Re: How Large is a BIG Relational DB

Kent S. Gordon (kgor_at_malibu.aadt.com) wrote:

: Does any relational database engine other than TeraData allow for
: multiple cpus and disks to work on a single
: query/insert/update/delete/load/indexing?
 

: I know Oracle 7.1 is supposed to support multiple cpu's id queries
: (maybe loading and indexing), but I believe it will not use more than
: one cpu in a insert/update/delete statement.
 

: I think Informix 6.x will have similar functionality.
 

: I am not sure when Sybase/Ingres/?? will have similar functionality.
 

: I am working on a large (300GB - 2TB) combined DSS/OLTP application.
: I would like to have an alternative to using TeraData (NCR 3600), but
: have found no other database that will use multiple cpu's well in
: handling large (10 - 100 million row) insert/update/delete statements.
 

: Any Comments, helpful hints, corrections are welcome.

Hi Kent.

My investigations would confirm your experiences so far. I have so far found no evidence that either Oracle, Informix, or Sybase can handle databases the size you are talking about for either OLTP or DSS. The quality of the products is low in my estimation and I would not trust any of the RDBMSs in their latest forms for an application of this size. System 10 is too new, and Oracle7 in its 15 months of general release has not really been that heavily exercised. Most of the Oracle7 users I have come across are migrating very slowly and are experiencing some degree of problems even in their gradualy migrations. There are so many variables in terms of platforms, release levels, operating systems, type of application, it is hard to pin down the sources of these problems so it is hard to predict the outcome of any single project.

I lean heavily toward proven solutions such as Tandem Nonstop SQL for heavy OLTP and large databases, Teradata or DB2 for large DSS databases. I would not try to do both on the same machine. An interesting, and probably very cost-effective alternative for online-analytical processing (OLAP) are the multi-dimensional databases such as Express. Their specialized data storing and data handling capabilities are proven for this type of work though I have not seen them used for a database of the size you are describing. However, when redesigned into multi-dimensional form, the databases may be compressed into much smaller size databases.

Hope this helps a bit.

• Rich

-- 
Richard Finkelstein                        Voice: 312-549-8325     
Performance Computing, Inc.                Fax: 312-549-4824 
Chicago, IL                                Mail: rfinkel_at_infochi.com