Date: Tue, 18 Nov 1997 18:29:01 -0500
Reply-To: "Jay L. Stevens" <jstevens@PRUDENTIAL.COM>
Sender: "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From: "Jay L. Stevens" <jstevens@PRUDENTIAL.COM>
Subject: Re: Justification for Oracle
Content-Type: text/plain; charset=us-ascii
> If you want to set up a double entry system Oracle is much better. (i have
> done it in both systems). In fact data entry via Oracle is in general
> eaiser. For tracking changes to data (if you need to determine when and who
> added, changed or deleted a record or field) Oracle is far and away a better
> choice than SAS (Again I have set up systems in both).
>
Indeed you're probably correct. In fact, in my original post, I
actually recommended Oracle and their development tools to those who are
looking to build operational systems (rather than analytical/decision
support systems). I'm sure there are some SAS-L'ers who are vehement AF
supporters who would disagree with you, though.
> NOT.
> SAS/ACCESS to Oracle makes the access of Oracle table to SAS datasets
> transparent
>
To quote a frequent poster and a 20-year SAS veteran:
"Not only is it not transparent, but the Oracle baggage slows down the
whole train."
I have to agree. In a decision support environment, why should you even
have to go through the headache of defining SAS/Access view descriptors
or writing Proc SQL passthru code when SAS can provide the same
functionality without Oracle. No matter which SAS/Access interface you
use, it only makes logical sense that a) Accessing your Oracle data via
Sas/Access or other query interface then b) cramming that result set of
data through an I/O pipe into SAS for reporting and analysis is going to
be slower than allowing SAS to access the data directly.
> In general SAS is probably the best report and analysis tool to use. Oracle
> is a relational database while SAS is NOT.
>
This is exactly the point. When you are interested in reporting and
analysis (i.e. data warehousing, data-mining application(s), why in the
world do you want to slow down the process with ReDo loggging and
Rollback segments, etc., etc. ad-nauseum, when you don't need that extra
baggage. You have to work extra hard just to get Oracle to stop doing
these things so it will perform.
> You need to evaluate your needs for a mainframe relational database (i know
> there is a PC version). If you need the things a relational database
> provides then Oracle is one of the best choices.
>
Again, this is the point. The relational database vendors realized a
few years ago that this thing now called Data-Mining and Data
Warehousing (something SAS has been good at for many years) is a hot
topic with many potential dollars involved. So...they pulled out their
old relational transaction systems and said, Ta DA! A data warehouse
repository. You can use SAS but still keep the good old relational
database alive and kicking! I seriously question the value of the
relational model in a Decision Support environment.
Jay L. Stevens
AVP Portfolio Analysis
Prudential Bank
jstevens@prudential.com
