Date: Sun, 29 Mar 1998 15:01:23 -0600
Reply-To: drewspen@imagin.net
Sender: "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From: "Kent Andrew (Drew) Spencer" <drewspen@IMAGIN.NET>
Organization: ImagiNet Communications
Subject: notes from SUGI 23 replacing Oracle w/SAS
Content-Type: text/plain; charset=us-ascii
replacing oracle with sas spds
jstevens@prudential.com
prudential credit card
gave a big bang for the buck
used a credit card data mart
users are credit card portfolio analysts and statisticians
hardware sparc 1000e
several tables 300-500 mb with largest being 4gb
small compared with the stuff I use...
the original plan
moving from mvs db2 to oracle
sas was analytical tool
oracle 7.3 was to service the data
the it team initially suggested a sas only solution
rejected because at the time sas did not use smp, the file size
limitation, and column name limit
what went wrong?
Oracle is tuned for transaction processing
a great deal of time and money was spent re-tuning oracle for decision
support
data conversion limitations
mvs data
performance problems
spds outperformed oracle
sas/oracle load process
3490 tapes
sas preprocessed the data
parallel oracle data loading of flat files created by sas
what does spds bring to the table
physical file partitioning
full use of smp
enhanced indexing
seamless integration with sas system
acl security
server and data set level auditing and logging
the conversion to spds
data _null_'s were replaced with sas spds data set names
performance
load test was biased against sas
oracle was loading a pre-processed flat file
sas was converting and loading
sas spds was twice as fast as oracle in loading
the objections
many have not heard about spds
sas does not have auditing and logging - wrong
sas is not a relational database
the 'relational religion'
being a full blown rdms is part of the problem
don't have the overhead of a lot of the transnational rdbms features
of oracle
performance improvement
faster load
parallelism and bitmap indexes
the primary power of spds
seamless, high performance access to data
if your is people are interested in business needs instead of their own
is agenda, then you are half way there
the issues are political , not technical
tuning with spds is the same as for oracle - maximize i/o
sun was a principal consultant in tuning
spds 2.5 and spds 3.0
star schema optimization
mddb inside spds
sas data step views on the spds server side
spds uses less disk space that sas, either compressed or uncompressed
spds footprint is only 2.5mb
//-----------------------------
// Kent Andrew (Drew) Spencer
// DrewSpen@Imagin.Net
// http://www.imagin.net/~drewspen
// http://www.jjt.com/ntsug
