We went with HBase. Cassandra would have been suitable too, but we already use Hadoop for data processing so it was a natural choice within the infrastructure ecosystem. We will write a followup about that.

Based on their use case, I'd expect they went with either HBase or Cassandra. I'm quite partial to HBase, it's insanely scalable and has a lot of pretty amazing features, but at the cost of knowing exactly what you want to do with your application beforehand.

I'm not too familiar with Cassandra, but the scalability of an HBase table is almost entirely dependent on your key design. Judging from their use case and requirements, they would likely use a incremental key design which would allow for super fast range scans, of course, this leads to region server hotspotting, which may or not may not be a big deal to them.

Look into Cassandra. Very fast writes and scales linearly (although there is some elbow grease involved in keyspace distribution). Our analytics platform guys are very happy with it, after flirting with several other options (including big, beefy RDBMS). Great for large, flat, denormalized tables.

Just a plain old RDMS?

Me too, except I skipped straight to the end :)