Conor Cunningham’s blog on SQL Server, data-driven applications, and pretty much whatever other random stuff he decides to post.
I found this paper interesting, so you might too. It covers some of the differences between parallel database systems and map-reduce style computation:
I think that Mike Stonebraker's remarks should be taken with a grain of salt since he has a vested interest in Vertica and debunking his competition. Looking at that list of authors, I can't help but notice that none of them would provide any kind of unbiased analysis of the situation.
Some of the problems brought up (sophisticated schedulers) have since been solved. Others (abstracting the language or ad hoc querying) were solved at the time the article was written (see Apache Hive and HBase respectively). Massive RDBMS parallelism can even be solved for free, without purchasing Vertica, through the use of HadoopDB.
Stonebraker frequently has good points, but much of the value is lost because he's constantly flogging Vertica out of the side of his mouth.