michael stonebraker

i really enjoyed this talk by michael “map reduce is crap” stonebraker. it is a terse and mind-blowing romp through the current database landscape, focusing on inability of one-size-fits-all “dinosaurs” to deal with streaming data (like tick market data) as well as data warehousing. he has some benchmark numbers and proposes some alternatives.

i have been looking into streaming databases and column data stores for a while, but I am yet to have a production hands-on experience with them.

i always had somewhat of an inferiority complex when it came to databases, since these guys have done so much work decades ago that tackled so many problems, and it seems to have had a hard time escaping the database world and affecting CS population at large.

so this new twist in the evolutionary spiral that comes due to the availability of cheap machines, lots of RAM, and the need to process lots of data should yield interesting results. i do hope that this will lead to a whole bunch of lean and mean specialized solutions instead of (loosely quoting stonebraker), “here’s oracle, what’s your problem?” approach.