Trace Point qe90

From: dori!wiz@uunet.uu.net (Karl Schendel)

On Apr 19, 11:22, Lou Chmielewski wrote:
> Subject: How do I read the output from trace point qe90?
> ...

very briefly -- it looks a lot like a QEP output except the labels are
different.  for instance, an FSM join might be labeled an SIJOIN in the qe90
output.  sometimes the qe90 output will have extra nodes, or will combine
nodes -- you have to be flexible :-)
Each node shows:

          ----------
          KJOIN
          at       2  --> actual tuples
          et    2093  --> estimated tuples

          ad     124  --> actual disk i/o in mystery-units
          ed     151  --> estimated disk in mystery-units

          ac       0  --> actual CPU, see below
          ec    2637  --> estimated CPU

          et       1  --> elapsed time
          ----------

the numbers are cumulative up the tree, so you have to subtract to get
the time for any one node.  The actual CPU number will be zero unless you
do something to turn on CPU accounting for your session; for instance,
a SELECT DBMSINFO('_cpu_ms') will do it.
The estimated numbers should match the SET QEP output.

-- 
Karl Schendel            Phone: (412) 963-8844
Telesis Computer Corp      Fax: (412) 963-1373
wiz@telesis.com



From comp.databases.ingres Mon Apr 22 15:04:27 1996
From: Roy Hann 
Subject: Re: How do I read the output from trace point qe90?

On Fri, 19 Apr 1996, Lou Chmielewski wrote:

> 	I've been looking at a query written 2 different ways.  The qep's are 
> identical, however, I ran the 2 with trace point qe90 and received different 
> output.  What I'd like to do is find out how to read the qe90 output.  Any 
> help would be greatly appreciated.

Unless you flushed the DMF cache between trials then you should expect to 
see differences because the second attempt could get a lot of cache hits 
on data loaded by the first trial.  Try setting trace point DM421 between 
trials to flush the cache.  (Maybe don't do this on a busy production 
system if you mind the resulting possible hiccup in performance.)

If you do this, expect identical query plans to result in identical qe90 
output.

========================================================================
Roy Hann                                   rhann@tnc.com
BKB Engineering Ltd.,                      NAIUA Member-at-Large
11211-76th Avenue,                 
Edmonton, Alberta                          http://www.tnc.com/bkb
T6G 0K2                                    Tel: (403)497-2409
Canada                                     FAX: (403)436-6055
========================================================================
Ingres Q & A
To William's Home Page
Email William