Re: Large databases

Keith Fulton (kfulton_at_mgrm.com) wrote:
: I've got a friend thinking about downsizing an enormous
: application from a 3090 to a Unix box, and he's thinking
: about using Oracle. I want to get y'all's reaction to
: this.
 

: This application will has, right now on the 3090, several
: tables with up to 100 Million rows in them. Several of the
: reports require sorts of up to 50 Million rows at a time.
: Some of the extract programs transferring data between one
: table and another write temporary files of 7 Gig. This
: system is on the biggest 3090 available and brings DB2 to
: its knees. When my friend talked to an Oracle salesguy
: about porting this to a (big) Sun box, the Oracle guy said
: "Nooooo problem."
 

: What do y'all think about this? Does this qualify as a
: big database anymore? Will Oracle for SPARC handle it?
: What are some issues that we may not be thinking about
: (i.e. not diskspace, MIPS, throughput, RAM, etc. but DB
: issues.)?
 

: Thanks for your advice in advance...
 

: Keith
:
: ---------------------------------------------------------
: Keith Fulton | "If we were in a court of law
: 711 Louisiana Ste 1300 | arguing this out, I think I
: Houston, TX 77002 | would win. It's that simple..."
: kfulton_at_mgrm.com | (He did.) Emmitt Smith #22
: ---------------------------------------------------------

It is difficult to provide an opinion without more information, but I can say that generally that I have not seen many cases of large databases being moved to Unix machines. This database seems to be in the _exceptionally_ large category which should raise some red flags. I would advise the customer to go back to the sales representative who said "no problem" and ask for references that have implemented a database of approximately the same: size, type of transactions and queries, transaction volumes, and response times. My guess that the sales person will come back with none. This does not mean that no one has tried, but more probably many customers have tried and failed. It is possible for this particular project to suceed but it should be acknowledged up front that it will be a ground-breaking project with lots of risks. Expect to spend alot of time and money on the conversion - probably double your best estimates.

• Rich

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