The system virtualization is arrested attention to the server consolidation for utilization and efficiency of system resources recently. This system virtualization means the software technology that offers a number of virtualized environments through ...
The system virtualization is arrested attention to the server consolidation for utilization and efficiency of system resources recently. This system virtualization means the software technology that offers a number of virtualized environments through separating physical hardwares of one machine. A virtualized environment provide independent virtual hardwares to each server of virtual machine although shared hardware for real. The system virtualization is already in practical use level and makes cost reduction with much less spaces at many commercial fields. So, there are many researches to improve the resource utilization and efficiency through using the system virtualization technique.
The system virtualization has two major types that are container-based and hypervisor-based virtualization. The container-based virtualization shares one kernel image and provides separated multiple user-space to each user. Because of the shared kernel, this type loads many virtual machines, but it can not support high-level resource isolation. Also, some fault tolerance of one virtual machine makes fatal fault to whole virtual machines due to shared resources. On the contrary, a virtualization layer of the hypervisor-based is on a physical hardware substrate so that supports high-level resource isolation to each virtualized environments. This type preoccupies resources to each virtual machine at creating virtual machine. Therefore a total resources of all virtual machine can not be over the real machine's resource capacity. This means the container-based type is able to load more virtual machine than hypervisor-based virtualization. And hypervisor-based type may product idle resources when some virtual machine needs more resource because of a characteristic of high-level resource isolation.
This paper perceives these problems and develops a system for improving resource utilization with high-level resource isolation. Especially, a memory makes a system to disable high-level resource isolation to container-based virtualization. Furthermore, it limits a number of virtual machines and generates idle resource to hypervisor-based virtualization. In this paper, I focus on this memory resource. Finally, I suggest the policy which predicts next memory usage and allocates memory dynamically to virtual machines.