OOW2K6: Sleepy Sunday

The first presentation of Open World for me was Tim Gorman's talk on Scaling to Infinity. Obviously I didn't read the abstract properly because this took me by surprise. Instead of being a way to gently ease myself into the spectacle Tim's talk gave me some very useful insights that are immediately relevant to my current project. "All databases are victims of their own success" was especially resonant. Tim's key point was that we can only scale VLDWs if we design them properly from the startpoint. In particular we need to avoid any heavyweight DML (i.e. UPDATE and DELETE) and use partioning smartly. One of my pending tasks is to improve the elasped times of queries in the searching datamart. I think in the light of Tim's talk I am going to have to revisit my implementation (which uses bitmap indexes and partioning) as I'm not sure it's going to scale to infinity.

After that it was up to the icy wastes of the Hilton's sixth floor to watch "Tiger" Mark Rittman dicing with death. Or at least exploring the slapstick possibilities offered by the trailing radio mike's cables. Beforehand Mark had expressed a concern that he wouldn't be able to fill the ninety minute slot on the new BI Suite EE. But in the end he ran out of time, with material still to do. Who would have thought Mark could be so talkative? I had better quickly say that Mark's presentation was coherent and informative and didn't feel incomplete. And he coped gracefully with the inevitable failure of one of his demos.

In the evening the ACEs dinner was nice, although I drank too much wine and didn't eat enough food. I almost got stuck with the bill after the manager mistook me for Andrew Carr. Well, I thought she said "Are you Andrew Clarke?". It doesn't help that we are both balding, bearded and British. Fortunately I eventually managed to introduce her to the right Brit, otherwise I'd be there still, washing dishes. I ended up seated next to Steven Feuerstein who wanted to know whether my presentation was going to focus on utPlsql (which it does). He thinks Code Tester (formerly QUTE) is a much better tool and he's certainly convinced me to take a second look.

I was awake again at 02.00am so I went through my presentation once more. Just as well: I ran my scripts and the first test failed with ORA-01438: Value larger than specified precision allows for this column.  I know I've joked about breaking my demo but this wasn't funny. Eventually I figured it out. The EMP.EMPNO column is NUMBER(4,0) and I had simply run the code so often that my EMPNO_SEQ had clocked 10000. After I dropped and recreated the sequence the scripts ran fine.