Re: tough choices

Serge Rielau apparently said,on my timestamp of 21/06/2004 9:42 PM:

> "organization" since it is supported in DB2 UDB for LUW at present.
> MDC is supported in DB2 + DPF. What happens is that first the table is

Is this DPF thing one of those famous separately priced options?

>
> I'm not certain if adding anny other partitioning but hash on the
> database level is a good idea. It invites data skew and would be a
> maintenance nightmare to add more database partitions.

Only if you implement it as an add-on option...

I get extremely worried everytime I hear the word "partitioning" mixed up with "clustering" and "databases".

Rather than the simple word "table". The reason for this worry is that one thing is database clustering, the other (and completely different one) is object partitioning. The two should never be mixed under the penalty of obfuscation. Let's not go into index partitioning and optimization across multiple nodes.

Last time I looked, adding a partition to a table in Oracle was a simple SQL statement that can be run without affecting other users of said table. Same goes for dropping a partition. Both tables and indexes work the same way. None of that multi-whatever clustering, orthogonal or not, or multiple databases: single instance is plenty enough (single license, no clustering add-ons).

And if it is not a hash partition, (a date range partition, for example) it is if anything easier. Which makes it extremely easy to implement very large "rolling date" transaction tables in a single database instance . Since V8.0, BTW. MOL 1997.

-- 
Cheers
Nuno Souto
wizofoz2k_at_yahoo.com.au.nospam