Ah ok. I understand why a columnar database would be better for analytical workloads, but I didn't realize that HBase and Cassandra aren't actually laid out in a columnar fashion on disk. I was confused by the 'column store' terminology. This makes Citus data's release even more compelling.

Thanks for clearing that up!