On Thu, 27 May 2010 18:58:33 +0300 Or Gerlitz <ogerlitz@xxxxxxxxxxxx> wrote: > Looking with some more care, I see that your patch adds parallelism > at the level of iscsi_tcp.c such that still there's one central > event loop (main) thread and the addition is per iscsi/tcp session > dedicated thread. In that case, some of my comments are less > relevant... Not session (nexus) dedicated thread. The patch creates one pthread per target. The patch also converts iSCSI/tcp to use it. > Still I would be happy to learn how you conducted the 500/850 MBs > test and maybe some comments that you have on my feedback. Also, > does the patch introduce another per IO context switch? I set up four targets. Each target has one logical unit (one SSD drive). I think that the patch would not improve the performance of one target having four logical units. Can't we tell people to use multiple targets if they want performance? I thought about pthread per nexus, pthread per connection, etc. Any plan to improve the performance without adding hug lock complexity works for me. -- To unsubscribe from this list: send the line "unsubscribe stgt" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html