UKOUG Day 3

After more work on the presentation and some very strange sleeping hours thanks to the raucuous ranks on Broad Street, it was time for the first of my presentations. But before that I wanted to attend JB's "Proactive Detection of Oracle Performance Events using Adaptive Thresholds" (skipped), the always entertaining Mogens Norgaard's "The Danish Experiment: 11g-Shock Upgrades and Massive Workload Reduction via C.O.B.S. (TM)" which was good if a little short (more on APC's blog) and Hüsnü Şensoy's "Backup and Recovery Tuning for VLDBs", which was cancelled. I was particularly disappointed about the two missing presentations because JB's stuff was probably relevant to my course and I'm currently involved in designing a Backup and Recovery strategy for a big new DW that's planned at my current site. On the other hand, it gave me a little more time to get ready for presentation and chill out a little.

Marco Gralike chaired the session and was good enough to bring a gift of this little chap.



Apparently he was keeping him awake all night by watching MTV in Marco's hotel room. I didn't really understand this story properly until later on Wednesday night when the little guy spent all night watching MTV in *my* room

I was unusually pleased with my "Performance Tuning Basics" presentation. It was definitely a push squeezing it into 45 minutes as best I could and there were definitely a couple of things that didn't go as I'd hoped but, fundamentally, it was one of those good days. It was an introductory presentation and so it was difficult to go into anything in too much depth, but I think that sometimes it's the basic approach that people get wrong, rather than the technical stuff and that was the main message I wanted to get across. One problem was that I had a good couple of demos which would have helped reinforced some of the points, but there just wasn't enough time. Another was that I'd planned to promote the DBA 2.0 session after mine because I knew those guys would show some of the Performance Pages in DB/Grid Control that I didn't have time to show. Although I'd checked this before the presentation, I only remembered when they came on stage at the end to set up. Graham Wood and JB were very nice about my presentation, though, which was comforting as they're two of the guys who design this stuff! In fact, later that evening, people kept coming up to me to tell me that they had both been very positive about my session during their own, which I was really pleased to hear. Maybe I should have attended DBA 2.0? Actually, hearing Marco's verdict afterwards, maybe not! I'll leave it to Marco to put his money where his mouth is (Updated later ... Marco has posted his thoughts on the presentation now.)

There was one other small correction that Niall Litchfield pointed out to me post-presentation. I think he's right, but I'm going to check it out at home before I upload the slides. (Updated later ... Niall was right. As described in 'Tales of the Oak Table', it was as a result of a problem with a benchmark, rather than a customer problem, that the wait instrumentation was introduced but I think I had the correct time-frame at least. I've replaced 'When faced with a customer problem ...' with 'When faced with a benchmark problem and uploaded the slides to the UKOUG website. Frankly they're very visual and content-light, though.)

A quick dash and I was off to listen to Steve Shaw of Intel talking about "Benchmarking Parallel Query Performance". I have a clear interest in the subject, both from a load-testing and PX perspective, so it was pretty interesting for me. I liked the closing line - 'Don't guess, test' - if I recall correctly, but my favourite bit was probably that he was another guy (like Eric Grancher yesterday) who had been able to test on Exadata. His findings on a TPC-H-like test (and it's very important to state that it's only a representation) using HammerOra was that Exadata produced a 5-10 times performance improvement over his Fibre-Channel SAN configuration and that it had changed the game completely for Oracle Data Warehouses. Now, whilst I hestitate to make such a bold, unqualified statement in a blog post, he was clearly very impressed, having actually worked with the thing. Maybe I'll need to go out busking (for a *long* time) to get the pennies together to buy one for my current client.

After that it was the main UKOUG party night but I'd have to say at the risk of sounding repetitive that the whole event seemed much quieter this year so it was a bit of a damp squib from my perspective so I didn't stay out too long because I had another presentation to prepare for in the morning. They announced the newly-appointed Directors whilst I was having a smoke outside but I heard that Lisa Dobson was voted in when I got back upstairs. Congratulations, Lisa

P.S. I had extremely sore calf muscles when I woke up on Wednesday morning. It took me about 30 seconds puzzling why that might be before I remembered running up and down stairs during James Morle's presentation!