It appears that recently there is growing interest in nested virtualization. A few days ago, VMWare released a technical report, titled "Bounding the Running Time of Interrupt and Exception Forwarding in Recursive Virtualization for the x86 Architecture", available at http://www.vmware.com.mx/files/pdf/partners/academic/hh.pdf This report discusses why nested virtualization is important, and estimates the performance of multiple levels of nesting. They show that the interrupt handling time is exponential in the number of levels, but propose hardware modifications which could make it linear. Their results are based on simulations, not on an actual implementation of nested virtualization in VMWare: They say that "current hardware is not performant enough for real implementation", and that "The design described in this paper is not in any VMware shipped product". As I wrote in a previous email, earlier this month we published our paper on nested VMX in KVM, "The Turtles Project: Design and Implementation of Nested Virtualization". The paper was not only accepted to the Usenix Symposium on Operating Systems Design and Implementation (OSDI), it was even chosen "Jay Lepreau Best Paper". Among other things, this shows that the program committee and reviewers found nested virtualization to be an important research topic. A few months ago, Olivier Berghmans from Antwerp University published an MSc thesis titled "Nesting Virtual Machines in Virtualization Test Framework" in which he, again, explained why nested virtualization is important, and how it can be done without new hardware or software (e.g., one option is to run a software-only hypervisor on a hardware-assited hypervisor, but he surveys many other options), and compared many different options. ScaleMP announced in May "VM on VM" (see http://www.scalemp.com/VMonVM), a feature which allows to run a hypervisor on top of their underlying hypervisor. I haven't seen any details published on what exactly they did, or how. In April, Nested SVM and Nested VMX patches (for hardware-assisted virtualization) were also sent to Xen. I am not sure what is the status of these patches. I put copies of all above mentioned documents (in case there's difficulty in finding them), in http://www.math.technion.ac.il/~nyh/nested/ -- Nadav Har'El nyh@xxxxxxxxxx +972-4-829-6326 -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
