The value of in-house seminars (run by the correct people)
Posted by John Hallas on November 2, 2010
Lately we have been very lucky where I work as we have managed to get Graham Wood from Oracle to pay a visit for the day and we have booked Julian Dyke to present three seminars to us on-site.
Getting the intro to Graham was rather fortuitous as we have another Oak Table member working on site regularly and he put me in touch with Graham. Graham’s role amongst other things is to circulate and present information on what Oracle are doing and in particular around the performance tuning available from within OEM.
He came on site and the idea was that we would give him an hour or so on a system that was obviously new to him and he would have a scout around and then later on he would repeat the process but to a larger audience but this time it would be be much quicker as he already knew some of areas that needed further attention.
Well we (me really) made two mistakes there in that we picked a system which needed work doing on it (which was actually quite intentional) and we got the developers in for that session. It is fair to say that a number of issues around how the application was written were picked up and discussed robustly. The key point was that Graham proved that he could look at a new system using just OEM tools (and we do have the performance packs and are therefore licensed for AWR) and perform some deep and meaningful analysis.
Perhaps the tool that was newest to most of us was the ability of the sqlmonitor tab under performance monitoring. I have shown an example of how powerful it is. Firstly I looked at Top Activity screen and selected a session that seemed to have done some work
Then I drilled down into it (note the green icon on the left shows that it is currently running and then I looked at the IO associated with the session.
The interesting thing to note is the diference between the number of rows the CBO thinks it should be working on and the actual number it found. The third line from the bottom shows the code that is causing most IO and the FTS for the table is estimated to read 246K rows but it has found 32 million rows – a considerable difference.
Later on that day , to a larger audience ,Graham re-worked through that system and selected several others and used the same methodology to highlight ‘interesting ‘ pieces of code.
I must mention that the highlight of the day for most of us was a discussion between Graham and one of our DBAs on the benefits of the Spotlight from Quest v OEM. You can probably guess where Graham stood on that one.
I also managed to get time with Graham to talk about my current pet project, an AWR snapshot database to hold AWR stats from a number of production databases for 366 days.
Anyway, it was an excellent day, we enjoyed having Graham visit and I think he enjoyed getting user feedback on the product and he was certainly very enthusiastic about OEM. The fish and chips went down well at lunch-time as well.
I met Julian at an UKOUG SIG a while ago and when internally we decided that a good way of educating a number of people would be to get speakers to come to us as opposed to sending people away on courses, his name sprang to mind.
I arranged for Julian to do 3 presentations for us, AQ and Execution plan half days and a one day RAC on 11G seminar. The two half-days were held last Friday and we had 13-14 attendees on each seminar. To be fair to Julian the execution plan was much longer than a half day, both in scope and content and so it was rushed through a bit but that was on our insistence. However the course material was very good and I think everyone got a lot from both sessions.
I believe that this type of session, if properly scoped and there are sufficient number of attendees can offer very good value and I would be happy to recommend Julian Dyke to provide this type of event on site.