I watched Larry Ellison’s live keynote at OOW 2011 last night. It’s all about Performance (Nice touch on the video about the America’s cup “Hardware and Software engineered to run together”) and “everything parallel”. A bit like the team that won the the Cup, it takes everyone to be coordinated and working together to achieve this kind of performance. I have to say, Exadata is really “Crushing it!”. Scanning of 5TB compressed in 5 seconds is quite impressive.
I’ still a bit puzzled by why Oracle hasn’t yet developed a way to spawn parallel processes in their PLSQL language, a bit like Java does with the Thread class. I’m not talking about the “Parallel Query Option” in SQL, but a true parallel feature, with thread synchronization APIs, etc. LCS has already developed it’s own version with a simple framework that we use to run conversions and batch jobs up to 10x faster. Our customers have enjoyed a mini software version of Exadata where a single job runs with concurrent threads to process the data. We’ve been able to demonstrate performance ratios of 10x, 12x and even 16x based on the underlying hardware.
Can’t wait to run this framework on Exadata soon. Are we looking at 10×10 performance increase? I’m sure that PLSQL will have some kind of parallel processing at some point. In the meantime, let us know if you want to boost your current jobs.
What you doing to increase the performance of your long running jobs?