Re: separate data/inidex

>
> C) They follow the advice they receive from Oracle Corporation.
>

Which is generally good practice.

> It feels a bit uncomfortable when you criticize people for doing exactly
what the
> documentation produced by Oracle Corp. tells them to do. It is unrealistic
to
> expect developers and DBAs to go against advice Oracle Corporation itself
> publishes; or to expect them, on their employers dime, to validate each
and every
> assertion Oracle makes.
>
> For example:
>

http://otn.oracle.com/docs/products/oracle8i/doc_library/817_doc/server.817/ a76956/create.htm#998233
>
> Where it states:
> ===================================================
> Distributing I/O
>
> Proper distribution of I/O can improve database performance dramatically.
I/O can
> be distributed during installation of Oracle. Distributing I/O during
installation
> can reduce the need to distribute I/O later when Oracle is running.
>
> There are several ways to distribute I/O when you install Oracle:
> Redo log file placement
> Datafile placement
> Separation of tables and indexes
> Density of data (rows per data block)
> ===================================================
>
> A reasonable person looking at the above quote would see "Separation of
tables and
> indexes" and infer from that that they should separate tables and indexes.
>

True enough. A reasonable person might well draw that conclusion. They might also draw the conclusion that rollback segments can be housed anywhere you fancy, since that particular quote is mysteriously silent on the subject -yet we all know (I hope) that rollback I/O is a potential killer unless housed in its own tablespace/datafile/disk.

I guess the point is that documentation is fallible, as are the people who write it. Provided the message gets through various forums [books, here, magazine articles, FAQs and so on] that point three in the above quote is not strictly necessary (though, let's face it, it does no harm to split the two, and has (as I keep saying) considerable merit from the point of view of backup and other management), I don't think it particularly fair to criticise Oracle for the odd line here and there in what is after all an enormous quantity of documentation of a complex product.

Regards
HJR
> I don't have any objection to being advised not to do this blindly. In
fact it is
> extremely valuable advice that is very much appreciated. And I am sure I
speak for
> many when I say "Thank you" for correcting the misinformation.
>
> But to criticize people for following advice given by Oracle seems a bit
unkind.
>
> Daniel Morgan
>
Received on Tue Apr 30 2002 - 19:04:31 CDT