Re: Oracle threatens to sue Standish over an article

From: Thomas Cox <tcox_at_qiclab.scn.rain.com>
Date: 3 Jul 93 02:14:08 GMT
Message-ID: <1993Jul3.021408.10121_at_qiclab.scn.rain.com>


timo_at_mits.mdata.fi (Timo Talasmaa) writes:
>shah_at_santur.nuo.dec.com (Amitabh Shah) writes:
 

>>In view of the recent discussion here about Oracle7's misleading TPC numbers
>>using discrete transactions...

FYI, the Oracle7 'discrete transaction' is apparently modelled after IMS/Fastpath.

In defense of the Oracle benchmark, two points:

  1. The benchmarks mentioned are usually run by the hardware vendors -- i.e., the Hewlett-Packard benchmark of Oracle7 was run by HP (I believe).
  2. There appears to be no doubt that all benchmarks of Oracle7 have been completely within the letter of the TPC (Transaction Processing Council) rules, and all have been audited. This is in contrast to DEC's RDB benchmark not long ago that was *withdrawn* for *not* following the rules.

>>It would be interesting to see if Oracle actually publishes TPC-A numbers
>>using Oracle7, but without using discrete transactions. I think if
>>Oracle does
>>go through with the threat, then these data will be disclosed as a
>>part of the
>>lawsuit. This is something that Oracle has resisted for a long time!!

While it's fashionable to bash Oracle, in this case I have to blame the TPC benchmark itself. If 'discrete transactions' can give Oracle a 100% increase in its benchmark scores, then _either_ DTs are good, _or_ the benchmark is useless.

vThe whole point of a benchmark is, I thought, to indicate how well a product would run my application.

>Can someone please tell me what a TPC-A or B is? Or
>better yet post it here so other's will benefit, too.

The Transaction Processing Council (TPC) includes reps from most of the major DBMS vendors, plus some hardware vendors. They put out a set of specs for two or three 'standard' benchmarks. You can run it yourself, but when you submit your results you must have the results and your methods audited by a third party.

The TPC-A is an OLTP simulation with real terminals and real i/o hitting   the database box, designed to show the overall performance of a   particular hardware-software combination.

The TPC-B is an OLTP simulation with simulated terminals, tending to   stress the DBMS; terminal i/o is minimal or nonexistant.

The TPC-C is supposed to be a combination of OLTP and decision support;   I don't know if the spec is out of committee.

Additional information and corrections welcomed.

-- 
Thomas Cox      DoD #1776   '91 CB 750 Nighthawk   tcox_at_qiclab.scn.rain.com
    The Platinum Rule:  "Do Unto Others As They Want To Be Done Unto"
Received on Sat Jul 03 1993 - 04:14:08 CEST

Original text of this message