Last week, I had a wonderful time participating in the 2010 Hotsos Symposium; a favorite annual outing to network and learn in the company of many of the world’s very best Oracle performance experts. It was my first public outing under my new trade name, EPiC Performance Associates.
It was encouraging for me that Tom Kyte’s keynote address quoted my definition of Best Practices from my 2001 Sun Online Blueprint “Sun/Oracle Best Practices”, which is a definition I still stand by a decade later. I sometimes define Best Practices as being “the things you have to do yourself until the software evolves to do them automatically”, but I’m adamant against the term being diluted to mean “some clever things we do or did that we believe are part of our success”. I also liked Tom’s candor about how even the best feel “stupid” when encountered with a problem they’ve never seen before. I liken that to my own metaphor about fear of bears and the value that can bring to one’s bear-hunting outcomes.
Although I missed making a timely submission during the original call for papers, the good folks at Hotsos latterly signed me up as a standby presenter. Since Henry Poras was unable to present his scheduled session on “Diminishing Resource Utilization and Saturation Limits Using AWR History and Queuing Theory” on Tuesday afternoon, I got to present my “CPU QoS (Quality of Service)” talk in that time slot. My session was very well-attended, and I received nothing but positive feedback regarding the material. I am passionate about the qualitative aspects of CPU usage, and I hope I motivated some attendees to go exploring in those dimensions. Ironically, my session in Hall B ran concurrent with “Performance and Sizing of Oracle on VMware” in Hall A, which was presented by my esteemed former colleague at Sun, Richard McDougall, now Chief Performance Architect at VMware. If it had been a Sun-centric conference rather than an Oracle-centric conference, Richard would have drawn an overwhelming share of the attendees!
Hotsos organizers chose Disco as the theme for Tuesday night’s dinner party, and I saw no choice but to attend in costume as The Disco Duck! Yep; I was in Atlanta when Disco was big, and I felt I just had to make a statement! Younger persons and attendees from overseas mostly thought I was a chicken, but the DJ set everyone straight. He made an on-the-spot iTunes purchase of the original 1976 Rick Dees’ Disco Duck track on his iPhone – which led to my command performance on the dance floor. A thousand thanks to the kind lady who joined The Duck on the dance floor; I believe she would prefer to remain anonymous!
On Wednesday, I had the pleasure of co-presenting with Tanel Põder on “Understanding LGWR, log file sync Waits and Commit Performance”, with my slides filling in the part about what happens to I/O requests once they leave LGWR to run the gamut of the I/O stack to be serviced by the target hardware.
Tanel Põder’s training day on Thursday was incredibly well-attended, and I’d say he proved his point that (paraphrasing) “The Oracle performance tool of choice for the 21st century is … SQL*Plus!”. The speed at which Tanel navigates Oracle is impressive, and he does a great job of empowering others to do likewise!
Marco Gralike captured numerous video clips from the conference, including my brief tour on the dance floor as Disco Duck on Tuesday night. Google “Hotsos 2010″ to find these and catch a glimpse of some of what went down in Dallas last week! The Hotsos Symposium has become my favorite annual outing! I’m already planning for Hotsos 2011, where I hope to present a session tentatively titled “Brute Force Parallelism”, the story of a rare 100X speedup payoff we got by adding a little C code with some PThreads logic to a monthly procedural batch-processing job that originally took far more than a month to run. Yep; I need a good excuse to write that one up properly and share it — and try to get on next year’s agenda well-before the deadline!