Good morning! It's Day 2 of the PASS Summit 2013 and it should be a busy one.
Douglas McDowell, EVP Finance of PASS opened up the keynote to welcome people and talked about the financial status of the organization. Last year's Business Analytics Conference left the organization $100,000 ahead, and he went on to show the overall financial health, which is very good at this point. Bill Graziano came out to thank Doug, Rob Farley and Rushabh Mehta for their service on the board, as they step down from their positions.
Tom LaRock introduced the new executive board, including Adam Jorgenson as the Executive Vice President, and Denise McInerney as the VP of Marketing, and he introduced the new incoming board members, Jen Stirrup, Tim Ford and Amy Lewis.
The PASS Business Analytics Conference will be in early May in San Jose, California, and next year's PASS Summit will be in Seattle from November 4-7. Tom invited everyone to the WIT luncheon here in the Cisco Crown Ballroom, to the Birds of a Feather lunch tomorrow, and to the Community Appreciation Party tonight at the NASCAR Hall of Fame.
Today's keynote speaker is David DeWitt, Technical Fellow at the Microsoft Jim Gray Systems Lab, to talk about Hekaton, What, Why and How.
Dr. DeWitt seems to think we'll be board with his talk and he couldn't be further from the truth. He always explains really complicated things in a way that the rest of us can really understand.
He calls Hekaton an "OLTP Rocket Ship". It's memory-optimized, but durable, and fully integrated into SQL Server 2014. It's architected for modern CPUs.
OLTP performance has benefitted from CPU performance improvements, etc., but the hardware improvements have pretty much maxed out. Hekaton essentially "means" they're going for 100X performance improvement. How do we get there?
Pinning tables in memory still has problems. Performance still limited by latches and locks, and interpretation of query plans. Latches must be used to protect data in the buffer pool, but cause contention for other processes attempting to read the same data.
Hekaton uses lock-free data structures, it uses versions with timestamps and optimistic concurrency control, and it's compiled into a DLL to improve performance dramatically. SQL Server now has three query engines under the hood: Apollo (the Column Store index processor), the relational query processor, and Hekaton. Essentially uses versioned views of data using the optimistic model to provide high-speed throughput. Dr. DeWitt's discussion was detailed and thorough, and it would greatly benefit you to view the recording if you didn't see it live.
Allen White is a consultant and mentor for Upsearch Technology Services in Northeast Ohio. He has worked as a Database Administrator, Architect and Developer for over 30 years, supporting both the Sybase and Microsoft SQL Server platforms over that period.