Re: Oracle/Veritas File System

Sybrand Bakker wrote:
>
> On Wed, 12 Jun 2002 14:05:50 +0200, "Knut Talman"
> <knut.talman_at_mytoys.de> wrote:
>
> >> Connor,
> >> Have you had a chance to compare direct io w/ qiock io? Veritas claims
> >> a near-raw-device performance.
> >
> >I am not Connor, but I do have experience w/ quick I/O. We made some
> >performance tests (UFS vs. VFS, raw vs. quick I/O, etc.) when evaluating the
> >Veritas Oracle Database Edition, and yes, it's really as fast as raw devices
> >(at least on Solaris 6).
> >
> >Regards,
> >
> >Knut
> >
> Knut,
>
> Could you provide more detail, if not in the group to my e-mail
> address.
> We have an customer on Solaris 2.6 that is using the Basic Edition of
> Veritas that comes with Solaris and the customer needs to be convinced
> he should buy the Advanced Edition. Using raw devices is not really an
> option: the database(s) grow tremendously fast due to lack of
> procedures to archive historic data (this is a 3rd party app, and we
> can not address that issue).
>
> Regards
>
> Sybrand Bakker, Senior Oracle DBA
>
> To reply remove -verwijderdit from my e-mail address

Sybrand (& Netcomrade)

I'm not a good one to ask since I'm a great fan of raw. In terms of 'raw is not an option', you still have to do some preliminary work with a QIO file, so its not really much different.

The one strong for argument for QIO (at the place I was using it) was that they had some largish machines (in terms of memory) and 32-bit oracle. Using QIO (actually "Cached QIO" allowed use of more of the ram - an advantage that 64bit Oracle would/should obviate. And yes - it outperformed the original vxfs files that were originally in use significantly

The strong argument against QIO - is that raw devices are free :-)

hth
connor

-- 
==============================
Connor McDonald

http://www.oracledba.co.uk

"Some days you're the pigeon, some days you're the statue..."