Blogbench RAID benchmarks

Matthew Dillon dillon at apollo.backplane.com
Mon Jul 11 12:39:46 PDT 2011


:> writes:           600
:> reads :        178372
:
:For good measure, I did the same test with UFS1+softupdates.
:The results are much lower:
:
:  writes:         282
:  reads:        20522
:
:there was very little disk activity and blogbench was almost constantly
:sleeping (process in nanslp state)
:
:-- 
:Francois Tigeot

    There's something very basic wrong.  How much memory does the machine
    have?  Blogbench clearly isn't running long enough to blow out the
    system caches.  Try more iterations:

	blogbench --iterations=100
	blogbench --iterations=200

    Otherwise you aren't testing the disk drives at all.

    --

    w/regards to the CAM fix, to determine if it is meaningful use
    /usr/src/test/sysperf/randread.c.  The default parameters should be
    sufficient.  Run it on the raw disk block device (it should work for
    FreeBSD too but I'm not sure lseek will properly size the drive).

    cd /usr/src/test/sysperf
    make /tmp/rr1

    Serialized disk reads will show up at around ~150-200 tps.  Multi-tagged
    disk reads will show up in the 400-500 tps range.

						-Matt





More information about the Kernel mailing list