Re: Oracle TPC-A Benchmarks

Dave Laudicina (dil.admin_at_mhs.unc.edu) wrote:
: I did read somewhere that Sequent had acheived 1000 on one of there
: boxes. Can't remember what db they where using could hav been Oracle.
: The box is pretty expensive but when you comapre to an IBM mainframe
: running DB2 Cics etc etc. it is probaly a bargain.
: Besides is the industry is truly going to rightsize at thge enterprise
: level expensive large smp
: servers will probably be the way to go but they don't come real
: cheap. The important thing seems to be that you can start small and
: expand easily by adding new processors.
: Thx Dave L

Thanks for your response Dave. I guess my main concern is that the ad seems to imply that Oracle can really achieve these numbers in a real-life production application. In my experiences, Oracle would be hard pressed to achieve 1/100th of these TPS numbers. While it is true that many users would look at these numbers at have a good laugh, there are many potential customers (who this ad necessarily is targeted to) who do not know that the TPC numbers are designed and cooked by RDBMS and hardware vendors who are simply looking to increase sales. Hopefully, a good evaluation effort would surface the fact that Oracle cannot come close to 1000 TPS and 10000 concurrent users on two low-end Unix machines (really an incredible tale by Oracle marketing). However, it bothers me to see Oracle resorting to these tactics once again as it did several years ago. I was hopeful that Oracle (and other vendors) were trying harder to get good information out to the public so that we would have more success stories and less fiascos.

Regards,
Rich

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