Oct 8: OOW 2014: Day 2
I picked up a few extremely useful things from this presentation but I think the most important one was the journaling area used when rows in the standard row-orientated buffer cache have been updated. Which, for starters, means that only 80% of the allocated memory will be available for your original data. Not a problem, but worth knowing.
What really jumped out at me though was when he discussed how the number of updated rows could affect the optimiser's decision to use In-Memory or not. I might not have explained that very well, but I believe the effect would be that the optimiser is likely to flip between using In-Memory or not depending on quite a few variables. Which means one thing to me. Potential Execution Plan instability. I'm not sure how Oracle could get around this because cost-based decisions are the sensible approach but I foresee lots of new performance analysis and tuning opportunities! Not quite "flick a switch and it just works", but who would ever believe that kind of thing anyway?
Great presentation, though. Exactly what Oak Table World is all about so thanks to Kyle Hailey and the various sponsors () and speakers for making it happen!
Judge for yourself. No need to go to San Francisco!
... or how about finding open_cursors set to 2000? A per-session limit of 2000 cursors? As Graham pointed out - good luck keeping track of the state of all of those! As soon as you stop and think about these things sensibly, you realise that it's almost certainly a sign of an application leaking cursors.
There were lots of similar examples but the interesting overall approach that I would say they were illustrating is something that I tend to do when I first arrive at a new client site and I've watched other experienced Oracle techies do the same.
An AWR report is not just the top 5 timed events and the sections at the top are a pretty good description of the actual system workload which, in turn, can tell you a lot about the application design. Then, based on potential application design issues, you can drill down into the report and look at later sections to see where all those leaked cursors or transaction rollbacks or (whatever) ... are coming from.
Updated Later: As Toon Koppelaars highlighted on Twitter later, you can see this version of exactly what I'm talking about here, for free. I should hang my head in shame because Andrew and Graham made a point of all the RWP videos being available online here. Watch and enjoy!
Lucky boy that I am, I was able to retire to the comforting surroundings of the Thirsty Bear to continue the conversation about all things performance related with Graham and JB, much of the conversation being me whining about why people don't use the *full* range of tools that come with the Diagnostics and Tuning Packs that they've paid Oracle good money for. That's why I've been slowly developing a presentation on that very subject.
Then it was back to Oak Table World to catch Greg Rahn talking about all that Hadoop stuff *again*! Even though I only caught part of the presentation, I do keep managing to pick up bits and pieces on the subject although I wonder when it'll become relevant to my day to day work. Probably whenever I'm too late to the party, as usual
But my main reason for showing up was to see Kevin Closson talking about using SLOB in some less obvious ways. Because SLOB is a good all-round Oracle workload generator, it shouldn't be seen as simply a tool for testing storage performance and that's probably it's main strength. Kevin is always a great speaker and I find listening to him a very different experience to reading his blog, but I'm not sure I can put my finger on why. Oh, he also had the most ridiculously bright SLOB buttons! (As I found out by making the mistake of looking to closely at it as I tried to switch it on )
At some point, all of the slides for the Oak Table World presentations should be available on the site, so keep a look out for those! (Oh, and I got my T-shirt which is deeply cool and was one of the few items of non-ACE swag I managed to pick up all week)
From there on, it was more or less party all the way.
- First quiet beers and snacks with lots of Oak Table and Oracle types.
- Then my very first ever Customer event that wasn't for a specific technology area, but a sales region. Man, *that* was a mistake! Suits *everywhere*! but I suppose it was useful to build contacts with the senior support managers in my new region.
- Instead, I headed towards the OTN night in Howard Street (until I realised I'd just dropped my bag with the entry ticket back at my hotel room)
- So instead I landed at one of the events of this and any other OOW - The Friends of Pythian Party'. As always, beautifully-organised, very generous on the liquid refreshments and the coolest crowd in town. Just because I find myself thanking Vanessa Simmons, Paul Vallee and all of the Pythian crew every year doesn't make it any less sincere.
I have to be honest, though, and say that the highlight of the night for me was spending much more time with Kevin's punchy, beautiful and fun wife Lori. If you think Kevin's smart, wait until you meet his wife! There's a lady who can hold her own and make me chuckle Problem is that I think she's used to scaring people but us Scots don't scare so easily
It was a great night anyway, as always, and although this is entirely unconnected to the Pythian party but might have had a *lot* to do with jet lag, I didn't wake up until 11:45 the next morning