Date: Wed, 12 Jul 2000 22:21:03 +0200
Reply-To: Mauro Morandin <mauro.morandin@TISCALINET.IT>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Mauro Morandin <mauro.morandin@TISCALINET.IT>
Organization: Tiscali Spa
Subject: Re: INSERT SASDATASETS into ORACLE Tables
Scott,
I don't believe that SAS has all the answers! SAS is really fast when doing
stupid things
on huge amounts of data -> I don't think there is anything which can beat a
PROC MEANS
or a data step merge in terms of efficiency........but have you ever tried
to do a star schema join
with SAS on multiple keys in one single view with SAS ... forget it ... use
Oracle.
SAS has done a lot in improving PROC SQL, it's becoming faster and more
efficient with every
new release of SAS, I haven't tried it on a big database in V8. In V6 I
often had out of
memory errors using PROC SQL... I had to find other less memory hungry
solutions like a
pair of SET and SET KEY=... I think you are better off understanding the
strength of both
systems and using them at their best. Cost may be an issue too.
Ciao.
Mauro.
--
-------------------------------------------------------
MATRIX
Castelfranco Veneto - ITALY
email: mauro.morandin@tiscalinet.it
-------------------------------------------------------
"Scott Chapal" <schapal@mail.jonesctr.org> ha scritto nel messaggio
news:p1z0zw59r.fsf@mail.jonesctr.org...
> paul_dorfman@HOTMAIL.COM (Paul Dorfman) writes:
>
> > Welcome to the wonderful world of the stellar Oracle performance! I hope
it
> > is worth those M$$ that migrated to the Oracle coffers. On a
constructive
> > note, please let me suggest that you should read the excellent SUGI 25
paper
> > by our fellow sasler Dianne Rhodes "Migrate to ORACLE? I need my SAS
> > Software!"? Suffices it to say that its conclusion consists of but a
single
> > sentence: "Stick with SAS".
>
> Please elaborate. We are also evaluating the coexistence of SAS and
Oracle.
> Is the paper available on-line somewhere? I searched www.sas.com but
didn't
> see it.
>
> > >I'm wondering if there are any tricks to tune this statement
> > >in order to decrease the run time (estimations are approx
> > >30h for 10000000 observations with 2 attributes each).
>
> How bad is this? Compared to what?
>
> Thanks.
> --
> Scott E. Chapal
| 
