UNSOLVED
J
jeff_browning
2 Intern
•
256 Posts
0
889
August 22nd, 2011 09:00
Virtualized vs. Physical Performance on Various Storage Configs
This issue arose in the context of review the presentation for Sam's and my up-coming VMworld talk. Basically the idea is to compare virtualized vs. physical performance in a variety of storage contexts. NetApp did something similar a recently, but it did not include virtualization.
I would dearly love to see EMC do something like this (although it would be a lot of work). This would compare the following:
- ASM / FC / physical
- ASM / FC / RDM-P
- ASM / FC / RDM-V
- VMFS / FC
- Same 4 above on iSCSI / 10 GbE
- Same 4 above on FCoE / 10 GbE
- dNFS physical / 10 GbE
- dNFS virtual / 10 GbE
- NFS datastore / .vmdk / 10 GbE
The last is probably the least interesting, but I include it for completeness. I would suggest a TPC-C workload, as that is the most interesting database benchmark for us. Let me know if you think this would be useful.
I am not only interested in the performance results but also interested in VMware and storage tuning to agile with database requirement, since I believe OLTP system should be a Cache-friendly and CPU-sensitive system, how do we lower bare lay impacting with Physical CPU/RAM converting to vCPU/vRAM, OLAP system is IO bandwidth sensitive system, but VMware FS is clustering FS with lots of lock, do we have any improvement on it?
Eddy
taceyr
98 Posts
273
0
Posted August 22nd, 2011 09:00
I'd certainly be interested in the results. Being able to provide test results during a physical vs. virtual discussion would certainly be a feather in anyone's cap.