[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference nsic00::eis_dw

Title:Executive Information Solutions & Data Warehousing Conference
Notice:Welcome to the Data Warehousing conference
Moderator:26002::HAGGERTY
Created:Thu Sep 01 1994
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:499
Total number of notes:2932

485.0. "TPC-D with more than 100GB" by VAXRIO::NOVELLO (Cristina Novello - Sales Suport/RJ) Mon Apr 28 1997 23:47

    Hi,
    
    our customer is asking why Digital has TPC-D numbers
    only using 100 GB? Cost reasons?
    
    Thanks,
    Cristina.
    
T.RTitleUserPersonal
Name
DateLines
485.1How much data reallyUTROP1::utodhcp-198-112-175.uto.dec.com::olthof_hSpellchecked Henry AlthoughTue Apr 29 1997 12:2035
Hi Cristina,

Running benchmarks is very expensive. You have to have all the equipment
together (a tough job when you manufacture the stuff yourself), all the
experts, the right software etcetera.

If you look at our TPC-C results you will notice that the majority of the
results where done with older equipment, like the 350 MHz CPU's. I guess
that now that the competition has come a little closer we could redo the
tests with the faster hardware that we have at the moment to regain the
top position firmly.

I'm told that we are working on the 300GB TPC-D result right now. However,
please keep in mind that each TPC-D database size is the pure data. So,
the 300GB is the amount of records multiplied by the record size. In
practice, you need a lot more disk capacity to store that data (index,
overheads, ..). If I recall well our 100 GB TPC-D result was done on in
total 361.2GB of disk storage, so the ratio (mandatory disclosure) is
3.612.

As an illustration, here's what a customer of mine wrote in an RFP:
- aggregations: 25% on top of raw data;
- indexes and workspace: 200% of raw data plus aggregations;
- RAID5 requirements: 20% of the sum of the above;
- Speed optimization: only 30% of disk space will be used.
This leads to about a factor of 10 to multiply the raw data with to 
calculate the total capacity!

Why do I tell you this? Verify how much raw data your customer really has,
it may well be below the 100GB level, in which case we currently have an
excellent TPC-D result.


Cheers,
Henny
485.2Thanks for the explanation...VAXRIO::NOVELLOCristina Novello - Sales Suport/RJTue Apr 29 1997 18:5210
    Hi Henny,
    
    I imagine it is a great effort to publish and audit all these
    numbers. Thanks for all the explanation regarding how much
    really we need for storage everything.
    I only needed a good reason for the many customers who are
    asking for bigger TPC-D numbers.
    
    Thanks again,
    Cristina.
485.3yup, they're coming26002::HAGGERTYKevin, NSIS, Stow MA USAMon May 05 1997 00:475
    Cristina, we are working on TPC benchmarks of several different
    flavors.  As Henny said, 300gb is included.
    
    If you need more info, send me mail.
    
485.4We need it soonULYSSE::VISCIGLIOPas a l'abri d'un coup de bolTue May 06 1997 15:3512
    Christina,
    
    HP just published impressive 300 GB TPC-D figures, as Sun did recently.
    
    Our customers actually require comparison numbers. I have questions 
    related to this at each customer presentation.
    
    Regarding the size, some industries need very large DWs. For the Telecom 
    industry, 300 GB is a minimum. I was even asked why we don't publish the 
    1TB TPCD bench.
    
    Pierre-Yves (Telecom IEC)