RE: explain plan, can you explain this?

From: Yasin Baskan <>
Date: Thu, 10 Jan 2008 14:30:28 +0200
Message-ID: <083667B535F3464CA0DD0D1DAFA4E3760F6673C2@camexc1.kfs.local>

John, are you sure that statspack reports sql statistics cumulatively?

As far as I know it lists them as delta values between the end snapshot and the begin snapshot. The package STATSPACK gets the current values from v$sql and stores them in stats$sql_summary and spreport.sql gets the difference between snapshots. From my spreport.sql from a home:

... lpad(to_char((e.buffer_gets - nvl(b.buffer_gets,0))

       from stats$sql_summary e
          , stats$sql_summary b


-----Original Message-----
[] On Behalf Of John Kanagaraj Sent: Thursday, January 10, 2008 6:09 AM To:
Cc: Oracle L
Subject: Re: explain plan, can you explain this?


> The very interesting part is that in a one-hour statspack, this
> generates 30 mil buffer gets, executed 111,388 times (about 273 buffer
> per exec). When combined with the facts below, it becomes a puzzler:

This is only slightly related to the issue, but STATSPACK's SQL is a capture of V$SQL "as-is" when the snapshot took place. SPREPORT does NOT perform a Diff as it does with other stats (i.e. end_stat - begin_stat). The 111,388 executions is *cumulative* since that SQL was last loaded/parsed. This is apparent if you look at the STATSPACK source at $OH/rdbms/admin/spcpkg.sql.

You should look at Tim G's "sphistory.sql" to determine the periodic progression of execution and other stats for a given SQL.

John Kanagaraj <><
DB Soft Inc (Sorry - not an Oracle blog!)
** The opinions and facts contained in this message are entirely mine
and do not reflect those of my employer or customers **

Received on Thu Jan 10 2008 - 06:30:28 CST

Original text of this message