Pushing calculations to the database is a great idea, good to see it gaining more traction. Too often people pull back "all" the data just to iterate it locally.
I looked at MonetDB as part of researching column databases in general. It was as fast as the commercial databases on some queries, I was really impressed. The main thing that put me off was that you couldn't seem to nudge data layout / the query optimizer to do exactly what you specify. They seemed to have the idea of an all-knowing system that would optimize for you, perhaps as a research idea that works but for commercial use it worried me too much.
The list seems out of date, InfiniDB has been open-sourced since Calpont went bankrupt, and Greenplum is open-sourced under Apache 2 license[1]. There's also MariaDB Columnstore, a fork of InfiniDB[2].
I've always liked the idea of moving general compute to the database, but usually in an organisation the databases are pretty tightly locked away and it becomes a pain to get the right versions of your libraries installed.
It's nice if you're looking for an alternative to a paid columnar database (ie Vertica, Redshift). I used it on a side project and it performed surprisingly well for large queries (think sum/group by across tens of millions of rows).
I looked at MonetDB as part of researching column databases in general. It was as fast as the commercial databases on some queries, I was really impressed. The main thing that put me off was that you couldn't seem to nudge data layout / the query optimizer to do exactly what you specify. They seemed to have the idea of an all-knowing system that would optimize for you, perhaps as a research idea that works but for commercial use it worried me too much.
For anyone interested in column databases in general I put this list of comparisons together: http://www.timestored.com/time-series-data/column-oriented-d...