I’ve been a bit busy recently because I’ve had a rolling 4 week cold (like lots of people in the office – obviously several colds in quick succession), a couple of presentations at UKOUG (where I’m sure the additional beer didn’t help recovery) and then I was in Vienna on Monday and Tuesday this week, teaching the 10g Performance course for Oracle Austria.
Personally I was a bit disappointed with my efforts because I could only see areas for improvement but that means that I’ve got the opportunity to do better next time. It’s true I’m naturally self-critical though and the reaction of the attendees seemed pretty good. They were a smart bunch, too, which I can always tell by the standard of questions and I’m pleased that one of those smart guys, Martin Berger, seemed reasonably content. I have a feeling he’s a guy I’ll be seeing presenting at some conference or other in the near future. The food was excellent, the conversation stimulating and Vienna is a smoker’s heaven! (for now …)
Oh, and I forgot to take a photo of the class for a blog post, but my mind really wasn’t on blogging this week …
With the performance basics presentation, the course, and the fact I’ve been looking at a few performance issues at work in between, performance has been on my mind even more than usual lately and I kept spotting related blog posts, too.
First up was Glenn Fawcett’s post sharing his Oracle Performance Analysis 101 presentation that he developed for delivery to his Sun colleagues. It’s pretty good, practical stuff for the right audience and I can imagine I’ll point colleagues to it in future.
There was also a minor flurry of posts around Data Warehouse appliance vendor’s benchmarking claims. Kevin Closson posted an interesting little puzzle that illustrated the folly of placing too much faith (faith being an appropriate word) in those ‘100x faster’ claims. Kurt the Dude guy picked the reason for the stunning performance gain. If I’d had my wits about me and taken time to read it, I would have had a chance too because I experienced a self-inflicted performance problems with some PX testing I was doing a while back when I set PCTFREE to 90 to inflate the space used by a test table. It seemed like a good idea at the time, but it meant a lot of i/o activity, but not much action when the data was aggregated! Actually, what I thought was clever about Kevin’s posts was that he put Exadata head-to-head with Exadata and gained a 485% improvement! 😉 Food for thought, particularly just after I’d repeated some 5-10x improvement Exadata results! Fair point.
Then Greg Rahn added a couple of his own posts on some of the more bizarre vendor claims. Of course it would be easy to say that both Kevin and Greg have a personal stake in the success of Exadata and so maybe they’re just biased? But they’re not marketing guys, that’s for sure, and they haven’t made those go-faster claims without some explanation about where the go-faster comes from.
Finally, someone on the course asked how they could invalidate individual cursors on demand and I said I didn’t think you could. The next morning, Martin came in and said he thought there was a way you could do it, at which point I immediately remember Fairlie Rego’s post that I’d read last year. If there’s one good side-effect of blogging, it’s that it increases the chance I’ll remember these things in future! (No guarantee, though …) So, here’s Martin’s follow-up post.
Rather than performance, I suppose I’d better start thinking about Christmas now! 😉