MOOW 2K7 – The Friday Sessions
I stood up at 07 o’clock and after a great meal in a restaurant, I attended my first session: Tapio LahdenmÃ¤ki’s presentation about "Rethink your Indexes".
A lot of programs experience performance problems because of topics like inadequate indexing, table design or optimizer issues. Tapio’s great presentation, showed his method by applying carefully crafted index designs in such a way that if done properly the cause, among others, for random table access would be reduced to a minimum. Starting of with all myths going around regarding index use and explaining semi-fat and fat index principles, he went along with explaining his "Quick Upper Bound Estimate", QUBE in short.
Because of the really nice timeslot of one and half hours, a lot of examples of (index) design could be demonstrated and explained in detail. Afterwards I heard that this session was a short example of what could be expected, when attending Tapio’s (3 day?) masterclass "Better Database Performance with Well-Designed Indexes". For me, well spend time, it triggered re-thinking OIT index problems on Object Relational XMLType storage contra B-Tree indexed driven access paths as currently is the default on Oracle 11g.
After a short coffee break, I went to Anjo Kolk’s presentation "Developers, When Will They Learn".
As you probably know, Anjo is a driven, humorous and gifted speaker, sometimes very straight forward but always knowledgeable when speaking about his the fundamentals he stands for. Given a hardware world were CPU speed performance went sky high in contrary to harddisk I/O rpms / speed, it is a miracle that nowadays there are CPU bound systems. So Anjo questioned, under certain circumstances, the idea that these kinds of environments should / would have a problem related to a topic that a DBA would be responsible for.
To enforce developers to think about what they are doing to a (database) system, why don’t we (DBA’s) show them the door by telling them that it couldn’t be our problem, thus forcing them to re-think statements, procedures or methods they produced? So instead of letting the DBA (not induced by her or him) be responsible for solving every problem out there, but radically sending out the message to the developer: It’s your problem. Re-think what you have done and how to fix it.
After enjoyed a great meal; on to Cary Millsap’s presentation. Since I had seen Cary during Oracle Open World in 2006, were his presentation saved my day regarding all the bad ones I had seen that day, I am a fan (al least regarding his top notch presentation style) despite you can argue if the method "R" is the sole solution to all and everything. Also having been on Hotsos, most of Cary’s presentation was know to me. Despite this I enjoyed most of it including the much heard story about "Bob" (and its implied message: stay on topic before profiling the "kissing" part )
The last Friday session I attended was from the hand of Graham Wood, one of the principle architects behind statspack and currently the architect regarding the AWR, ASH, ADDM and DBTime principle / method implementations in 10g. During his presentation he compared the statspack and ASH (Active Session History) methods. ASH evolved and has great possibilities as probably a lot of you know, but despite the nice trending and graphical interfacing (via eg. OEM), as a lot of tools out there, it comes with a cost. It is "only" looking at active sessions and it is very database centric in its approach.
It doesn’t say anything about its surroundings, the environment where the database instance and its sessions live in. Even so, I realized that the latest OEM products regarding ASH and AWR don’t give you much information about instance trends over time. As I was informed by Graham, this isn’t a requirement anymore because of the automatic memory structures nowadays in place. Now you know why I sometimes hate wizards / automatic stuff; it comes with a price .
Speaking of prices, Graham couldn’t address the issue. As most of the nice features nowadays, they come with an extra Oracle license. ASH, for instance, is part of the license of using the Diagnostic Pack. Asking the audience to raise their hands to see who already used ASH, AWR, ADDM, etc. Only a few people raised their hands. If I would be the principle architect of a nice method like ASH, AWR or ADDM, this result would really hurt. And believe me, there were a lot of people attending the presentation.
After the Grand Gala, with again, a lot of nice food, I tried also to uphold my social 80%, enjoying the swimming pool. Apparently Doug knew what he was talking about, because I still have sore ribs after enthusiastically using the water slides in the Water Park.