Re: Database benchmarking...references

From: Daniel Druker <ddruker_at_agsm.ucla.edu>
Date: 19 May 93 12:19:40 PDT
Message-ID: <1993May19.121941.6448_at_mic.ucla.edu>


In article <SHAH.93May19092228_at_santur.tay1.dec.com> shah_at_tay1.dec.com (Amitabh Shah) writes:
>
>On 18 May 93 06:16 PDT, Richard Finkelstein <rfinkelstein_at_igc.apc.org> said:
>
>> What Oracle did was the an extreme case of what all vendors were doing
>> (including DEC) which was to provide benchmark results which were
>> deceptively high and which were misleading to the public.
>
>Unlike many other benchmarks, the TPC benchmarks require that the sponsor of a
>result completely document how they achieved their performance numbers,
>including disclosing the complete application code and the special parameters
>set in the system. This requirement of the Full Disclosure Report in the TPC
>benchmarks is a significant contribution of the TPC, IMO. If the goal was to
>mislead the public, why would the TPC require FDR's?
>
>> To Sybase's credit they have aggressively tried in their
>> last press releases to get away from these silly benchmarks and get back to
>> reality. I hope other DBMS vendors do the same.
>
>As far as I know, Sybase has been and still is a strong supporter of the TPC
>activities. Yes, they have been hurt recently, when compared to the sleazy
>results from Oracle7. But they continue to support TPC, and publish TPC
>benchmark results. I have not heard anything to the contrary.
>

Mr. Finkelstein. Howabout a little disclosure of your own ? Did you receive any funds from Sybase in the last year ?

If you'll look at the Sybase full disclosure for their TPC benchmarks, you'll see that they used 32 and more history tables to get around problems with page level lock contention and escalation. Do you know of any real app that splits its transaction log into 32 seperate segments like this ? Sybase didn't even have a union command at the time so a user could get a view of all the transaction history !

How does this make Sybase more creditable than anyone else in terms of TPC Benchmarks. Isn't it much more likely that they're de-emphasizing their TPC results now because other vendor's are better (Perhaps they aren't able to cheat as well as everyone else anymore.) It wasn't too long ago when Sybase was crowing about being the fastest TPC database out there...

  • Dan

Daniel Druker
Anderson Graduate School of Management at UCLA                    


| Dan Druker                                                               |
| agsm mail 	: ddruker                                                  |
| internet 	: ddruker_at_agsm.ucla.edu                                    |
| oracle*mail	: unix:ddruker_at_agsm.ucla.edu                               |
----------------------------------------------------------------------------

Disclaimer: None. I'm a student now and I don't care what you think. Received on Wed May 19 1993 - 21:19:40 CEST

Original text of this message