Research And Implementation Of Resource-Container-based Virtual Machine Technology

The recent increase in the amount of global servers has posed a severe challenge to many aspects of sever technique such as server costs, resource management, data integrity, and system security. The traditional use of the Virtual Machine technique, which was originally used in mainframe, meets the demands for many aspects of system security and server consolidation. This vigorous technique has been widely used in server research and development, and gets an unprecedented substantial development.This thesis compares and analyzes thoroughly some essential techniques of Virtual Machine and the two greatest trends of its development by combining some examples of existing virtual machines. Then we point out that most current studies of the Virtual Machine technique focus on improving server utilization and lowering TCO through server consolidation, however, they have neglected resource reservation and multi-grained resource management. As a result it has been difficult to meet the dynamic requirements of sever applications, making servers weak in the face of DoS attacks.To solve these problems, we propose a Resource-Container based Virtual Machine technique which includes three sub-techniques providing functions of application isolation, resource abstraction and resource control. The application isolation technique provides a virtual environment for applications and insulates the devices details; the resource abstraction technique abstract resources into virtual resource pools which consist of more fine-grained resource containers; and the resource control technique places boundaries on resource usage, preventing resource principal from over-consuming resources. In addition, we adopt a share-driven scheduling based on CPU-reservation to implement the multi-grained dynamic abstraction of CPU. And we make an improvement by adding priorities to that kind of scheduling.Finally, aiming at the requirements of the domestic Kylin Server OS to high-performance, high-availability, and high-security, we apply this technique into Kylin. KVM framework is designed and implemented in its kernel.