Hotsos symposium 2006, day 1...
I got in late Sunday night – the flight attendant that should have been in Washington was stuck in NewYork – making our flight much delayed. So I missed the opening event on Sunday night – but was there bright and early on Monday for the kickoff of the week.
After Ray Lanes keynote – the technical sessions began. I attended my session first (of course) and it went pretty well. I like to say that giving a presentation for the first time (new material) must be a lot like a blind date. You don’t know how long it will actually last (do you have 30 minutes, or 90 minutes of material – or did you get lucky and have exactly one hours worth). You don’t know if you’ll have anything to say. You don’t know what is coming next (I usually talk about the next slides before I get them on screen with new material) and so on… But this one went well. Exactly 60 minutes, the demos all worked and the material flowed.
Then I attended David Kurtz’s talk on instrumentation in PeopleSoft applications. It was a good follow on to my talk which was just about instrumentation of applications in general. It was interesting to see what they had done (and it appears rather comprehensive, it definitely makes figuring out “where the time is being spent” in PeopleSoft much easier.
After that – time for lunch and then off to Julian Dykes inside RAC presentation. He had an excellent slide in there I’ll have to borrow – about how the buffer cache does a full scan. I knew the details, but his picture says it well. I guess sometimes a good graphic can be useful! He also had a nice analogy about how the buffer cache works in cache fusion compared to “kids bugging you for an ice cream”. I like analogies myself – they help me remember how things work conceptually. This one a good one – describing how locks are downgraded cross instance – but only if one of the instances really bugs the other one.
Then I switched my schedule and went to Dominic Delmolino’s talk “Optimization as an Organizational Discipline”. It was excellent – optimization (performance) must be a holistic thing – at all levels. I really like the stories he brings from his job – they drive the points home. The one story I took away is how a disconnect between the business and the technical people can lead to disaster (it ties in nicely with my “WHY WHY is probably the right answer” talk). One week – they were asked for an API that returned a set of customers that had purchased products and those products were set to expire soon (so they know who to email presumably – to give them a change to re-up their service). A little while later – they were asked for an API that dumped information about a product. No problem – easily done. No one asked “why”, just supplied it. Now – a short while later – there is this 8 hour long running process. What was it doing – it was using API one to get the list of people and then for every product calling API two to get the details. Brilliant (not). The simple fix – “just join”, run time massively decreased, new API, everyone happy. This is a good reason why “WHY” is probably the right answer.
Then the last session for the day was Cary’s “Lessons Learned, version 2006.3”. What I took away from this session: Cary was wrong. He described a mistake he made – that was corrected – and the story behind it. The point – we all make mistakes, don’t take things for granted, the obvious answer is sometimes right and sometimes wrong. We all make mistakes.
After that, we had the speaker panel (those are fun) and got some good questions. The day ended with a party – Margaritaville was the theme and fun was had by all.
Now for day two…