The author mentions the "trick" of closing a poorly performing instance and re instantiating it again in the hope of locating away from bad neighbor. This soun

The author mentions the "trick" of closing a poorly performing instance and re instantiating it again in the hope of locating away from bad neighbor. This sounds reasonable, but the method could be improved by instantiating a new instance first and then removing the old one - this ensures you don't instantiate in the same location as before. I don't know EC2 architecture well enough to know, but there may even be some ways of telling if your new instance is located in the same problematic instance as the original one (perhaps by tracerouting the problem instance and finding it to be v near by). If this happens presumably you can instantiate a 3rd and repeat until you find a suitable instance at which point you kill the others.