Re: Where is Oracle’s Grid ?

"Daniel Morgan" <damorgan_at_x.washington.edu> wrote in message news:1072796803.85192_at_yasure...
> > What is the COMMON thread in all of the above? I HAD access
> > to the code and design and I COULD change it.
> > Not ONE of the Oracle tuning toolsets in all those revisions
> > would have helped me one bit had I not had this.
>
> How can Oracle give you access to compiled code they didn't write?
>
> > And that is the gist of my claim: the tuning toolset from
> > Oracle is USELESS until 10g for anyone WITHOUT access
> > to source code or the ability to influence changes in
> > said code. Which is 90% of the production DBAs
> > out there. Therefore, the existing toolset is useless to
> > 90% of the production DBAs.
>
> If you think 10g, or 100z, is going to give you access to compiled code
> written by another vendor you are begging for disappointment. No product
> other than a decompiler will do that.

Well you'd have thought so. However in at least limited circumstances (and from 8i) Oracle can intercept the sql that the application sends it, and rewrite it and execute the rewritten code, this is all that is needed (mostly). Those cases are stored outlines and materialized view query rewrite. The tuning tool that leccotech sell (can't recall the name) automates the 'tuning' process of trying hundreds of possible ways of rewriting the queries to select the 'best' one. So it is certainly *possible* for Oracle to provide something like Nuno suggests, at least in theory by preprocessing the sql passed from the app and running an alternative formulation that it expects to perform better.

One problem is likely to be the cost of so doing, especially for queries that perform well already (probably 80% of them). Another problem would be performance tuning in the cases where this process failed - how do you tune if you can't tell ahead of time what will get executed? Backward compatibility and 3rd party support would be issues as well.

-- 
Niall Litchfield
Oracle DBA
Audit Commission UK