Oracle FAQ Your Portal to the Oracle Knowledge Grid
HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US
 

Home -> Community -> Usenet -> c.d.o.server -> Re: 2.5 billion row table scan...

Re: 2.5 billion row table scan...

From: Big Al <db-guru_at_att.net>
Date: 2000/06/05
Message-ID: <393C1325.36307534@att.net>#1/1

sergey_s_at_my-deja.com wrote:
>
> Due to time constraints, we may have to stick with scanning 2.5 billion
> row table (and figure out another access path later). It's partitioned
> (each partition about 50-60 mil rows). Current memory is at 4 Gig, and
> there are six processors (Sun machine). We will probably double the
> memory soon and add two more processors.
>
> With it's current configuration, however, it was still cheaper to scan
> the table than use any indexes, and it takes roughly 18 hours.
>
> Can you suggest anything to speed up the scan?
>
> Thank you!
>
> Sergey

A few questions come to mind. I assume that the table is partitioned by Oracle, but can you verify that? What's your partitioning key? Can you make it more atomic so you have more partitions? What version of Oracle are you running? Have you heard of the Parallel Query Option (PQO)? If you have it, what are the values for min and max parallel servers in the init.ora. Is your disk striped to spread out the I/O? More information would help someone to help you.

Big Al Received on Mon Jun 05 2000 - 00:00:00 CDT

Original text of this message

HOME | ASK QUESTION | ADD INFO | SEARCH | E-MAIL US