That latency tip is so naive.... it is assuming that performance profiles are randomly distributed. In the real world, this simply isn't the case.
The end of the article has an edit saying how time-correlated events would throw off the math, but it also would be thrown off by the real-world case that all CLIENTs are not created equal, either. If there is a slow connection between a client and a server, their requests will be served slower.
The end of the article has an edit saying how time-correlated events would throw off the math, but it also would be thrown off by the real-world case that all CLIENTs are not created equal, either. If there is a slow connection between a client and a server, their requests will be served slower.