Optimization of Network Bandwidth Allocation in Xen

Through virtualization, a plurality of servers can be simulated on a single physical machine. The sharing of hardware resources of one physical machine by multiple virtual servers greatly improves resource utilization. The key to virtualization is how to handle the allocation and management of hardw...

Full description

Saved in:
Bibliographic Details
Published in2015 IEEE 17th International Conference on High Performance Computing and Communications, 2015 IEEE 7th International Symposium on Cyberspace Safety and Security, and 2015 IEEE 12th International Conference on Embedded Software and Systems pp. 1558 - 1566
Main Authors LiRong Mei, Xiang Lv
Format Conference Proceeding
LanguageEnglish
Published IEEE 01.08.2015
Subjects
Bandwidth
bandwidth allocation
Channel allocation
Linux
Resource management
Servers
Virtual machining
Virtualization
Xen
Online AccessGet full text

Cover

More Information
Summary:Through virtualization, a plurality of servers can be simulated on a single physical machine. The sharing of hardware resources of one physical machine by multiple virtual servers greatly improves resource utilization. The key to virtualization is how to handle the allocation and management of hardware resources among competing virtual machines. One type of hardware resource that has not been fully studied in this regard is network bandwidth, which, if dealt poorly, can have significant negative impact on the performance of a server cluster. This paper analyzes the bandwidth allocation strategy of Xen, an open source hypervisor. We find that Xen uses a static network allocation approach that is based on the parameters specified in a configuration file. Such approach greatly limits the utilization of network resource. To overcome this drawback, this paper presents a dynamic bandwidth allocation strategy based on the demand and priority of each virtual machine. Our evaluation demonstrates that the proposed solution makes much better use of network resources with relatively low overhead.
DOI:10.1109/HPCC-CSS-ICESS.2015.35