CSR: Small: Multi-hypervisor Virtual Machines - Enabling an Ecosystem of Hypervisors in the Cloud

点击翻译按钮获取中文摘要

Modern cloud computing platforms offer users an array of choices in operating systems, database systems, financial software, virtual network routers, and so forth, all of which can be easily customized and executed within Virtual Machines (VM). Unfortunately, this level of user choice has not extended to the underlying systems software, called the hypervisor, which enables the cloud infrastructure to run VMs. On one hand, cloud providers have an opportunity to differentiate their service by offering rich hypervisor-level services such as rootkit detection, live patching, intrusion detection, high availability, and VM introspection. On the other hand, it is difficult for a cloud provider to develop and maintain a single hypervisor that incorporates every feature that each cloud user wants. This research is to enable a software ecosystem in which third-parties can also contribute hypervisor-level services to the cloud platform and augment the services of a base hypervisor. To use multiple hypervisor-level services, a cloud user can run a multi-hypervisor virtual machine, which is envisioned as an unmodified VM that simultaneously runs atop multiple co-existing hypervisors. There are two thrusts of this project. The first is to develop the systems software to transparently run multi-hypervisor VMs, where multiple hypervisors co-operatively exert control over the memory, virtual CPUs, and I/O devices of a VM using nested virtualization. The second thrust is to investigate mechanisms for the co-existence of various hypervisor-level services and demonstrate a multi-hypervisor ecosystem of services such as those listed above. By enabling an ecosystem of third-party hypervisors, the technology behind multi-hypervisor VM can significantly improve the functionality and performance of the cloud platforms. Integration of research with education includes course modules and projects, and an advanced topics graduate course. Broader impacts include mentoring undergraduate and under-represented students in research, outreach to local high schools, and technology transfer.