A potential issue with benchmarks like this is that there are instances where the initial findings don't scale. I would be interested to see how it does over a

A potential issue with benchmarks like this is that there are instances where the initial findings don't scale. I would be interested to see how it does over an operation that takes 1 minute, 5 minutes, 10 minutes.