“AI is everywhere” – that is what I took from the opening keynote at the IDUG Db2 Technical Conference in Charlotte, NC.
Though ML methods start to empower performance optimization and are about to allow for self diagnosis and tuning, many of my colleagues at the conference agreed with me in thinking that serious DBA work is still the key to maximize performance and availability.
Attending the technical sessions, I collected 27 findings to improve performance and availability of our Db2 installation. These take-aways are going to be analysed, then transformed into stories and will eventually be implemented. Some of them are simple checks of system parameters, but others also include updating some of our homegrown health checks with new Db2 12 features and structures or allow us to use profiles to solve problems and to tailor system parameters to different workload groups rather than simply following the “one size fits all” paradigm. So, the outcome of the conference will surely keep us busy during the next weeks.
In addition to attending many technical sessions, visiting the booths at the exposition and previewing or discussing the newest vendor products and features also contributed to the conference experience: Following a hacker attack to the mainframe and interpreting the alerts and actions produced from the attacker’s traces in real time – very interesting!
And obviously the social events helped to maintain and extend my network of Db2 specialists worldwide.
Special thanks to IBM to invite all conference attendees to the NASCAR hall of fame: Simulating Db2 buffer pools is ok, but simulating a NASCAR race in an original muscle car is even more fun!
For me personally it was a pleasure to speak at the conference about our performance results with the new Db2 12 version.
We managed to save 20% CPU during the first four months after the migration. The key success factor was to identify objects which match the sweet spots for contiguous buffer pools and insert algorithm 2. Can you imagine what helped us to identify these objects? Right, you guessed it: we use ML methods to help us with our daily DBA activities!