Re: MS SQL Server 6.0, TPC-B world record performance
Date: 1995/07/18
Message-ID: <Pine.SOL.3.91.950718134621.4768A-100000_at_exeter>#1/1
On Thu, 6 Jul 1995, Gene Kligerman wrote:
> In article <3t5jf3$plu_at_oslo-nntp.eunet.no>, olavt_at_microsoft.com (Olav Tollefsen) writes:
> >AT&T Global Information Solutions Announces Industry's Best
> >Price/Performance TPC-B Benchmark Result
> >
> >AT&T Globalyst(tm) S40 PC Server and Microsoft SQL Server 6.0 combine for
> >new world's record
> >
> >The TPC BenchmarkB, developed in 1992 by the Transaction Processing
> >Performance Council (TPC), is considered to be a leading workload benchmark
> >test for evaluating real-world application performance.
...
> >--
> >- Olav Tollefsen
> >- Microsoft Norway
> >-
>
> TPC-B "a leading workload benchmark for evaluating real-world performance"?
>
> Surely you jest.
>
> This must explain why TPC council has announced that no new TPC-B benchmark
> results can be published as of July, and all of TPC-B results that
> have been published (including this one) become obsolete and unreferencable
> by the end of the year.
>
> Clearly TPC council (owner of this benchmark) knows more about what is
> "real-world performance" than the author of the referenced press release.
>
> > The TPC-B benchmark simulates the activities found in complex OLTP
> > environments
>
> Well, perhaps TPC-B is a "complex OLTP" environment for Microsoft SQL Server
> customers. Most other database vendors (Sybase, Informix, DB2, Tandem ...)
> have been duking it out using a more complex (and, dare I say realistic?)
> TPC-C benchmark.
>
> --
> Gene Kligerman
> DB2 Planning, IBM Software Solutions Toronto Lab
> genie_at_vnet.ibm.com
>
>
Gene is absolutely correct. Performance work undertaken here at City U as part of a PhD (not mine incidentally, I'm working on Object DBMS performance) resulted in the City Benchmark (City BM) for OLTP. This work was undertaken by empirical means (collecting over 40 million transactions worth of data and analysing several thousand discrete applications) at three of the largest computer sites in the UK (a large international airline, a "Big Four" UK bank, a local authority computer centre). The results of this work conclusively showed that the TPC-A and TPC-B benchmarks were not representative of OLTP applications and the City BM was significanly cheaper to set-up and run than the TPC-C benchmark. This work is now being used by a number of organisations in the UK.
Regretably, published papers and the PhD thesis are _not_ available electronically at the moment. But if there is enough demand, it may be possible to put some of the papers on-line (depending on my own research workload, etc.)
Regards,
Akmal.
Akmal B. Chaudhri, Systems Architecture Research Centre (SARC), Computer Science Department, The City University, Northampton Square, London EC1V 0HB, United Kingdom
Tel. 0171-477-8551 Fax. 0171-477-8587 Email akmal_at_sarc.city.ac.uk http://www.cs.city.ac.uk/finger/akmal | http://www.city.ac.uk/~ad533/info.html
Received on Tue Jul 18 1995 - 00:00:00 CEST