Re: ADDM advices Oracle's application logic

On 22 jan, 11:06, Frank van Bortel <frank.van.bor..._at_gmail.com> wrote:
> Helma wrote:
> > On 20 jan, 11:47, Frank van Bortel <frank.van.bor..._at_gmail.com> wrote:
> >> Helma wrote:
> >>> BTW i use raid 0+1
> >> I have seen Compaq hardware RAID controllers that have less
> >> throughput that an old RLL controller. I'd concentrate on that.
>
> >> Comment of Compaq on that particular event was something along
> >> the line of "for that money, you do not expect a real RAID
> >> controller, do you?". Followed by the usual comment on which
> >> beautiful controllers were on sale, for a hefty price, of course.
>
> >> The company that hired me responded with a "get me a decent
> >> one now, or pick up every damn piece of Compaq tomorrow".
>
> >> Believe the hardware was a 360, but it could have been a 570.
> >> It was the basic model. The new RAID controller did make
> >> a difference!
>
> >> I explained earlier, I use LVM on top of two disks, which
> >> would qualify as a stripe set (RAID-0), then.
> >> --
>
> >> Regards,
> >> Frank van Bortel
>
> >> Top-posting in UseNet newsgroups is one way to shut me up
>
> > I'm looking now with the sdparm tool if there's something wrong with
> > the IO setup. This tool isn't as intuitive as hdparm, but let's tinker
> > with it.
> > Meanwhile, the filesystemIO_options was set to 'none', i put it on
> > setall, although i didn't test the result, as the main problem seems
> > to be OS related.
>
> > (alter system set filesystemIO_options =setall scope=spfile )
>
> Thqat's the same here.
>
>
>
> > Regards,
> > Helma
> > ps controller here is a p400 , i'm told i should be 'very fast'.
>
> SQL> @helma
>
> Tablespace created.
> Elapsed: 00:04:28.76
>
> Tablespace created.
> Elapsed: 00:04:38.38
>
> Tablespace created.
> Elapsed: 00:04:48.57
>
> Tablespace created.
> Elapsed: 00:04:51.29
>
> Tablespace dropped.
> Elapsed: 00:00:02.64
>
> Tablespace dropped.
> Elapsed: 00:00:01.01
>
> Tablespace dropped.
> Elapsed: 00:00:00.94
>
> Tablespace dropped.
> Elapsed: 00:00:00.94
>
> Knocked some 45~90 secs off the creation, but the drops are
> lightning fast!
>
> On the RAID controller; if you have time and resources, all that
> will be conclusive would be your measurements. Reinstall the lot,
> without using the RAID capabilities, and define RAID01 yourself.
> Then rerun the iozone tests (no need for oracle, besides, iozone
> originates from Oracle).
>
> Then draw your conclusions (that would be interesting...)
> --
>
> Regards,
> Frank van Bortel
>
> Top-posting in UseNet newsgroups is one way to shut me up

:) Ah, a world with plentora of time and resources. Unfortunately this is my last week here and the machine is in production, hurries to reload and recalculate December's records. The IO is still unpredictable to me. I just added 2 more datafiles, it took about 8minutes each. Yesterday - before taking in production - it took about 54 minutes. And the IO is still at around 33Mb per second. Let's see what i can find. But thank you for staying with me Frank!

Regards,
Helma Received on Tue Jan 22 2008 - 05:41:57 CST