Hotsos Symposium 2008 — Still On

The symposia is still ongoing and my head is slowly filling up — relieved from my presentation, finally, I’m able to focus on others’ sessions.

Yesterday, Tanel Poder presented his new tool Sesspack and his integration with Excel. 3 years ago, I created a similar tool and collected session waits and statistics transformed into differences per the interval and organized in the star schema to simplify the analysis. I tried to write a front-end in PHP — it was taking ages and I didn’t have time. Then I tried APEX (HTMLDB 1.6 back then) and it wasn’t flexible enough. I ended up querying the data directly and copy & paste to Excel where I could use pivot charting. What a great feature of Excel — it let me organize the data easily and visualize the problems to management, system and storage administrator and other DBA’s. I moved on and didn’t have time to continue this project, clean it up and put into public domain. I’ve still had it in my mind but there is no need now since Tanel already did far better job. He put the first version of the Sesspack on his web-site about half a year ago. What excited me more this time was the integration with Excel that he did — what a powerful but simple tool in the hands of a smart DBA. I’m looking forward to use it when it becomes available on his web-site.
Update: Tanel put the material on his web-site here.

Since I went to Tanel’s presentation I missed the first half of Tapio Lahdenmki’s two hour talk on “Rethinking Your Indexes”. Since taking only the last half didn’t make much sense, I took a chance to update the blog. After lunch, I went to “RAC Performance Taxes” by Mike Erwin — he focused a lot on parallel processing in RAC. I myself didn’t have enough experience in this area (my preferred approach is to parallelize within single node only) so it was interesting to listen and learn several interesting things.

Wolfgang Breitling did very god job with his presentation “Active Statistics”. A lot of insights into statistics gathering options and emphasized lots of new features in 11g. I was particularly impressed with his finding that sample size AUTO works very quick in 11g and it seems to provide excellent result most of the time providing results of 100% sample size most of the time. Apparently, Oracle completely redesigned AUTO sampling approach in 11g — it’s not just automatically chosen percentage anymore. Another highlight was how to avoid recollecting statistics when adding new partition and how to transform statistics in a smart safe way so that you can take existing statistics and apply it to the new partition. Excellent stuff!

The educational part of day finished with the second part of Hotsos RAC Campground. Interesting discussion again but it went way over the schedule — I left the room one hour after the schedule. My favorite part was when one guy walked into the room when we almost finished and asked a question — “why would I need RAC?”. The answer that anyone seemed to agree on was — “unless you *know* why, you don’t need it”. In this example, systems were stable, running on the huge Solaris box with lots of capacity and no availability problems — excellent case when you definitely don’t need RAC. However, he was very persistent with his question and it seems just couldn’t get it — he was keep asking it for another 30 minutes dragging the panel off the schedule. Apparently, his management knows better than him and keep running with the idea of implementing RAC. Well, life is life I suppose.

There was the Hotsos dinner yesterday evening and it was good chance to relax and talk about something outside Oracle and IT. The dinner finished with dancing but the party was going on far beyond the dinner. I saw James Morle playing guitar and as far as I can say — he did pretty well (what do I know?). It seems that some hotel guests were not as much impressed as me but… life is life.

I woke up at 7am this morning and felt quite fresh — I didn’t really expect that. I was able to grab some breakfast but managed to be 10 minutes late to Joze Senegacnik session on partition pruning. He managed to cover a lot of material in one hour and I liked few 11g new features he covered, particularly, information on Bloom Filtering used in 11g very often instead of old partition pruning mechanism. Well done Joze!

I enjoyed Robyn Sands’ presentation — An Industrial Engineer’s Approach to Oracle Management and Performance. It was her first Hotsos presentation and I know she invested lots of efforts and sleepless night. Well, it was definitely worth it and the presentation went great. I realized that I’m using quite a few techniques without even thinking about it — I guess that’s my engineering education kicking in. ;-) I especially liked the part about variances and we can introduce variances of the conditions in the testing to assure that we choose the best option suitable for the possible range of conditions. Interesting ideas. Thanks Robyn!

I was sitting in the back of the room during Bod Sneed’s talk on capacity planning and writing this blog post. That means I didn’t really focus on the slides but grasped few interesting points and I also have a new book to read which I ordered right away — Are Your Lights On? Thanks for reference Bob!

Oh… I’m missing lunch — I’m 30 minutes late. I better rush to grab something — the next one on my schedule is Kyle Haley with his session on “Average Active Sessions: A Simple Solution to Complex Performance Data”. I’m actually sitting next to him right now — he is polishing the GUI which looks fantastic to me! I’ve gotta run now — more details to come — stay tuned!