Re: TPC-D benchmarks (was Re: Sybase MPP shows excellent scalability to 128 CPUs

In article <grimes.816454862@access2>, grimes@access2.digex.net says... > >grimes@access2.digex.net (Seth Grimes) writes: > >>nraden@aol.com (NRaden) writes: > >>>I don't think TPC results are even relevant. What do they have to do with >>>decision support, which is clearly the target market? > >>What is your opinion of TPC-D, which is supposedly designed to test >>complex, decision-support-like queries? I haven't seen the spec or any >>published benchmarks. Is it good for DSS and are vendors adopting it? > >I was informed that the TPC-D specs are on >http://www.tpc.org/bench.descrip.html but that no vendors have yet published >results. I'm sure that all the major SQL database vendors are running >the benchmarks but have any committed to publishing results? > >The benchmark specifies that "the TPC-D database must be implemented using >a commercially available database management system (DBMS), and the queries >executed via an interface using dynamic SQL." "TPC Benchmark D is comprised >of a set of business queries designed to exercise system functionalities in >a manner representative of complex business analysis applications. These >queries have been given a realistic context, portraying the activity of a >wholesale supplier to help the reader relate intuitively to the components >of the benchmark." > >It's not acceptable to the TPC to translate the queries into proprietary >languages and the data structures into non-relational hypercubes but some >vendors in the decision-support arena support only proprietary languages >and non-relational structures. Do any of them plan to run the benchmark >anyway? Even if unaudited, the results could substantiate claims that >a multidimensional database structure is preferable for DSS work to a >relational structure. > >I understand that the first stage of the OLAP Council's API has been >released. What about their proposed benchmarks, which would presumably cover >a lot of the same ground as TCP-D? > > Seth >-- > >Seth Grimes -- grimes@access.digex.net -- http://www.access.digex.net/~grimes Presumaly the council would want to drive at many of the same result needs, but the major differences would probably highlight the fact that the TPC-D requires interactive SQL. I sometimes wonder if there would even be an MDDBMS industry if it were not for the limitations in SQL. SQL does a lot of things great, but among them are not time trends and conversions, compound nested cross-tabulations, or hierarchical relationships. Everybody who has attacked a significant ad hoc decision support system spec with just SQL knows this. With enough coding you can do something in each of those areas, but development is onerous, performance is usually disappointing, and range of possible queries is sometimes limited. Of course, you have to balance that against potential issues of synchronicity and integrity when you seek your summary cross-tabs from an MDDBMS but must go to a different data source when supporting detail is required. But that issue is a structural one, not well addressed by performance benchmark suites. -- Dirk Coburn