Re: Choosing a database for a DSS system (40-100G range)

WRONG! If you are stricktly looking at benchmark numbers, Oracle's TPC-D numbers are far superior to IBM's. Oracle's latest TPC-D is over 27,000 on a 1 TB database, where IBM is less than 20, 000.

If you are going to quote numbers, please use something other than a specific vendors web page.

    Eyal Kattan wrote in message <36412BA5.D5CE1C3D_at_infomall.co.il>...     Ken Chiu wrote:

>* how many clients will be connected to the database ?
        1 user, the Crystal Report Server

>* How long the rows will be ?

        15-50 columns (depends on which table)

        50million records per table, 16 tables for now.

>* What kind of data will you store (i.e. will you use BLOBS
etc..?)

        ASCII imports, i.e. Dates, Chars, Numberic

    hmmm......50 million records per table.....that's a big one. And if you have to join some tables in your queries....well.....you really need a strong stable database.

      


        >and there is also....... OS/2 + DB2 

        There is noone in our shop that knows OS/2 well enough to stick 
his 
        neck out...  =)

    But it is much better solution for Intel based machines than NT. 


        >I wouldn't trust NT to handle such size of database. 

        > 

        > 

        I know, but it seems to be cheap to start with.. but might 
eventually 
        get myself burnt!?

    I assume that you wish to stay with Intel based system, however, NT is not even reasonable alternative. My experience with DB2 shows superiority in performance over Oracle (on the same Intel servers) and certainly over SQL-Server(which didn't even come close), along with full salability, stability and many other features, DB2 is a true OODBMS. It has longer history of handling large databases (on MainFrame) and is very mature product. Almost every benchmark test, shows that DB2 is far ahead of Oracle which is the closest in performance to DB2.

    Even though it might be expensive at the first purchase (HW + SW), you will find out that in the long run, the investment pays itself, specially if your data is 'mission critical'.

    I hope this gives some idea....

--