Re: Parallel Queries

Salut Victor,

"Victor Kaminsky" <pleasenospam_at_nospam.com> schrieb im Newsbeitrag news:3b137bc0.2401533_at_news.newsguy.com...
> Hi,
>
> I have one point to discuss, please help me if you have experience or
> correct if I am wrong.

I don't have this experiance but i think you may try the following:

>
> I am currently working with a huge database (about 1.5 billion
> records).
>
> The database runs on a powerful Sun machine, has tons of RAM, Disk
> Arrays, other wonderful stuff and costs a zillion dollars.
>
> Currently, parallel queries are active and, every time I run a query,
> the system creates 22 different sessions to gather data quickly and
> efficient.
>
> We have used partitioning, paralellism, etc and I can say that the
> database is optmized to run about 90% of the queries.
>
> However, some queries take a long time, no matter how good is your SQL
> statement (explain plan).

Very BIG!!!

>
> Since most queries have date periods as main criteria, I was wondering
> if it would be a good idea to create a different connection and, for
> example, have 2 different connections working on different date
> periods at the same time.

Did you try to change the table definitions? Try to use a integer or long instead of a date field, or a field type long for the date part and a field type integer for the time part. perhaps a udf for date/time conversion?
Comparisation on numbers normaly are much faster. Create views for those who want to see a date field.

>
> I would be responsible for joining the results and produce a
> consistent table or view (probably by using Access). The common sense
> leads to think that 2 different connections would create 44 sessions
> instead of 22.

I think you have to try this

>
> Since the machine has some idle power, I think it would double the
> parallelism.
>
> It it correct? Has anyone tried that before?
>
> Thanks in advance,
>

Greetings
Matthias Received on Mon May 28 2001 - 16:46:28 CDT