On Thu, 24 Sep 2009 12:48:50 +1000 ronnie sahlberg <ronniesahlberg@xxxxxxxxx> wrote: > For certain workloads, when/if the entire lun fits in page-cache you > get better performance with only one single worker thread than > multiple ones. > This eliminates the context switching and bouncing between cores on a > many-cores machine. bs_aio would be faster for such workloads? BTW, I think that the signalfd patch decreases the overheads of using worker threads. > For these workloads, the single thread used to drive the eventsystem > quickly becomes a bottleneck which is why I will also follow up with > additional patches to make the portal address and the port > configurable, to allow running multiple tgtd's on one system. One > instance for each core, or pair of cores. I don't recommend running multiple tgtd daemons but it should work if administrators understand what they are doing (devices can't be shared in tgtd daemons, etc). Anyway, if you send patches to make the portal address and the port configurable, I'm happy to review them. Thanks, -- 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
