We use TokyoTyrant to store lots (TBs) of tiny (<100bytes), constantly updating bits of data. The record overhead on Mysql and Postgresql was huge, and the continual updates were leaving lots of dead tuples that had to be cleaned, and killed performance.
I use MongoDB for greenfield app development and experiments, so I can quickly alter "tables" and "attributes", without having to bother with first thinking about schema, or having to write migrations.
Had the same experience with Tokyo Cabinet/Tyrant on a project. We had hundreds of millions of small records to store, and TCH could pack it into space fall enough to fit into disk cache on an XL EC2 instance.
I use MongoDB for greenfield app development and experiments, so I can quickly alter "tables" and "attributes", without having to bother with first thinking about schema, or having to write migrations.