Performance Evaluation of Virtualization Technologies for Server Consolidation

TLDR

Server consolidation reduces cost and improves efficiency, and virtualization enables multiple applications to run on shared physical servers, yet selecting the appropriate technology and configuration remains a challenge. The study evaluates Xen and OpenVZ virtualization technologies across different configurations. The authors consolidated multi‑tiered systems onto one or two nodes, used the RUBiS auction workload, and compared Xen and OpenVZ against a base system on application performance, resource consumption, scalability, cache misses, and Domain‑0 usage. Experiments show that as the number of application instances increases, Xen’s average response time rises over 400% while OpenVZ’s increases only about 100%, due to higher virtualization overhead in Xen manifested in greater cache misses, instruction misses, and CPU consumption, as confirmed by kernel‑symbol profiling.

Abstract

Server consolidation has become an integral part of IT planning to reduce cost and improve efficiency in today’s enterprise data centers. The advent of virtualization allows consolidation of multiple applications into virtual containers hosted on a single or multiple physical servers. However, this poses new challenges, including choosing the right virtualization technology and consolidation configuration for a particular set of applications. In this paper, we evaluate two representative virtualization technologies, Xen and OpenVZ, in various configurations. We consolidate one or more multi-tiered systems onto one or two nodes and drive the system with an auction workload called RUBiS. We compare both technologies with a base system in terms of application performance, resource consumption, scalability, low-level system metrics like cache misses and virtualization-specific metrics like Domain-0 consumption in Xen. Our experiments indicate that the average response time can increase by over 400% in Xen and only a modest 100% in OpenVZ as the number of application instances grows from one to four. This large discrepancy is caused by the higher virtualization overhead in Xen, which is likely due to higher L2 cache misses and misses per instruction. A similar trend is observed in CPU consumptions of virtual containers. We present an overhead analysis with kernelsymbol-specific information generated by Oprofile.