Re: Capacity planning for VERY large table

The short answer is it depends :-)

Nothing is going to read 1.5 Tb of data in 1 second. Using indexes will reduce the amount of data that needs to be read, but with 25 billion rows, I think you can pretty much throw out 1 second queries all together. If fast performance is of prime importance you will need a way to get the object being queried smaller.

Many large company's entire datawarehouse is a fraction of this one table. That being said, if you put this table on 128 disks on a server with 128 processors and run with a parallel degree of the same number the response time may be amazingly fast, but who knows, I haven't worked with any objects (or databases for that matter) that are this large.

Unless your company has an extra 3-4 million laying around for a loaded E10000 and EMC disks I would look at creating a smaller object to query.

--
Dave A


"Michael Bialik" <michael_bialik_at_my-deja.com> wrote in message
news:8qoc4j$cfc$1_at_nnrp1.deja.com...


> Hi.

>

>  I have a very large table ( 25 * 10**9 rows = 25,000,000,000 rows ).

>  The row length is 60 bytes, so the table size is 1.5Tb - 2.0 Tb

>  ( including Oracle overhead ).

>

>  The table contains 24 numeric fields and 1 description.

>  Each numeric field ( F01, F02, ... F24 ) may contain values

>  from 0 to 255 only.

>

>  The SELECT is going to look like that:

>

>    SELECT description FROM large_table

>    WHERE

>     F01 between 24 and 44 AND

>     F02 between 66 and 86 AND

>     F03 between 1 and 21 AND

>     ...

>     F24 between 235 and 255.

>

>   Does anybody worked with such big tables?

>   What response time ( with what type of computer ) may I expect?

>   Is it possible to achieve 1 second response time with Oracle ( 8i )?

>   On what computer?

>

>   TIA. Michael

>

>

>

>

> Sent via Deja.com http://www.deja.com/

> Before you buy.