Oracle OpenWorld 2009, the first two days

This year I'm visiting Oracle OpenWorld for the second time. Everything is now familiar, except for two differences. First one is the location of Oracle Develop, which has moved back to the Hilton Hotel, instead of the nearby Marriot Hotel. It is a nice walk, but sometimes you cannot chat after a session if you have a next session at the Moscone Center. Second difference is that it seems to me less people are attending the conference than previous year. But it's still a lot of people. One thing hasn't changed: my excitement about the conference.

The first two days were very busy. I'll quickly describe the sessions I went to. Sunday started early at 8:30 AM.

Alex Gorbachev - Demystifying Oracle Real Application Clusters Workload
A topic that with which I am not too familiar: my shop doesn't use RAC and I'm not a DBA. So I thought to use this one to get to know a little bit more. I had some trouble getting the context right in the beginning of the session, but when Alex started demonstrating the stuff on his laptop, everything fell into place. The demos were simply great. The sometimes complex load balancing algorithms were made perfectly clear by showing exactly the right things. A good start of the conference.

Steven Feuerstein - Why you should care about PL/SQL in Oracle Database 11g Now!
This was the first time I attended a session by Steven Feuerstein. Although I heard nothing new about PL/SQL, it was great to see how he presented. Very interactive and with a lot of humour.

Dimitri Gielis - Mastering an Oracle Application Express Page
In this session we learnt about what happens when APEX shows or processes a page. The session really helped to get some of my thoughts straight. One of the new things I heard was a simple button on the page design screen called "View" where you can see the order of the events that will happen. I saw that page lots of times now, but never bothered to use that.

Andrew Clarke - Designing PL/SQL with Intent
This session was totally different than I expected. It was a rather philosophical one comparing design patterns in PL/SQL with design patterns in town planning. At the end a few PL/SQL examples were shown that I liked a lot, since I preach them myself often as well, although not as well as Andrew did.

Michael Hichwa & David Peake - Oracle Application Express 4.0
A lot of nice new features were presented from the APEX development team themselves. The ones I liked most were Dynamic Actions, with which you can create client side validations declaratively, and the APEX listener. I'm not entirely sure, but it looks like the APEX listener opens up easier and supported ways to have proxy database authentication. Other new features include integrated Team Development for feedback and bug reports and the like, improved charts, websheets applications and improved tabular form handling.

Bryn Llewellyn - Online Application Upgrade
This session was renamed to "Edition Based Redefinition". Bryn very concisely talked about the 11g Release 2's killer feature: editions. Without using a demo he managed to make it perfectly clear what the feature is about and what it is not about. A few random things I wrote down: from now on an object is not defined by owner and name but by owner, name and edition. Only synonyms, views and PL/SQL of all kinds are editionable. You need to rename the table, create an editioning view with the same name as the table's former name. Then move all triggers, all privileges and all VPD policies to the editioning view. And last but not least: in 11gR2 it's now possible to add an extra column while the table is in use, whereas previously you would get an ORA-04021 lock timeout message.

Scott McNealy - Extreme innovation
This keynote was mainly aimed at assuring all Sun customers that Oracle is not trying to kill Sun's products but instead Oracle is increasing the investments in Sun's products. Larry Ellison did a good job at depicting IBM's range of hardware as inferior, slow, and not very green. He even offers $10M if your Oracle application doesn't run at least twice as fast. Now that's confidence!

ACE dinner
An evening at the Sens Restaurant for all Oracle ACEs and ACE Directors, where I've met a lot of old and new friends. Thanks to Lillian and all others at OTN for the perfect organization and a nice evening!


On Monday I attended these sessions:

Jonathan Lewis - How to become an expert
A very entertaining story about what you should do to become an expert. My rather free interpretation of the answer to this question is, to learn and really understand the basics first. If you understand the architecture of version 6, when you could still carry the documentation with you, then everything afterwards is really just a variation on the same theme. Understand exactly what Oracle will have to do when you send it a SQL statement and you'll be in a good position to know where the root cause of your problem is.

Tom Kyte - All About Metadata: Why Telling the Database About Your Schema Matters
A good story about why should have constraints, even in data warehouses. They may slow down DML a little, but it's extra information for the cost based optimizer to make smarter decisions. Extra access paths will become available. A new thing for me was to see the effect of creating a dimension object in combination with materialized view query rewrites. I have to see that one for myself one time.

Greg Rahn - Chalk & Talk: The Core Performance Fundamentals Of Oracle Data Warehousing
A story about Exadata and data warehouses and benchmarks. A different world than my usual one, which was nice. For example, this exadata machine is so fast, Greg recommends to not even use indexes anymore in most data warehouses. Of course, it all depends. I must say I was impressed by all numbers he mentioned. Those are not numbers I usually see in my clients OLTP database.

Jonathan Lewis - How to Hint
Again a very good presentation in which I heard quite a number of new things. To name a few: use_hash is only half a hint; it's better to combine this hint with the swap_join_inputs hint to explicitly state which query block should act as the in-memory table and which one as the probe table. Jonathan also had a great example of the no_merge hint, which finally became clear to me. And the parallel hint was not what I thought it would be. It's not instructing to always go parallel, but rather: if you are full scanning the table AND it's cost is lower than the serial plan, only then go parallel. Which means there might be a day your query will run serial even if it ran in parallel before, or vice versa. In this session I also learnt how an alien looks like ...

Chen Shapira - Visualization Session - Make your charts more effective for troubleshooting and presenting
This one definitely categorizes as "something completely different". Although the running example was about duration of exports, the session was about how to use graphs effectively and how they can effectively support your message, or not. Luckily I knew just enough about statistics to understand all graphs. I especially liked the Q-Q one.

Lucas Jellema - Continuous Application Application Evolution with Oracle Database 11g Release 2
Exactly the same topic as Bryn Llewellyn's session yesterday. Lucas also had a clear story and even included a few live demos that really helped. A very well explained introduction. The only unfortunate thing about both sessions is that no downsides were mentioned. And there must be several. Like test efforts being multiplied or worse performance because crossedition triggers are row level triggers. And we know what that means ... But still I liked the session for being very clear and the nice metaphores.

OTN Night
Here I just grabbed some food, but I could not find any familiar faces so I decided to leave early and catch some sleep. Which doesn't work by the way: I'm still awake each morning well before 4AM. Which is not uncommon for Europeans, it seems ...

More to come later.