CIDR 2009 is shaping up to be as interesting as I had hoped.
If you don’t know about CIDR, it’s a conference founded by David DeWitt, Jim Gray and Mike Stonebraker 8 years back, for data systems work and other data-centric topics that would have a tough time getting accepted to the more hidebound conferences like SIGMOD and VLDB. It’s usually a good conference — everybody stays in one room and pays attention, and it tends to attract a smart mix of old hands and young upstarts who really dig in and engage. I ran the program this year, and was really happy with the papers and speakers we got. Only have time to comment on a few things here.
Jeff Heer kicked off with a keynote on Social Data Analysis, based on his work at Berkeley, IBM and elsewhere. I’m a huge fan. If you’ve never used Swivel or ManyEyes or the like, you should check them out. Then go read Jeff’s work, which does the best job to date of setting a research agenda around this very provocative set of ideas for harvesting the social effort around data analysis, turning that work into reusable information that can improve the analysis process both for users and for the software trying to analyze the data.
It’s also good to see lots of energy around novel uses of declarative languages — both for new apps, and in an “eat-your-own-dogfood” context, applied to DB infrastructure.
- The Cornell Data-Driven Games group pitched their language for batch specification of state update in computer games. It’s a compelling story: they point out that when a game has lots of computer-controlled characters, the game AI has large-scale batch updates between time ticks. So that logic is more simply programmed and optimized/parallelized via a data-centric language. Their language is declarative (compiles to relational algebra) but has an imperative “skin” to make it look more familiar and Java-like. In the CIDR spirit, it’s early days for their project — they don’t have any games implemented yet, and are still focused on simple select/project/join/aggregate queries. The argument for their specific language (over SQL, say) is currently just programmer familiarity — they assert that game designers are allergic to SQL. I buy the argument that syntax matters. But I’m looking forward to hearing about the features they evolve for the games domain that force them to go beyond vanilla relational algebra.
- Folks in Boon Thau Loo‘s group at Penn, joint with folks from LogicBlox, presented their work on LBTrust — a declarative framework for specifying trust languages. The application to neatly generalizing trust languages is interesting, but the ideas they use to get there are even moreso. As in our Evita Raced work, they represent code as data; they call this a metamodel. They leverage that in new and very clever ways, including doing code synthesis in (meta)logic. They also handle types and other constraints in an elegant way. There is some convenience syntax for all this that’s of interest as well. It’s a very powerful use of the code-as-data framework.
- Folks in Sam Madden‘s group at MIT talked about their work on RodentStore, a declarative framework for designing and optimizing storage system layout. This is great dogfood-eating (to mix animal metaphors): a database system whose storage layer is implemented in a declarative language that can be co-optimized with queries. Like most good CIDR talks, there’s a lot of proof yet to be put in the (dogfood) pudding, but the direction looks very interesting.
Data in Beta 
2 Comments
Thanks for the post. This was my first time at CIDR. I’ve been to security conferences before, and the difference between those and CIDR was clear. The emphasis on one-talk-at a time was very useful, and we had more interaction than most other conferences. I think the location also has something to do with it … :)
Yep, CIDR is special. Much less diffuse than the big DB conferences like SIGMOD and VLDB. Single-track conferences are far far better than smaller ones, especially when the room is filled with smart folks.