Date:         Sat, 22 Feb 1997 23:16:41 -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>
Organization: Media Shower, Inc.
Subject:      Re: SAS Institute's Scable Performance Data Server - How good is
              it?
Content-Type: text/plain; charset=us-ascii

Douglas McManus wrote:
>
> I now work on a unix platform with data sets on the order of 8 million
> observations and just fit under the 2 Gig constraint for file sizes.
>
> I have an application that will soon require working with data sets on
> the order of 30-50 million observations.  I saw an ad for something
> that sas appears to have developed with SUN a server engineered for
> data warehousing (Orlando II release).
>
> Has anyone used this product?
 
 
I've been working with the SPDS since beta.  We've used version 1.0
heavily and I've just installed version 1.1.
 
 
> Is it good or does it stink?
 
I think it's amazing.  We are currently running the SPDS with SAS 6.12
on a SPARC 1000 with Solaris.  Our largest dataset at this time is 16
million rows and 3.2 GB.  I have been very pleased with its performance.
We currently use Oracle as our data repository.  We have been extracting
portions of this data to the SPDS for some of our analysis and
reporting.  The results have been so encouraging that I will soon begin
the process of replacing Oracle entirely with SPDS datasets.  Version
1.1 seems to have made some huge leaps in both performance and
adiministration sophistication (security, indexing, data management,
etc).  Basically, the server is FAST.  Even when our programs require
full table scans of the dataset, the SPDS is faster than straight SAS.
When the programs are able to use any of the defined bitmap indexes,
performance improves substantially.
 
 
> Is it overpriced?
>
 
It is expensive.  We licensed Version 1.0 at half-price (the Institute
was trying to generate interest for 1.1), but not when you consider how
much Oracle or other decision support RDBMS are.
 
See the article(s) in SAS Communications (4th Qtr 1996) for some more
information.
 
 
Jay L. Stevens
The Prudential Bank
 
jstevens@prudential.com
 
--