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: lots of waiting on 'db file parallel write'

Re: lots of waiting on 'db file parallel write'

From: peter <p_msantos_at_yahoo.com>
Date: 24 Feb 2006 19:02:09 -0800
Message-ID: <1140836529.570685.203430@e56g2000cwe.googlegroups.com>


The application I maintain is not an OLTP database nor a warehouse..it's sort of in between with
large data uploads and large selects going on. Most of the data access in our system is Index Fast Full Scans and full table scans.. (we hash join large data sets ). I'm not sure I understand what you mean by thrashing buffers...

By looking at AWR reports, my buffer cache ratio is never about 92%, but at the same time there aren't lots of waits for free buffers either...only when the db writers get hung up. The waits on undo blocks seems insignificant compared to everything else..

As far as the size of my SGA .. we generate histograms on a large part of our data..which is constantly changing distributions based on regular data uploads..so the SGA needs to keep lots of information in the shared pool. On average the memory usage of the shared pool is always above 93% so even that leads me to believe that my shared pool could probably benefit from being slightly larger... I've looked at AWR reports where the INDEX namespace has had as much as 15% misses for index information in the shared pool...

Our system is very much IO bound...it's always been, and we are in the process of making some data model changes to reduce that, but I still can't make sense of the db writers...

Can you explain what you mean about v$bh and buffer thrashing and veritas housekeeping??

thanks
-peter Received on Fri Feb 24 2006 - 21:02:09 CST

Original text of this message

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