HOTSOS 2K8 – About Filling the Gaps in Performance Tuning
After a quick breakfast, the Hotsos symposium got on has way with a small introduction of Gary Goodman on reviewing the history of the symposium event. This time also some symposium campgrounds are held instead the annual panel sessions. Four symposium campgrounds will be given at the end of the day regarding: SQL, Instrumentation, Internals and RAC. The Keynote of Cary Millsap gave a good review of Cary’s expected enhancements in performance tuning.
The keynote session was based on a presentation once originally given in 2003, besides "the story of Bob, bob’s boy and the bicycle that became a Ferrari", he compared the advancements and lacks between performance issues (and how it should be handled) between 2003 and its current state.
Toon Koppelaars presentation hugely impressed me again and being based on his (and Lex de Haan’s) book, "Applied Mathematics for Database Professionals", it only points out that it is time to spend time to read the book. Also his presentation showed what Semantic Query Optimization can do for you regarding performance optimization. If applied and supported by the Optimizer logic (and probably applied within a defined scope), the impact can be huge, because, as clearly demonstrated by Toon, the optimizer can deduct the answer before hand, without touching any data (and be aware that there is now magic involved here, but basically determined by the environment constraints already known in the database.
This also makes me wonder again, what would happen if more business rules / constraints were NOT set / known, in the middle tier but also in the database back end where CPU and hardware resources could have been better utilized if information was knows as well in the database tier.
Another session a really enjoyed was Tanel Poders presentation called "Advanced Oracle Troubleshooting", I was very interested from the start, if not only for the fine examples he demonstrates on his blog site. Tanel demonstrated several demo’s and his approach how to figure out what is actually going on, even given the fact that the normal standard routines will leave you with no answers at all (as he clearly demonstrated).
Via his systematic approach he led us beyond the database internals into the realm of OS calls and their breakdown. I liked the approach very much because it follows the flow of processes as they are handled in each level of the architecture using the proper tools for each level. It is also very appealing to me because that is more or less the way I try to solve these kinds of issues since a long time, but this time even further at a lower level. Also I was happily surprised that, doing it this way, the needed knowledge for each level is manageable / doable.
Gerwin did a good job with his presentation "Another Way of Thinking". He presented the content regarding GAPP (General Approach Performance Profiling) with a lot of confidence. As said before, GAPP is a performance profiling approach that makes use of easy to obtain performance markers that gives insight in the broad spectrum of a business process / architecture. Although still in concept, and you will have to fulfill to some restrictions, it has the promise to being able to predict architecture behavior and/or find the most intrusive performance drains in the system. Even in nice discussions at the end with Neil Gunther, the presentation withstood the critics. Maybe the story still needs some tweaking but the content stands and Gerwin the a hell of a job during his first presentation here in Dallas.
By the way, I am not kidding, currently it is snowing outside. When we arrived it was still 21 degrees Celsius on Saturday and now on Monday evening the drop in temperature hits 2 degrees Celsius. Cold enough to produce some wet snow…
A good start for Hotsos 2008