Quantitative system performance: computer system analysis using queueing network models
Quantitative system performance: computer system analysis using queueing network models
Xen and the art of virtualization
SOSP '03 Proceedings of the nineteenth ACM symposium on Operating systems principles
Intel Virtualization Technology
Computer
A comparison of software and hardware techniques for x86 virtualization
Proceedings of the 12th international conference on Architectural support for programming languages and operating systems
Quantifying the performance isolation properties of virtualization systems
Proceedings of the 2007 workshop on Experimental computer science
Towards modeling & analysis of consolidated CMP servers
ACM SIGARCH Computer Architecture News
Generating Adaptation Policies for Multi-tier Applications in Consolidated Server Environments
ICAC '08 Proceedings of the 2008 International Conference on Autonomic Computing
The evolution of virtualization
Communications of the ACM - Being Human in the Digital Age
Database systems on virtual machines: How much do you lose?
ICDEW '08 Proceedings of the 2008 IEEE 24th International Conference on Data Engineering Workshop
HotCloud'09 Proceedings of the 2009 conference on Hot topics in cloud computing
SR-IOV networking in Xen: architecture, design and implementation
WIOV'08 Proceedings of the First conference on I/O virtualization
Workload management for power efficiency in virtualized data centers
Communications of the ACM
Communications of the ACM
| Hi-index | 0.00 |
An application's performance can suffer from significant computational overheads when it is moved from a native to a virtualized environment. Adoption of virtualization without understanding such overheads in detail can dramatically impact the overall performance of hosted applications. The rapid adoption of virtualization has fueled the development of new hardware technologies, which promise to optimize the performance and scalability of processor and network I/O virtualization. However, no comprehensive empirical study of the effectiveness of these hardware assistance technologies is publicly available. In this paper we focus on x86 architectures and study empirically the performance improvements introduced by Intel's VT and PCI-SIG's SR-IOV on a Xen-based hypervisor. Using a range of benchmark programs, we compare benchmark scores and resource utilization between native and virtual environments for two different testbeds, one with hardware assistance and one without. The results indicate that hardware assistance indeed eliminates most overheads, especially those relating to network I/O, but non-negligible CPU overheads still remain. Also, there is no hardware technology with specifically deals with disk I/O virtualization, and significant overheads do arise in workloads requiring intensive disk usage.