Re: Simulate AWR like repository in 9i

  • From: Tanel PÃder <tanel.poder.003@xxxxxxx>
  • To: "ORACLE-L" <oracle-l@xxxxxxxxxxxxx>
  • Date: Wed, 14 Dec 2005 12:51:07 -0600

Hi,

Direct SGA access would probably the most CPU efficient method, but you have to 
account for inconsistent data as you're not using any latching for your reads.

Otherwise, if you are sampling x$ksuse tables, make sure you include the 
restricting predicates on:
1) inst_id - that sampling would be done on each instance locally, otherwise 
you'll get lot's of interinstance calls, have to spawn PQ slaves etc
2) if sampling is needed only for specific sessions, query the x$ table using a 
fixed table index (the indexed x$ table columns can be seen from 
v$indexed_fixed_column) 

Tanel
  ----- Original Message ----- 
  From: rjamya 
  To: Oracle Discussion List 
  Sent: Tuesday, December 13, 2005 5:06 AM
  Subject: Simulate AWR like repository in 9i


  I may have this requirement to implement something that captures session 
details and sessions waits at very short intervals in a 9204 database. AWR 
kinda does that in 10g, but this is no 10g. 

  I am wondering what might be the lightest process that I can run maybe at a 
minute interval? Too many and too frequent selects against gv$ are not good, 
and I understand that. But I have a feeling that this requirement is coming 
down the line. 

  Question for experts ... if write a really (well as light as I can) process 
to get info from x$ksusecst (for waits) and x$ksuse (for session), what should 
I look for as potential problems (any specific latches?). Oh yeah and this will 
be running on a RAC system. 

  Feel free to tell me that this is a bad idea, but if the request comes down I 
will then use those arguments to say why we shouldn't do it. 

  TIA
  Raj 
  ---------------------------------------------- 
  This space is available for rent.

Other related posts: