MOOW 2K7 – Finishing up

After surviving the water slides the day before my colleague
Gerwin and I enjoyed again a great breakfast and a lot of coffee. For me this
is an essential part, to get focused (the getting coffee part)… Almost all of the participants started
the agenda on this Saturday with Jonathans Lewis presentation “Understanding
Statspack”.

Jonathan started of with the essentials regarding the "why using
Statspack in the first place". How does one troubleshoot and create awareness
that all tuning methodologies are incomplete, also statspack. Most of the methods out there,
especially statspack are database centric. So if problems are at hand, statspack will also give
an incomplete picture of questions like: What is working too hard? What are we
waiting for?

The general silver bullet of using a 10046 trace doesn’t always
give you the complete overview, especially about the “what’s wrong” part. I
learned that, regarding statspack, there are some things that I overlooked over
the years, like that the statspack report can also be a means of doing some
trending which can be very helpful. When using statspack you should know your
code, understand your Oracle bit and understand the outcome of performance test cases
before hand and/or understand the results afterwards. This would lead to a
better general understanding of what is happening in your system and statspack
can help you achieving this (if applied properly).

My choice for the second slot was Mogens’ and Anjo’s “TBD”
(to be determined) presentation. I had already seen Cary Millsap’s presentation
during Hotsos 2007, so I skipped it because I was curious about Mogens Nørgaard’s presentation. Anjo
and Mogens session contained a lot of more practical information about
hardware, design and the lack of… At the end an open discussion was started in
which Toon Koppelaars was invited to share his view about how to deal with
database independent data store thinking like J2EE / Java framework evangelists
nowadays tell their disciples. Mogens also launched his idea about a 3 month
boot camp at Miracle HQ learning the real DBA job based on Mogens most preferred / classic books
like Cary Millsaps “Optimizing Oracle Performance”, Toon and Lex de Haan’s new database
design book “Applied Mathematics for Database Professionals“.

After the (again) great Lunch (indeed with a capital L), I
joined Graham Wood’s last presentation “The DB Time Performance Method”. Graham’s
presentation is interesting, because the concept “DB Time” is the underlying concept
of the ASH and ADDM methods. Average Active Sessions, as can be seen in the
Oracle Enterprise Manager Performance panel, is defined by: Average Active
Sessions = DB Time / Wall Clock Time. As Graham told his audience, the term “DB
Time” was in the end the best descriptive regarding describing “the time spend
by an active oracle session” per time unit (thus avoiding the time per time unit description).

To wrap it up. 

I liked the atmosphere. As far as I know now, the atmosphere makes
Miracle Open World (or whatever the name will be next year) unique among its “rivals”.
Most of the presentations were of high standards. Though some of them had problems
with the 1½ hour time slot, being originally created for only 1 hour time slots. This
extra time was sometimes filled with open discussions, which on itself is not a
huge problem, if people are prepared for it. Unlucky enough, this sometimes
also resulted in an early coffee break.