A Study on CPU Time Reservation for QoS Guarantee in a Virtual Machine Environment

我们研究了一个能够支持调度理论和CPU时间预留的资源管理框架的虚拟机环境。调度理论使得通过将CPU时间预留请求集成到虚拟机中来为该虚拟机提供足够的CPU时间成为可能。在虚拟机环境中，有两个调度级别，一个在操作系统上，另一个在虚拟机上。操作系统调度一个虚拟机，并且在该虚拟机中调度一个任务。因此，必须集成CPU时间预留请求，以便操作系统可以调度虚拟机，以满足其中的CPU时间预留请求。资源管理框架的作用是根据调度理论实际分配CPU时间。该框架将虚拟机器的调度参数传递给操作系统内核，由虚拟机器为每个任务重新分配CPU时间，从而完成任务的CPU时间预留请求，并开发了一个虚拟机器环境原型，为上述调度理论和资源管理框架奠定了基础。该原型运行在Intel IA-32处理器上，支持Linux和RTOS作为操作系统。

We studied a virtual machine environment, in which the scheduling theory and the resource management framework for CPU time reservation can be enabled. The scheduling theory makes it possible to give enough CPU time to a virtual machine by integrating CPU time reservation requests within the virtual machine. In a virtual machine environment, there are two scheduling levels, one at an operating system and the other in a virtual machine. The operating system schedules a virtual machine, and a task is scheduled in the virtual machine. Therefore, CPU time reservation requests must be integrated, so that the operating system can schedule a virtual machine in order to meet the CPU time reservation requests within it. The resource management framework takes a role of actually assigning the CPU time based on the scheduling theory. The framework gives the scheduling parameter of a virtual machine to the operating system kernel, and the virtual machine reassigns CPU time to each task, so that the task's CPU time reservation request is fulfilled.We also developed a prototype of a virtual machine environment, which can become the base of the scheduling theory and the resource management framework described above. The prototype operates on the Intel IA-32 processor, and supports Linux and RTOS as operating systems.