> Could you please send me the log in separate Email I'll let our
> production team know. It is not that common version nowadays but
> anyway.
I'll try to compile 4.1.9 and send the error messages if any
> > The anticipatory scheduler was used on 2.6 kernels and the tag depth of
> > the aic7xxx controller was clamped at 4 via the following modprobe.conf
> > command line
> >
> > options aic7xxx aic7xxx=global_tag_depth:4
>
> Why did you reduce TCQ Depth ?
Nick Piggin has mentioned it on lkml that AS with high TCQ doesn't work
well. For Database servers, it's recommended to use deadline. See
Documentation/block/as-iosched.txt in the linux-kernel source tree. I
got similar numbers with deadline and default tag depth of 32
I'll try to setup a RAID 10 disk mirror and then re-run the same tests
with deadline/cfq/as.
> Yes. 2.6 kernel seems to do much better. It is however barely depending
> on IO subsystem as default table size is just 10000 rows so it will well
> fit in memory.
I changed the table size to 100000 rows. If/When I get a faster i/o
system on the boxes. I'll try with even larger table sizes.
> Very interesting. 2.6 kernel runs very fast but it allows some threads
> to starve - so you can see higher average response time as well as much
> higher 95% percentile response time. Also Thread fairness is pretty
> poor - this parameters measure what is the difference between number of
> requests threads was able to get. In perfect case it should be close 100
> meaning all threads processed about same number of requests.
