On: How not to benchmark Cassandra: a case study

[jbellis]

This was the toughest part to make clear as well as brief. I was not entirely successful.

For Cassandra I used the default 50ms batch window. I'm not aware of tunable windows for the others.