Comments on: Initial reactions to IBM acquiring SPSS

By: Historical notes on analytics — terminology | Software Memories

Historical notes on analytics — terminology | Software Memories — Tue, 17 Jan 2012 08:02:31 +0000

[...] Anything in the predictive analytics area (but see the first point in a 2009 SPSS post). [...]

By: It’s On: IBM Acquires SPSS « Market Strategies for IT Suppliers

It’s On: IBM Acquires SPSS « Market Strategies for IT Suppliers — Mon, 03 Aug 2009 02:05:02 +0000

[...] and integrate customized functionality with R and Python in any module.” Curt Monash has pointed out that SPSS is much more SQL-friendly than SAS. Why does that matter? In the DBMS space, a key current [...]

By: RC

RC — Wed, 29 Jul 2009 18:35:42 +0000

Oracle has already bought so many companies, it can’t go on and on.

By: Shawn Fox

Shawn Fox — Wed, 29 Jul 2009 16:09:11 +0000

I have no experience with SPSS but I constantly see SAS trying to pull a full copies of huge tables out of the database to perform some operation on it.

Maybe the SAS/Teradata integration reduces that quite a bit on Teradata, but with other databases I have a lot of doubt as to how good of a job SAS does at pushing analysis into the database instead of pulling all the data out to run the analysis on the SAS server itself.

By: Guy Bayes

Guy Bayes — Tue, 28 Jul 2009 21:04:01 +0000

SAS can be pretty sophisticated in the way it issues SQL selects the RDBMS, has a sort of pushdown optimization capability similar to informatica. However there is no doubt that SAS has a tendency to pull data down into work areas to sort and process it, that is certainly true.

Whether SPSS has that same tendency, I do not know but I would not be surprised if it does. Any external engine is going to have to stream the data from the RDBMS through the engine somehow, and once it is streaming, it will find advantage in locally cacheing. Even Informatica does this.

SAS also has a feature where entire enterprise miner (EM) graphs can be exported and recompiled as UDF’s, run natively inside Teradata. At this point, all execution should happen inside Teradata. However, to get this to work, you have to be using EM against only teradata sources.

By: Curt Monash

Curt Monash — Tue, 28 Jul 2009 20:51:18 +0000

My sense is that SPSS just naturally relies on the database manager to get it data, while SAS prefers one to do extracts into its own proprietary data format first.

Am I wrong in that?

CAM

By: Bence Arató

Bence Arató — Tue, 28 Jul 2009 20:41:40 +0000

Kurt, I’m somewhat surprised at your point of SPSS being more SQL-oriented than SAS.

I feel that actually SAS has the edge now in the
area of in-database mining with its Teradata integration and shared development. SPSS could use and does benefit from the Oracle Data Mining option, but this is a combination which rarely used.

By: Jerome Pineau

Jerome Pineau — Tue, 28 Jul 2009 18:42:32 +0000

@Curt: “I feel another “Future of … ” post coming on.”

I started the thread on my blog

Thanks for taking that task up I think it’s an area that’s been perhaps underserved for a while and I look forward to your insight on that issue!