Comments

• 

  Frank Petersen March 13, 2012 at 7:37am

  Hi John !  Yes, the 401 record it is indeed. What I see is a *big* difference in the numbers collected in the 401 record compared to the DB2 accounting record. I have made an assembler program that collect this and anchor it to the program in our performance datawarehouse. The intention was that the programmer from our intranet can look at the performance for his/her program (this is what they do today) and now they should be able to zoom into the program and see that timing broken down to statemant numbers. However it has proven impossible to correlate the 401-numbers to the "good old" accounting numbers.I have an open PMR on this issue but the comments from the service team point to "working as designed" (perhaps an early conclusion, sorry).

  It is also very difficult to keep track of changes in the program because the numbers contains the accumulates timings. So if a program has a very bad accesspath where statement number 4711 consumes 99.99% of the time and a rebind suddenly makes it the most cheap one it will not reflect itself in the 401-records. It would have been nice to be able to "reset" the timings at rebind...

  Thanks !
• 

  Frank Petersen March 12, 2012 at 7:53am

  Hi John !

  I think the new DB2 10 statement level monitoring deserves a detailed walk through. It seems to act quite differently than what I thought before I started worked with it. I can give you some input if you need it.

  Best...

  Frank Petersen

  Bankdata/JN Data

   
• 

  Cristian Molaro March 10, 2012 at 9:35pm

  Hi John,

  I would suggest:

  1 - Differential cost of starting the same trace but with a different destination (SMF / GTF)

  2 - What is more expensive? GTF vs OPx vs SMF?

  3 - Are there traces that are tight together? For instance, are we missing anything when using EXCLUDE filters?

  4 - What about getting some traces started when there is a specific event? For instance, I'd like to have some performance traces set ON only when a specific event takes place.

  Thanks a lot in advance.

  Looking forward for such a paper.

  Best regards,

  Cristian Molaro