A CPU Allocation Method Considering the Control Domain Overhead in Xen Virtualization Environment
Hyun-Jun Yun, Young-Rok Ko, Junyoung Kim, Sungyong Park
Abstract
Hyun-Jun Yun, Young-Rok Ko, Junyoung Kim, Sungyong Park
Abstract
In virtualization environment, automated management frameworks are necessary to meet users' requirements such as expected response time. The existing management frameworks working in the Xen virtualization environment estimate workload of each virtual machine. Based on this estimated result, the management frameworks make a decision of allocating CPU to each virtual machine in order that requests are responded within expected response time. But existing management frameworks' CPU allocation policies of the virtual machines do not reflect the overhead of domain-0 into the CPU allocation. By that, responding to request within expected response time can be failed. In this paper, we propose the CPU allocation technique that considers the overhead of domain-0. Proposed technique attempts to improve the I/O performance of domain-U, one of the primary cause of the domain-0's overhead. Proposed technique reflects I/O workload and CPU workload to the CPU allocation rate with different ratio. After that it tries to prevent situation that domain-0 cannot obtain enough CPU resource by work conserving scheduling of Xen credit scheduler. With the technique, domain-0 can obtain enough CPU resource to make domain-Us respond within expected response time.
OpenAlex reports 2 citations for this work. Citation counts describe recorded attention and do not establish research quality.
A contribution statement is not available in the OpenAlex record.
Method details are not available in the OpenAlex metadata.
Findings are not separately available in the OpenAlex metadata.
Limitations are not available in the OpenAlex metadata.
Application details are not available in the OpenAlex metadata.
In virtualization environment, automated management frameworks are necessary to meet users' requirements such as expected response time. The existing management frameworks working in the Xen virtualization environment estimate workload of each virtual machine. Based on this estimated result, the management frameworks make a decision of allocating CPU to each virtual machine in order that requests are responded within expected response time. But existing management frameworks' CPU allocation policies of the virtual machines do not reflect the overhead of domain-0 into the CPU allocation. By that, responding to request within expected response time can be failed. In this paper, we propose the CPU allocation technique that considers the overhead of domain-0. Proposed technique attempts to improve the I/O performance of domain-U, one of the primary cause of the domain-0's overhead. Proposed technique reflects I/O workload and CPU workload to the CPU allocation rate with different ratio. After that it tries to prevent situation that domain-0 cannot obtain enough CPU resource by work conserving scheduling of Xen credit scheduler. With the technique, domain-0 can obtain enough CPU resource to make domain-Us respond within expected response time.
Key concepts: Computer science, CPU shielding, Virtualization, Central processing unit, Virtual machine, Workload, Overhead (engineering), Domain (mathematical analysis)