Noticed that they have ‘overflow’ rooms where they are broadcasting sessions that max out attendance, very nice. Wonder how much they get used? Lunch was BBQ, chopped brisket (not pulled I think – more Alpo’ish), pork. Not as good as yesterday, but not bad.
I requested an interview with Torsten Grabs to learn more about StreamInsight – the process being to walk over to the staff table in the press room and ask! That falls under the BI Conference so they forwarded the request to them, a bit later I got a reply that he wasn’t available and offering a substitute, Herain Oberoi, a Group Product Manager on the SQL team. Set that up for 12:30. They’ve got a room set up for interviews that has portable rooms within it, think a cube 8 feet tall and maybe 10×10, round table with four chairs, conference phone. Functional, not fancy.
Catch up on a few other things, head next door for the interview. You can see the results from my post yesterday. I was all set to dig into technical trivia, but really needed Torsten. I suspect that the PR people are used to business focused interviews and are just trying to get as many done as possible, missing the distinction that I am – more or less on any given day – a technical guy.
More wandering and chatting after writing up those notes, went into a performance session for a while. They had replaced a bunch (4, 5?) racks of SAN storage with a single rack of SSD. MS of course, nice to have the budget to give that a try. Lots of interesting tidbits, one that caught my attention was a recommendation for max performance (think TPC benchmarking) was to create one file in temdb per logical processor. Not core mind you. At the same type they said with the Nehalem processor hyperthreading was a better play. Still possible to have workloads where it could hurt you, but in general due to waits you could get higher throughput by enabling HT.
From there off to an Open Data on the Web by Douglas Purdy and Jonathan Carter. Easily the best session I attended. Fun guys without being over the top, great to the point demos, and clearly passionate about OData. Essentially oData is a way to query data that you expose on the web – think of the power of RSS feed, only with structured data. One of the demos was a static html page with about six lines of code, using Jquery to bring back a list of movies from Netflix. No asp.net, very elegant. Moving to asp.net, also easy to implement and there you write the queries in LINQ, and it transforms the LINQ query to the Odata syntax automatically. Very compelling, both on the ease of implement, but also on the idea of opening up data. On my todo list to try soon.
Dropped on in the Database Design session by Buck Woody. Good session, and very fair (and Zen) answers that appropriate normalization is a call based on situation and experience. Buck really doesn’t like the diagramming tool in SSS, for me it’s there and with clients it’s just the most direct way to show them stuff. No doubt MS has totally neglected this part. At the end of the session I had an attendee ask me about guids for primary keys, and of course I’m a fan of that, but always try to explain the tradeoffs. Brent Ozar was aghast that I’d be willing to sacrifice some performance (he doesn’t like replication either, making him my polar opposite in many things!), but for me performance is just one part of the puzzle and I don’t obsess over it – I can get the performance if I have a good design, guid or integer keys.
From there we were going to dinner, me, Brent, Denny, Buck, only we got on the wrong bus. So we had to ride the whole loop back to the convention center, get on another bus, and then to Brent’s hotel to drop off his stuff, and then finally to the restaurant. By the time we were seated it was 7:30, and I decided to skip a long dinner to get to the Tuesday night party that ran 7-9 pm, having promised to meet people there. Party was at the aquarium, got to meet some new people, had a fun conversation on transparency, saw my friend Andy Novick, and finally called it a night about 10:30.