event waits in 9.2 OEM

From: Joel Garry <joel-garry_at_home.com>
Date: 14 Jul 2004 11:10:26 -0700
Message-ID: <91884734.0407141010.26f37a9e_at_posting.google.com>



All:

I'm looking at a performance issue with OAS10g, and I'm using OEM standalone to look at top wait events (to verify something support told me about performance and local v. remote data sources). 9.2 OEM to 90151 db, W2K.

So OEM says the top wait events are SQL*Net more data from dblink (about 9 on the bar chart), followed by library cache pin (about 3).

But I see this:

SQL> select event, total_waits, time_waited, average_wait, max_wait, time_waited_micro from v$session_event where sid = 70;

EVENT TOTAL_WAITS




TIME_WAITED AVERAGE_WAIT MAX_WAIT TIME_WAITED_MICRO
----------- ------------ ---------- -----------------
latch free

    1

          0            0          0                98

log file sync                                                         
    3
          0            0          0              1652

library cache pin                                                     
    4
          3            1          3             32135


EVENT                                                           
TOTAL_WAITS


TIME_WAITED AVERAGE_WAIT MAX_WAIT TIME_WAITED_MICRO
----------- ------------ ---------- -----------------
SQL*Net message to client

   37

          0 0 0 171

SQL*Net message to dblink  2127

          1 0 0 5619

SQL*Net more data to client

   15

          0 0 0 3022

EVENT TOTAL_WAITS




TIME_WAITED AVERAGE_WAIT MAX_WAIT TIME_WAITED_MICRO
----------- ------------ ---------- -----------------
SQL*Net message from client

   36
     151800 4217 35818 1518001237

SQL*Net message from dblink  2127

      31349 15 41 313485659

SQL*Net more data from dblink

   87

         14 0 2 137806

EVENT TOTAL_WAITS




TIME_WAITED AVERAGE_WAIT MAX_WAIT TIME_WAITED_MICRO
----------- ------------ ---------- -----------------
SQL*Net break/reset to client

    6

          0 0 0 970

10 rows selected.

SQL> I guess OEM is displaying just the wait events between 5 second data refreshes? (The help says since the session started). It seems odd that it wouldn't catch those other, longer waits. So is OEM bogus here? Or does it just not consider sqlnet messages worthy of consideration? Are they? Timed_statistics is true.

jg

--
_at_home.com is bogus.
"...so sloppy, it's devious." 
http://www.theregister.co.uk/2004/07/14/atak_stealth_virus/
Received on Wed Jul 14 2004 - 20:10:26 CEST

Original text of this message