Re: partiioning option not worth it?

Marc Blum <marc_at_marcblum.de> wrote in message news:3c777f73.11911187_at_news.online.de...
> 1) what about testing, bugfixing, becoming more robust and the
> maintainabiliy? Each time you implement partitioning on your own, you
> have to test it.

The additional testing required is small assuming that you have isolated the code required to a central data access component - it becomes one additional unit test. This is almost certainly less expensive than the additional support cost of the partitioning option.

> Oracle Corp does the testing for you. Every new release of you
and you believe that ?!!

> buils a further application which requieres hat functionality, you
> have to reimplement it again and again and again...
True, but cut and paste is the most efficient code reuse mechanism I've ever encountered.

> 2) What about the difference between a declarative feature and a
> programmatic solution? You can declare constraints inside the database
> via a simple "ALTER TABLE blablabla" or you can bulletproof your code
> to avoid logical corruption of your data. What's cheaper?

I would tend to agree, except that it is not relevant in this case. It is a straightforward matter to declare constraints that any given partition should contain only appropriate data - this was, after all, one of the implementations of partition views.

I have had amusing conversations with people making this point, when the same people were saying that we shouldn't implement referential integrity constraints because "it's too expensive at run-time" - not that they'd actually tested it. As I recall, the database experienced about 1 database corruption per day where a constraint would have prevented it.

>
> just my 2 cents

You under-charge. Received on Sat Feb 23 2002 - 06:55:06 CST