[austinsharp]

My thought exactly. This is just a scale that can be solved either way; when you really can't fit your data on even a handful of machines with acceptable performance, then Cassandra can start to shine.