Re: Tuning Question

Did you ever change your init.ora parameters? The default init.ora parameters that come with the software haven't been changed in 10 years or so, and they are laughable. Start determining db_block_buffers, if that's still 200, increase it by a factor 10 at least.
One particular problem you could be having in your database, because of the long raw, is chained rows. You should also start to determine whether you have chaining. If you run utlbstat and utlestat you will get the hitrate of your instance, and it will also list the number of chained retrieval requests. Individual tables need to be analyzed, the column chain_cnt in user_tables will tell you how many rows are chained. Depending on the severity of this issue you may consider to recreate your database with a larger blocksize.

Regards,

Sybrand Bakker, Oracle DBA

<pendell_at_hotmail.com> schreef in berichtnieuws 8d4r11$lqi$1_at_nnrp1.deja.com...
> We are developing a large-scale computer simulation which uses an
> Oracle database for storage and retrieval of data. In particular, we
> have a construct called an "Instrument" which, during the course of
> the simulation run, will collect sample data and insert it into an
> Oracle table. Due to the nature of the data, it must be stored in
> binary form. So, our table is of the form:
>
> DATA_KEY (VARCHAR2) | DATA_OBJECT (LONG RAW)
>
>
> Insertion occurs one row at a time, one transaction at a time. That is,
> a "commit" occurs after each individual insertion.
>
> However, we have one small problem: A simulation run without
> instruments takes an hour or two. A run with all instruments turned on
> can take *DAYS* to complete, most of the time spent in Oracle accesses.
>
> I'm about to attempt to tune our Oracle accesses, using
> "Oracle Server Tuning" as a guide. But before I do,
> a question for the group: Do the symptoms I have been describing
> sound familiar? Has anyone encountered anything like this before?
> Does an obvious solution to this problem present itself? If not,
> what areas would you look at first?
>
> And a related question:
> Part of the fix may involve re-writing our application's C-level call
> interface -- where can I find a good guide to this architecture, and to
> interacting with Oracle at this level? The manuals shipped with the
> server are useful, but somewhat non-intuitive.
>
> Many thanks for any assistance you can offer.
>
> Very Respectfully,
>
> Brian P.
>
>
> Sent via Deja.com http://www.deja.com/
> Before you buy.
Received on Thu Apr 13 2000 - 19:59:15 CEST