Re:Well, (Score 1)

What would be very interesting is a study of the bottlenecks. If you are considering a system with certain characteristics (processor/processors, memory, storage/raid etc) and you have X dollars, where is that best spent? By paravirtualizing production machines, we found the greatest perfomance killer are complex bridging setups with virtual NICs on heavy traffic. This affects not only physical cpu load, more relevant for customers is some kind of 'feeled' performance ;) where network latency caused by dropped packets counts as well. Another bottleneck is of a more physical nature. We noticed, that threaded app's on SMP VM's are remarkably slower if the VM uses cores on different physical cpu's. If vCPU's are pinned to pCPU cores, benchmarks are not very different from physical machines with comparable cpu's. Taking the focus only on virtualization, an overhead produced by a hypervisor is hard to measure. Thus spoken, we're using Xen. And yes, this can't be compared against the article about iis ... we don't have any colorful charts ;)