Update as promised: I ran the same tests on 2 low end Centos 7.1 workstations with latest updates applied. I got exactly the same behaviour, increasing the iodepth has the same effect as with previous tests. Absolute values were about 5% lover, probably due to low end integrated nics (RTL 8169). Changing the mtu had no measurable effect. At this point I think that we can safely rule out network as a problem. I would be grateful if anyone can replicate my results on other distributions. At this point I'm out of ideas, maybe running blktrace on server can provide some pointers? Any suggestions on how to proceed further ? Dragan -- 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