I'm surprised that neither HBase nor Hypertable are mentioned. Both HBase and Hypertable have much larger community and deployment than many of the mentioned alternatives.
I would not pick either HBase or Hypertable as an option for most of the uses people are switching to non-relational DBs/KV stores for (which I think is driven more by simplicity than needing multi-TB scalability.) They're not particularly accessible to just running make and getting started with either (I have a very large hadoop cluster at arm's reach and I found them sort of painful for other reasons...)
Based on that I'm kind of surprised they put Cassandra in there, though; I would class that with Hypertable/HBase etc in one article and CouchDB/memcachedb/tokyo cabinet in another.
Hypertable running in single node/local mode requires zero configuration and runs orders of magnitude faster than these alternatives when RAM to DB ratio is small.
MemcachedDB is a memcached-compatible database backed by Berkley DB for persistent storage.