The Design And Realization Of Hardware Virtual Machine

Virtualization Technology can be used to divide or combine the computer resource (CPU, Memory and Disk). It represents the old resources as one or more new OS running environments. It helps to optimize the resources. In one word, Virtualization Technology can help to rearrange the computer resource, and then we can take full advantages of them.As we know, emulator is one of the solutions in Virtualization fields. It can create an abstract and pseudo hardware, and then the guest OS can run in this environment; it can support almost all the guest OSes. This is a pure software solution, all the instructions of guest OSes are translated dynamically by the Hypervisor, so the performance is the fatal issue. Para-Virtualization is another software solution, but it doesn't emulate anything, it let the guest OS run in the true hardware, but we must remove the "Sensitive Instructions" from the Guest OS, we must modify the code of the guest OS and recompile it. That means we must have the foll source of guest OS, all the guests can not be supported if we have the binary release only. That is the disadvantage in Para-Virtualization. Full Virtualization can combine the advantages in both emulator and Para-virtualization. That is the main purpose of this thesis.We focus on the principle and theory about how to design and implement a hardware virtual machine monitor (HVM). We will talk about the key features and technologies in foll virtualization. After the reading of this thesis, we hope the reader can understand how the VMM works, and he can write a simple VMM by following the directives and the references in this thesis.At first, we will introduce some basic background related VT, give the reader some overview concepts. These contents include the new theory about VMM: modeling view, the latest international trends, the categories, and the standardization.In Chapter 3, we will introduce CPU virtualization methodology with the support of Intel Vanderpool technology. The VMCS structure and the VMX instructions will be referred. We will highlight the SMP CPU virtualization finally.We talk about memory virtualization in Chapter 4; it is the most complicated module in VMM. It is a big challenge for us. Shadow Page Table is almost a general way. It was applied in many VMM, Such as VMware and KVM. We will show you how we design and implement it in our project.After we finished the design and implementation, Our VMM basically works. But it has some performance issues. We designed a test scenario to handle this issue, and then we found a performance bottleneck. Based on the root-cause analysis we got a perfect solution finally. In the end of this Chapter, We will show an integrated performance testing report about our VMM. This report comes from our QA team. We will compare the performance ratio between HVM and Para-virtualization, HVM and Native OS.The test report shows that we boot up the unmodified Guest OS on our VMM successfully. It has good performance. It reaches our expectations in the design. It proves the correctness of our suppositions and design. But there are still some promotion points, we can do it better and make it much more robust. We show that what we should do in next step.