I thought I'd better leave this until 3rd April or it would have seemed too much like an April Fool's joke. That's the first version of the course slides finished now.
It's a relief given the amount of work involved and Monday's impending deadline, but note that I used the words 'first' and 'version'. I'm already unsure about the slides but I still have a week or so to do more work on the demos, which I'd already been developing on and off over the previous couple of months, so hopefully they'll add some value to the course. The problem is that calling anything a 'Master-class' fills me with dread. If people expect some deep-dive into optimiser internals a la Jonathan Lewis or three hour bind variable presentations in the style of Tom Kyte then they're going to be disappointed!
I found myself covering more 'traditional' tuning features than I expected, but it makes sense in retrospect. All of the new features are really the natural development of what's gone before and my experience is that people often get the fundamentals wrong and getting the best use out of any performance analysis tool or method is best achieved by applying a sensible approach. Better to understand what performance analysis is all about before getting bogged down in data and pretty screens! (But I must say, they are *very* pretty!)
I also found myself wanting to cover more 11g features than I could squeeze in, but I can see the course mutating into a 10g/11g course over the coming months.
Regardless of my feelings about the technical balance of the material, I must say I'm really excited about teaching again. It's one of my favourite activities. I've been able to keep it up a little through events at work, but two full days teaching a bunch of people I haven't met before is always inspiring and I'm sure to learn something new too, based on the questions that will be asked.
My current site have been good enough to let me teach a dry run of the course next week which will help me nail down the timings and iron out any bugs before it hits paying customers (well, they are paying really, but not as much!), then it's off to Prague a week on Sunday. I would have liked to have spent more time there, as Tim Hall suggested, but I'm afraid time is money and I already take enough days off going to conferences and the like. Prague is going nowhere soon and I'd rather go back with Mads, rather than through work, and enjoy the place properly.
I've hit a couple of trivial issues whilst putting the demos together so I'll probably blog about them over at the other place. I expect things will pick up a bit over there. I've had some ideas rattling around my head recently but I delayed them until the course was done.
It still sounds like April's Fool joke.
But seriously, well done! Hey, it's interesting that you are going to teach in Prague these days... there are so many things about Prague going on.
IMHO, there is too much material around in too much detail and very few people doing an educated, solid round-up of performance-related work.
Not every problem needs a full 10057 trace or delving into the intimacy of obscure X$ tables and memory dumps, or screenfuls of OEMgrid-like confusion.
In fact, from what I've seen in quite a few years now of production support work, most problems have quite trivial solutions.
By contrast, very little work is spent making sure they don't happen again or that the conditions causing the occurrence are not repeated. Or even that the conditions causing the problem have been investigated at all!
If I get the drift of this course you're putting together, it reads like an emminently sensible approach to addressing these subjects.
Last week wasn't good - I looked through the slides as delivered and thought "where's the technical detail" but then as I've been developing the examples a little more, there's going to be enough in there. I have to remind myself that a course has to serve a purpose beyond impressing people (primarily myself
Besides, there are only two days and I insisted to myself that I had to touch on how we got where we are today - not everyone deals with 10g exclusively - and how to approach performance sensibly.
Your comment made me chuckle to myself and think back to one non-technical slide very early in the course.
"Performance Tuning is both
Complex
Simple
Your two most important tools
Information
Common Sense
Performance Tuning is informed common sense"
Although, as I noticed Richard Foote pointed out on his blog recently, the problem is it's not that common
Oh, and step one of my method for solving performance problems is 'Avoid them' although that's easier said than done with some of the Third Party apps that land on our systems
Well, I'll be teaching the course next week and the one after, to two very different groups, so we'll see how it goes.
Thanks again for the reassurance.
Slides look good so far, but having a "deep-dive into optimiser internals" wouldn't be bad at all
Really?
Only kidding