Systems support for virtualization

近年来，人们决心将硬件虚拟化作为一种技术来部署，以帮助管理现代计算机系统的规模和复杂性。硬件虚拟化允许对单个物理计算机进行分区，并使其看起来像是一组孤立的虚拟计算机，每个计算机运行独立的操作系统和应用程序堆栈。IBM AMD和英特尔已经发布了CPU扩展来帮助虚拟化，所有主要的操作系统供应商都宣布在他们的系统的下一个版本中支持虚拟化。该提案是为了一个研究计划，该计划探索随着虚拟化的大规模部署对系统软件的挑战和机遇，企业环境。在工业数据中心和租用服务器式的托管提供商等设施中使用虚拟化将极大地改变这些系统的许多方面，这些系统必须突然扩展以支持相同物理资源上的更多虚拟主机，并且由于入门成本较低，可能由经验较少的用户管理。此外，虚拟化提供了一个狭窄的介入层，系统软件可以独立于覆盖的操作系统和应用程序进行部署。作为短期目标，我描述了三个已经在UBC的硕士学生中进行的虚拟化项目。一组正在运行的服务器的远程副本，并允许此副本在出现故障时接管系统的操作。其次，我对数据包对称性强制执行的调查旨在对服务器传输某些类型的恶意流量的能力施加限制，保护托管设施不会无意中成为拒绝服务攻击的来源。最后，我考虑使用基于网络的内存的大缓存作为社区资源，以避免使用机械磁盘。

Recent years have seen a decided effort to deploy hardware virtualization as a technology to help manage the scale and  complexity of modern computer systems.  Virtualization allows a single physical computer to be partitioned and made to appear as a collection of isolated virtual computers, each running an independent operating system and application stack.  AMD and Intel have released CPU extensions to aid virtualization, and all major OS vendors have announced virtualization support in the next releases of their systems.This proposal is for a research program that explores the challenges and opportunities for systems software as virtualization is deployed in large-scale, enterprise environments.  The use of virtualization in facilities such as industrial datacenters and rent-a-server-style hosting providers dramatically changes many aspects of these systems, which must suddenly scale to support more virtual hosts on the same physical resources, and may be administered by considerably less experienced users due to lower entry costs.  Additionally, virtualization presents a narrow layer of interposition at which systems software may be deployed independent of overlying operating systems and applications.As short term goals, I describe three virtualization projects that are already underway with masters students at UBC.  "SecondSite" is a disaster recovery system that maintains a live, remote copy of a running group of servers and allows this copy to take over operation of the system in the face of failure.  Second, my investigation of packet symmetry enforcement aims to impose limits on the ability of servers to transmit certain types of malicious traffic, protecting hosting facilities from inadvertently being the source of denial-of-service-attacks.  Finally, I consider the use of large caches of network-based memory, as a community resource to avoid the use of mechanical disks.