| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
 |
 | |||
| Oracle FAQ | Your Portal to the Oracle Knowledge Grid | |
|
| |||
Home -> Community -> Mailing Lists -> Oracle-L -> RE: SQL*Net message waits
Robyn,
Ok, of a 30 minute window, 0.05 seconds total was attributed to SQL*Net message waits. How many total SQL*Net waits were in the trace? I'm a little concerned because with the scoping error you admit is present in your tracing, it's going to obfuscate the solution a bit. I'm also a little concerned that "the users can't tell me exactly which actions they believe are slow....". That statement tells me your system has no real measurable SLA. Without that, how will you know when you've succeeded? Ok, ok, I know, it wasn't your idea but you're stuck with it. Which business process is the most business critical and performing the worst. Start there. Try to get an accurately scoped trace of that process. This will reveal much in determining where the problem is.
Now, going back to the trace you already got: What does the profile of the other 29 minutes 59.95 seconds look like? How many database calls are in the trace file? Database calls are PARSE, EXEC and FETCH. Also, how much CPU time is represented in that 30 minutes? How much wait time? What events were you waiting on?
Also, consider what's going on the in the O/S. Are the CPUs hammered? Are there lots of processes waiting on I/O?
Hope that helps,
-Mark
-----Original Message-----
From: Robyn Anderson Sands [mailto:robyn_at_iTeamTech.COM]
Sent: Tuesday, March 16, 2004 3:40 PM
To: oracle-l_at_freelists.org
Subject: SQL*Net message waits
Hi all,
Performance concerns have been raised for a particular application. The users can't tell me exactly which actions they believe are slow, but it has been suggested that the problem may be the connection between the app server and the database. This was a fairly small app, with all components running on the same server. When we upgraded, the database went to a Unix server, and the other app process were split across several Win2003 servers. App servers are in the same rack, DB is connected with a gb line.
I ran separate 10046 traces against the most active sessions on a typical afternoon. Connections are pooled, so several different user sessions are represented in the trace files. The SQL*Net messages from the client include a range of wait times, but since the sessions included periods of inactivity, I'm considering this user think time, not wait time. 99.9% SQL*Net messages from the database back to the client are sub .00 seconds, a very few took .01 second. Total time for all SQL*Net messages in the traced sessions in a 30 minute window was .05 seconds. I'd like to be able to use this information to show that the connection between the database and the application server is not the performance block. Is this a valid way to use the data or am I reaching here? Can latency be hiding in the 'from client' messages that wouldn't appear in the 'to client' messages?
There appear to be too many database calls; still trying to investigate that, but it's a vendor app and we have limited ability to change their code. Performance issues seem to be related to the client workstation size, etc. but I need to prove that the issue is not between the app server and the db so we can move on and find the real problem.
TIA ... Robyn
FAQ is at http://www.freelists.org/help/fom-serve/cache/1.htmlput 'unsubscribe' in the subject line.
-----------------------------------------------------------------
----------------------------------------------------------------
Please see the official ORACLE-L FAQ: http://www.orafaq.com
----------------------------------------------------------------
To unsubscribe send email to: oracle-l-request_at_freelists.org
 |
 |