Further investigation of the performance overhead for hypervisor- and container-based virtualization.
(2020) EITM02 20192Department of Electrical and Information Technology
- Abstract
- This thesis work aims to find a suitable environment for different virtualization system to achieve enhanced performance. Virtualization plays a vital role in cloud services, and it is essential to help cloud users comprehend the brief distinction between different virtualization technologies. The hypervisor-based technology has been used as the primary selection for cloud services in the past, but container-based virtualization starts receiving more attention and is regarded as the substitution of hypervisor-based technology. This thesis, contains research and discussion of the performance overhead and performance variability overhead that a Virtual Machine or Docker container introduce under certain conditions. The challenge was to find... (More)
- This thesis work aims to find a suitable environment for different virtualization system to achieve enhanced performance. Virtualization plays a vital role in cloud services, and it is essential to help cloud users comprehend the brief distinction between different virtualization technologies. The hypervisor-based technology has been used as the primary selection for cloud services in the past, but container-based virtualization starts receiving more attention and is regarded as the substitution of hypervisor-based technology. This thesis, contains research and discussion of the performance overhead and performance variability overhead that a Virtual Machine or Docker container introduce under certain conditions. The challenge was to find a feasible solution to measure the parameters based on the performance, acting on the CPU, memory, hardware disk, and network throughput on the physical machine. These parameters will be used in the performance comparison of these two virtualization techniques. Moreover, in the 5G network, network function virtualization (NFV) is an efficient solution against the massive traffic, but it suffers from software-based solutions. The main concept in NFV is to decouple the network functions of the dedicated hardware working only for the given purpose. When setting up a virtualization proxy or deploying a web server within a virtual environment, it is necessary to determine if the virtualization method should use hypervisor-based (KVM) or container-based (Docker) technology. Therefore, we need to investigate as a special condition which technique will maximize the network data throughput and minimize the physical machine’s overhead. Our experiment presents the performance overhead when a webserver is running in the KVM and Docker virtualization environment. In most cases, the Docker virtualization performs better. KVM has similar performance overhead as Docker when the network is not under heavy load. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9003143
- author
- Cui, Zheng LU
- supervisor
- organization
- course
- EITM02 20192
- year
- 2020
- type
- H2 - Master's Degree (Two Years)
- subject
- keywords
- Hypervisor, Container, Virtualization, Network function virtualization, Software Defined Network
- report number
- LU/LTH-EIT 2020-743
- language
- English
- id
- 9003143
- date added to LUP
- 2020-02-06 15:38:32
- date last changed
- 2020-02-06 15:38:32
@misc{9003143,
abstract = {{This thesis work aims to find a suitable environment for different virtualization system to achieve enhanced performance. Virtualization plays a vital role in cloud services, and it is essential to help cloud users comprehend the brief distinction between different virtualization technologies. The hypervisor-based technology has been used as the primary selection for cloud services in the past, but container-based virtualization starts receiving more attention and is regarded as the substitution of hypervisor-based technology. This thesis, contains research and discussion of the performance overhead and performance variability overhead that a Virtual Machine or Docker container introduce under certain conditions. The challenge was to find a feasible solution to measure the parameters based on the performance, acting on the CPU, memory, hardware disk, and network throughput on the physical machine. These parameters will be used in the performance comparison of these two virtualization techniques. Moreover, in the 5G network, network function virtualization (NFV) is an efficient solution against the massive traffic, but it suffers from software-based solutions. The main concept in NFV is to decouple the network functions of the dedicated hardware working only for the given purpose. When setting up a virtualization proxy or deploying a web server within a virtual environment, it is necessary to determine if the virtualization method should use hypervisor-based (KVM) or container-based (Docker) technology. Therefore, we need to investigate as a special condition which technique will maximize the network data throughput and minimize the physical machine’s overhead. Our experiment presents the performance overhead when a webserver is running in the KVM and Docker virtualization environment. In most cases, the Docker virtualization performs better. KVM has similar performance overhead as Docker when the network is not under heavy load.}},
author = {{Cui, Zheng}},
language = {{eng}},
note = {{Student Paper}},
title = {{Further investigation of the performance overhead for hypervisor- and container-based virtualization.}},
year = {{2020}},
}