Re: Help me read my TKPROF output

From: craig sivils <casivils_at_lescsse.jsc.nasa.gov>
Date: 15 Jan 93 21:46:07 GMT
Message-ID: <casivils.727134367_at_node_508ba>


In <C0v5Gz.LnM_at_inews.Intel.COM> kortikar_at_mipos2.intel.com (Aniruddha Kortikar) writes:

>select * from vendor where vendor_name like '%ALD%'

> count cpu elap phys cr cur rows
>Parse: 1 0 0 0 0 0
>Execute: 1 0 0 0 0 2 0
>Fetch: 1 0 0 0 3 0 1
 

>Execution plan:
>TABLE ACCESS (FULL) OF 'VENDOR'
>============================================================================
 

>1. cr means buffers recieved in consistent mode. (why is this > 0 even though
> there is nobody using the system, hence no update, mod or select on vendor
> (why all of it is not retrieved in cur - current mode ?)
 

>2. what does 2 in cur-Execute mean ? (what does execute phase do in select ..)
 

>3. why phys 0 everywhere ? does it mean it was all in memory ? (count(*) is 150
> so with 16 MB memory that should be possible. - SGA is much smaller though!)
 

>will anybody provide me with detailed meaning of all these terms.

#1, If your process altered any blocks (either your tables and/or the

    system tables through some operation) then the cr buffers will need to     be read, this is not unusual.

#2 Means it took oracle 2/100 sedons of cpu time to figure out which rows

    satisfied the querry criteria.

#3 Most likely you do not have the kernel timing parameter turned on so

    you are getting 0's rather than the real data.

                                  Craig 
Received on Fri Jan 15 1993 - 22:46:07 CET

Original text of this message