RE: measuring TPM

2004-01-29 Thread Jamadagni, Rajendra
And these would be CPM (commits per minute) and RPM (rollbacks per minute). If you really want a transaction, you have to code it yourself, otherwise all you can get is CPM and RPM. If your transactions_per_minute ( or commits_per_minute) is low use this handy script to bump it up. create

RE: measuring TPM

2004-01-28 Thread Stephane Faroult
Charlie, I understand a transaction as a succession of SQL statements between two successive COMMITs or ROLLBACKs - you will find inside V$SYSSTAT how many COMMITs and ROLLBACKs were issued. If you are interested, besides transactions proper, in the number of statements executed, then have

Re: measuring TPM

2004-01-28 Thread Mladen Gogala
Well, as you are well aware of, you cannot measure without impacting. I know of the following methods: 1) Turn on auditing, count all transactions from dba_audit_trail table within a day and divide by the number of minutes in 9 hours. That will give you an average TPM number during the working

RE: measuring TPM

2004-01-28 Thread Niall Litchfield
Check out 'user commits','user rollbacks' and (maybe) 'user calls' in v$sysstat. These get collected by statspack so you can plot a chart over time. Niall -Original Message- From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On Behalf Of [EMAIL PROTECTED] Sent: 28 January 2004 15:29

RE: measuring TPM

2004-01-28 Thread Goulet, Dick
Charlie, I use the following to determine this: EXEC SQL SELECT ROUND(VALUE/((SYSDATE-STARTUP_TIME)*1440),1) INTO :tp FROM V$SYSSTAT, V$INSTANCE WHERE NAME='user commits'; Dick Goulet Senior Oracle DBA Oracle Certified 8i DBA -Original

Re: measuring TPM

2004-01-28 Thread Jared . Still
My reply would be something along the lines of A transaction as you would like it to be measured is best measured in the application. I can provide you with IO per minute, broken down into reads and writes, and a number of other statistics. What they are asking for cannot be measured from

RE: measuring TPM

2004-01-28 Thread Post, Ethan
Charlie, What is the perceived relevance of gaining this information? You would be much better off correlating statistics such as overall non idle wait time and database workload (# Users, Ion's/CPU etc...) to actual business functions the database is performing (invoices, sales orders, etc...).

Re: measuring TPM

2004-01-28 Thread Mladen Gogala
On 01/28/2004 12:34:26 PM, Post, Ethan wrote: Charlie, What is the perceived relevance of gaining this information? The information is necessary so that manager and director can make a lovely excell spreadsheet for the VP, who will, in turn, insert it into a slide show for the CIO. -- Mladen

Re: measuring TPM

2004-01-28 Thread Daniel Fink
If that is what this is for, the formula is very simple. TPM = x*42 where x is a number sufficient to justify the really cool hardware system you want. Mladen Gogala wrote: On 01/28/2004 12:34:26 PM, Post, Ethan wrote: Charlie, What is the perceived relevance of gaining this information?

RE: measuring TPM

2004-01-28 Thread Jesse, Rich
Hey Charlie, I made a DBMS_JOB here that runs this procedure every 5 minutes: CREATE OR REPLACE PROCEDURE QT_TX_MONITOR AS -- 06/17/2001 REJesse Created. v_value NUMBER; BEGIN SELECT SUM(VALUE) INTO v_value FROM V$SYSSTAT WHERE NAME IN ('user

RE: measuring TPM

2004-01-28 Thread Thater, William
Daniel Fink scribbled on the wall in glitter crayon: If that is what this is for, the formula is very simple. TPM = x*42 where x is a number sufficient to justify the really cool hardware system you want. as a serious question, is TPM a valid measurement for a database? or are there other

Re: measuring TPM

2004-01-28 Thread Daniel Fink
I would say that it depends on the system. If we are talking about a stock trading system, then TPM is very important as is transaction-time-to-completion. For a data warehouse, this may be absolutely meaningless. Of course, does TPM describe the width of the database pipe or it's depth? In the