Some time ago I made a benchmark exercise to compare the performance of an Oracle Database running in a bare metal environment versus a virtualized environment to clear some of the main questions our team and clients in Revera have, such as:
- Does an Oracle Database performs well on a virtualized environment?
- What virtualization technology is more stable and allows an Oracle database to perform faster?
- What is the performance difference between using a bare metal and a virtualized guest?
- Is it safe to run a production database in a virtualized environment?
Here you can find the results and the answers to the questions above: http://oraclenz.org/wp-content/uploads/2008/06/Why-use-OVM.pdf
Regards,
Francisco Munoz Alvarez
Francisco – This is a nice summary of your research. I have a couple questions.
(1) On VMs, was the storage a virtual file system, or a physical lun passed through to the guest?
(2) Did you use any sort of CPU pinning on the VMs?
(3) Was hyper-threading enabled either in the bare metal, or virtual environments?
Thank you.
Hi Pete,
Thank you so much for your feedback and here are the answers to your questions:
(1) Physical LUN passed directly to the guest (no virtual disks) for the OVM. And paravirtual storage driver used for the non-ovm machine due that it was the only way to improve performance in the non-ovm (Present the LUNS did not increased performance on it).
(2) No CPU Pinning used on this benchmark.
(3) Hyper-Threading was enable on all 3 physical servers.
Regards,
Francisco
Thanks for your research Francisco. Would there be a difference with hyperthreading off?