s through VSWITCH LACP. While folks claim that is because CP must copy the data from one buffer to the other, copying data can be done in 1% of an IFL for 100 MB/s, so there is more happening there. On top of this is the time that Linux itself must spend to manage the data and do useful things with it. That increases a bit when you let Linux do the ethernet bondng, but nowhere what you save on CP. Rob -- Rob van der Heij Velocity Software http://www.velocitysoftware.com/ ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to lists...@vm.marist.edu with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390