Fair point @mattbillenstein.

The point we were trying to make was that we were serving queries with low latencies from disk in a cloud environment, where folks have told us the random read latencies are higher (presumably due to multi-tenancy and virtualization).

We had run this experiment at the request of a customer who had benchmarked a number of systems and shared this concern about higher latencies especially when having to read from disk. They were very happy to see this result (pleasantly surprised even), because performance in cloud environments seems to be an issue for multiple people. So sharing widely.

BTW, we did some performance analysis/benchmarks against Cassandra here: https://blog.yugabyte.com/building-a-strongly-consistent-cas...