Re: Are we in the ballpark?

On Wed, Feb 02, 2011 at 10:06:53AM -0700, Wayne Conrad wrote:
> On 02/01/11 18:30, Greg Smith wrote:
> >>Bonnie++ (-f -n 0 -c 4)
> >>$PGDATA/xlog (RAID1)
> >>random seek: 369/sec
> >>block out: 87 MB/sec
> >>block in: 180 MB/sec
> >>$PGDATA (RAID10, 12 drives)
> >>random seek: 452
> >>block out: 439 MB/sec
> >>block in: 881 MB/sec
> >>
> >>sysbench test of fsync (commit) rate:
> >>
> >>$PGDATA/xlog (RAID1)
> >>cache off: 29 req/sec
> >>cache on: 9,342 req/sec
> >>$PGDATA (RAID10, 12 drives)
> >>cache off: 61 req/sec
> >>cache on: 8,191 req/sec
> >
> >That random seek rate is a little low for 12 drives, but that's probably
> >the limitations of the 3ware controller kicking in there. Your "cache
> >off" figures are really weird though; I'd expect those both to be around
> >100. Makes me wonder if something weird is happening in the controller,
> >or if there was a problem with your config when testing that. Not a big
> >deal, really--the cached numbers are normally going to be the important
> >ones--but it is odd.
>
> I also thought the "cache off" figures were odd. I expected
> something much closer to 120 req/sec (7200 rpm drives). I probably
> won't investigate that with any vigor, since the cache-on numbers
> are OK.

You may want to look into the "cache off" figures a little more. We
run a number of battery backed raid controllers and we test the
batteries every 6 months or so. When we test the batteries, the cache
goes off line (as it should) to help keep the data valid.

If you need to test your raid card batteries (nothing like having a
battery with only a 6 hour runtime when it takes you a couple of days
MTTR), can your database app survive with that low a commit rate? As
you said you ar expecting something almost 4-5x faster with 7200 rpm
disks.

--
-- rouilj

John Rouillard System Administrator
Renesys Corporation 603-244-9084 (cell) 603-643-9300 x 111