On Mon, 2013-08-19 at 14:50 -0400, Jörn Engel wrote: > On Mon, 19 August 2013 12:36:43 -0700, Nicholas A. Bellinger wrote: > > On Sun, 2013-08-18 at 22:43 -0600, Scott Hallowell wrote: > > > > I recommend pursuing a few different things.. > > > > First, you'll want to bump the default_cmdsn_depth from 16 to 64. > > > > The second is to try with write cache (buffered writes) enabled. > > Total agreement on these two. Depending on the storage and initiator > even 64 might not be enough, so you can set default_cmdsn_depth even > higher, if it makes a difference. > > Quite likely your results will still fall short of the samba > comparison. I suspect some reasonably low-hanging optimizations in > the iscsi target stack, but don't have time to get on my ladder and > start picking. > For 1 Gb/sec streaming writes, it's most certainly the fact of buffered writes vs. non buffered writes, and not any type of CPU bound code. --nab -- To unsubscribe from this list: send the line "unsubscribe target-devel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
