No, I'm looking at Twitter's numbers in production. Before and after very much tells the tale -- and it was mostly the same engineers, so it wasn't "oh well you

No, I'm looking at Twitter's numbers in production. Before and after very much tells the tale -- and it was mostly the same engineers, so it wasn't "oh well you had the B team doing Rails and the A team doing JVM code" or whatever other excuse you want to look for. Microbenchmarks aren't the end-all-be-all, but you see the same story there.