Well, you should benchmark on similar setups to what you will deploy on: benchmarking on "real servers" or "bare metal instances" is also going to be highly mis

Well, you should benchmark on similar setups to what you will deploy on: benchmarking on "real servers" or "bare metal instances" is also going to be highly misinformative if you expect most of your deployments to be on virtualized hardware under KVM or Xen.