Re: Oracle and Raid setup

Frank van Bortel <frank.van.bortel_at_gmail.com> wrote:
> xhoster_at_gmail.com wrote:
> > Frank van Bortel <frank.van.bortel_at_gmail.com> wrote:
> >
> >>Noons wrote:
> >>
> >>>Cris Carampa apparently said,on my timestamp of 9/06/2005 4:41 PM:
> >>>
> >>>
> >>>>But Raid 1 gives you only fault tolerance, no performance gains
> >>>>because there is no striping. Or am I wrong?
> >>>
> >>>
> >>>Absolutely right. In fact, it is slower for writes...
> >>>
> >>
> >>Why would a mirrored write be slower for writes? OK, as slow
> >>as the slowest disk, but those differences can (should!) only
> >>be marginal.
> >
> >
> > If each write takes a random amount of time uniformly distributed
> > between 0 and 1 (in whatever units of time would make sense), then the
> > average wait for one write is 0.5, while the average wait for slower of
> > two writes is 0.6667. That is a 33% penalty. I don't know if that
> > qualifies as marginal or not. I don't know what a realistic actual
> > distribution of write delays is, but I think uniform from 0 to 1 is not
> > too outrageous of an assumption.
> >
> > Xho
> >
> I don't know what you are talking about, but it's not going to happen.

If you don't know what I am talking about, how do you know if it is going to happen or not?

...

> There will be differences, but these will not be evenly distributed
> in a 0-100% range of the time.

So how *will* it be distributed?

> Actual timing would be 8.0 millisecs vs 9.0 millisecs,

Hunh? One disk is going to take 8 ms for every single write it does, and the other will take 9 ms for every single write it does? I find that hard to believe.

> so the OS would
> wait 9ms instead of 8 - so it could have been 11% faster.

You appear to be taking the average of many single-disk reads and than taking the max. You need to take the max, and then average.

Xho

-- 
-------------------- http://NewsReader.Com/ --------------------
Usenet Newsgroup Service                        $9.95/Month 30GB