----- Original Message -----
Sent: Friday, November 16, 2001 10:30
AM
Subject: RE: Internal Benchmarking
I've had some luck with the tools described in Scaling
Oracle8i by Morle. It has and awk script to parse out 10046 event trace
sql and bind variable information and produce a tcl script suitable for
dbaman. Dbaman is an extended tcl shell sotra-likea oratcl. I've
used it to run sql that I have captured by setting event 10046. I turn
on tracing for the system and for sessions already running. After the
time period has elapsed I turn it off and proceed to collect all trace files
into a working directory. I then feed each trace file into the awk
script to produce the dbaman tcl script. I use a separate instance of
dbaman to process each tcl file. I then reload the database from a
golden copy and re-run the scripts, then repeat until I reach my tuning
goal.
A few gotcha's; the current dbaman works with 7.3
libraries. And it has a quirk with the interpretation of the bind
variables. If the bind variable begins with a number then it assumes
that the rest of the characters are also numbers, so it chokes on "123ABC"
when it fires of an execute operation. I dug into the dbaman code, found
the IF-THEN line that was checking the first character and commented out the
entire statement. I rebuilt dbaman according to the instructions and it
worked. I'll post the changes if you are going to pursue
dbaman.
The only thing I didn't try was to simulate the think/latency
time. But that time can be extracted from the tim column in the sql
trace file. I would do this in the awk script that is doing the initial
translation of raw trace to dbaman tcl. The tim value is in 100'ths of a
second but I don't know how to convert it to actual time of day. But the
relative time between statements can be derived from it.
HTH.
Tony Aponte
-----Original Message-----
From: Orr,
Steve [mailto:[EMAIL PROTECTED]]
Sent: Wednesday, November 14, 2001 4:00 PM
To: Multiple recipients of list ORACLE-L
Subject: Internal Benchmarking
Howdy All,
I want to create some database-only benchmarking scripts to
reflect a
typical day in the life of a custom
application. I'm thinking about using
LogMiner to get
the redo and v$sqlarea to derive a representative mix of
queries. Maybe we can also sniff/parse the network traffic to the DB
server.
Since I'm looking only at database activity
I'm not too keen on applications
which merely record
and replay end-user keystrokes and mouse clicks but I
would like to mimic delays in transaction commits due to network
latency and
user indecision or whatever. I'd also like
to be able to increase the load
intensity by factors
of 10 to 1000. Has anyone created any application
specific benchmark routines and can you share some tips on how to do
this?
Any good tools that you have used? Any comments
on Mercury Interactive
stuff? Other ideas?
AtDhVaAnNkCsE,
Steve Orr
--
Please see the official ORACLE-L
FAQ: http://www.orafaq.com
--
Author: Orr, Steve
INET: [EMAIL PROTECTED]
Fat City Network Services -- (858)
538-5051 FAX: (858) 538-5051
San Diego,
California -- Public Internet access
/ Mailing Lists
--------------------------------------------------------------------
To REMOVE yourself from this mailing list, send an E-Mail
message
to: [EMAIL PROTECTED] (note EXACT spelling
of 'ListGuru') and in
the message BODY, include a line
containing: UNSUB ORACLE-L
(or the name of mailing
list you want to be removed from). You may
also
send the HELP command for other information (like subscribing).
