Sep 23: OOW Day 3
Well, 2.5 too, I suppose. I spent most of Monday afternoon 'networking'. i.e. Sitting around the OTN lounge chatting away to several people I know, most of them the strange old gents who like to complain about things a lot
I did, however, attend Andy Mendelsohn's General Session, or keynote really. It was the first time I'd sat in the dedicated Bloggers area of the big hall, but it all seemed a bit pointless as attendance was very sparse anyway. Bizarre, because I would have thought people would be more interested in the agenda at a technical conference than some of the other main hall presentations, but maybe it's just the crowd being attracted by the word 'Keynote'.
After that, it was an Oak Table dinner in the Grand Cafe, where I was actually able to find something to eat for a change amongst good company. But there was some confusion over the cheese menu - Soft Sheep's versus Sheep's. What exactly is a soft sheep and why did everyone think a Scot was the most likely to know?
Afterwards, Alex and I tied to hit the OTN night, but it was drawing to a close and we only just managed to get a beer. No problem, Frits and Jacco were up for some intense chat and some more drinking. Unfortunately, we spent so much time looking for an appropriate bar and walking around that we might as well have sobered up! Still, I didn't get back to my room to 2am so it seems to be getting later and later!
I saw a few good presentations on Wednesday.
First was the Optimiser Development Group's "Inside Oracle Database 11g Optimizer: Removing the Mystery". They talked about the problems that we've all seen that 11g improvements are trying to solve.
Unexpected plan changes
Wrong Cardinality estimates
Gathering Optimiser Statistics takes too long
Bind peeking doesn't work when there's data skew
and the proposed solutions. The slides will be available on line so I'd recommend you take a look and they highlighted that if you type 'Oracle optimizer' into search.oracle.com, you'll find a few White Papers including one on SQL Plan Management. Here are a few quick thoughts I noted down, though ...
I enjoyed the section on SQL Plan Management, which they described as controlled plan evolution. I liked the term and it made me think back to all the problems I've faced over the years with *un-controlled* plan evolution! LOL. SPM does look pretty good and solves long-standing problems, but it occurs to me that as skills plummet in the industry, people aren't going to really do this stuff unless they're very smart or they use the OEM interface so, without wanting to be too grouchy, it's more fuel to the OEM fireplace.
Oh, and there was a very strong recommendation to use AUTO sample size in 11g to the extent of a challenge to prove it doesn't work well! I can almost hear Jonathan Lewis digging around across the ocean
A few of this year's presentation have had a 'I've heard this before' feel to them but, in fairness, there's a sense that people need to play cards extremely close to their chest these days in case they screw up keynote presentation announcements and tight rules on revealing any share price-sensitive information.
The next one was "Advanced Performance Diagnostics: What the GUI Doesn't Tell You" with Kurt Engeleiter and Jon Waldron. Kurt works for Oracle and used to be on Andrew Holdsworth's Real World Performance group but now works in the Manageability area, which Jon is an architect at the Commonwealth Bank in Australia. During the first half, Kurt described how ASH and AWR work together and talked about the supplied scripts. I suppose I shouldn't have expected to learn much here because I cover the subject in more depth in the course and I didn't, but I did note that he talked about moving AWR data between instances. He didn't mention that this isn't intended for customer use until 11g so I don't know whether that was an oversight or an indication that it's something people should use. What bothered me a bit more was that when he showed examples comparing before and after performance improvements, he always focussed on number of logical reads or various percentages, rather than DB Time. Maybe I'm becoming obsessed, but it diluted the message a bit that he didn't focus on time. It was still a nice overview, though, given that he had less than 30 minutes. Jon's stuff was pretty interesting, too, and he showed how the bank had identified strange performance problems over time by generating graphs against long-term AWR history using SQL Developers graphs. So, I'm not sure the presentation was really that advanced, but they did illustrate the fact that AWR is just a big bunch of data which means you can report on it in many interesting and useful ways, without just sticking to awrrpt.sql
Andrew Holdsworth's annual "Current Trends in Real-World Database Performance" was only 10 months after the last one and it's not completely surprising when an annual review keeps highlighting the same old problems. I can imagine myself saying the same things in 10 years, sometimes! However, as I was discussing with Tim Hall afterwards when we bumped into each other, I like this presentation because it's based on what they've seen at real customers with very large systems and I don't necessarily get to work with them all the time and because I like the heavy dose of opinion in the presentation, too. Andrew feels like a fellow cynical old man, in other words As for the content, well Jason Arneil is going to help me out with that one! LOL. But I noticed he mentioned a paradigm shift in the Storage area and pointed us towards Larry Ellison's upcoming keynote without saying more. Oh, and for any friends who might have attended when I ranted on about the Real World Performance group presentations in the past, don't be worried if it didn't seem very technical, as Andrew pointed out, the real meat will be in tomorrow lunch-time's presentation at 12:00. It was terrific last year.
Next up was a visit to 'The Cave', which is the location of Moans Longball Nogood's annual mobile Miracle office at Openworld. It's nice to pop in there for a few beers and maybe something to eat and I never know who I'll meet as people nip in and out all day. It's probably one of the best innovations of this conference, so cheers to Moans for that. Of course, a few beers later, I realised I should really have a nap or a shower so decided to skip the ASM presentation I wanted to see, just to meet the authors of the book I liked. Back at the hotel, I set my alarm for 30 minutes later. Then slept. Then hit snooze. Then (I don't remember this bit, though) switched the annoying alarm thing off. Marco eventually phoned me from the ACE dinner but, as my phone was still on silent, I didn't know. Until I woke up 10 minutes later, 40 minutes after the party had started Oh, well, a quick dash and I just about made it in time for the starters. It was a nice night, good food and plentiful drink and an annual gift (more on that later). The only problem was that because I'd turned up so late, everyone was seated and there was no chance to mingle. Of course, who was the only empty seat next to? Alex-bloody-Gorbachev! I really have to meet some new people who understand the meaning of the phrase 'can't handle it' properly. (Sorry a private joke moment there.) For example, I just noticed an old friend, Carl Dudley, in Eddie Awad's pictures and I hadn't even been aware he was at the dinner
After dinner a few of us piled into a taxi and headed down to Yet Another Miracle Event, just in time for last orders. I seem to be late for everything this week Eventually I think Jacco, Frits, Marco and I called it a day at about 1 or 2 and I'm starting to wonder whether Jet Lag or Hangovers are worse!
A quiet day for me today. There's a Meet the ACEs thing between 12:00 and 1:30 which might be fun with Mogens, Anjo Kolk and Cary Millsap there (I'm the hungover Scotsman with the OTN T-shirt on if anyone actually does want to say hello) and then there's the big keynote. Oracle have certainly done a great job of building up excitement this year, and for the Database people for a change! So, yes, this time I'll be watching and listening.
As you may have noticed, I've been slacking on the blog postings, so heaven knows when the next one will appear ...