I think you may be over generalizing here. Yes there are vast similarities between MongoDB and other NoSQL databases but many of the issues he pointed out here are specific to MongoDB (and certain versions of it at that).

I think the real lesson here is to choose technologies based on actual experience rather than hype.

So that means extensive testing before nailing your colours to the mast.

Or choose technology based on requirements.

If the port to SQL was so easy and used half the storage and didn't appear to impact performance - it was possibly the better solution in the first place.

Mongodb is awesome - one of best technologies I've played with in a long while - but don't be too quick to write SQL off - especially if you have a lot of relationships.

Off topic - prototyping in Rails with Mongoid and the schemaless Mongodb underneath is frictionless - it's a step change in speed from prototyping with SQL.

Would you happen to have any anecdotes of a project or dataset that MongoDB worked particularly well for you?

I'm working on a project where documents seem to work particularly well, though the core of the project is still in MySQL. I decided to give Mongo a shot for what I'm working on, but posts like this are a reminder to constantly be evaluating the technology you choose to use.

Take a look at this: http://www.percona.com/software/percona-server/for-nosql/

There is one for mysql: https://github.com/ahiguti/HandlerSocket-Plugin-for-MySQL

But i would urge to go with XtraDB.

I use mongo for everything but the money (where cash is not involved) and for things not needing transactions. Have been sticking to MySQL for everything else (now moving to XtraDB).

For big big big projects and pure nosql I go with Cassandra (as long as you can feed it enough RAM). Otherwise, XtraDB can go the distance.