After a lot of feedback on the first posting about the improved iSCSI performance in vSphere 4, I decided that I should re-test the values and see what performance I can get. So here it goes...
Now for the setup we are using a RAID 1/0 array with 4 1TB Western Digital Black drives, this does two things gives lots of space and some fast performance on the cheap. The last test I did was centered on the jumbo frames v. non-jumbo frames; a lot of that data was based off observations and timings of events (i.e. deploying VM's, cloning VM's, etc). That tested the vSphere implementation of the iSCSI protocol without mixing in the virtual layer. Now for the next set of tests I wanted to see what did the guests see for timings and speed.
Here is the testing setup (the hardware and such is the same):
- Windows 2008 Standard Edition
- 2GB RAM
- Dual Processor VM
- 1 Virtual Hard Drives for the OS 40GB Thin Provisioned
- 1 Virtual Hard Drives for testing Thick Provisioned Eager Zero'ed
Results with 1500 MTU:
Not too bad average of 54MB/s and access times of 7.5ms
Results of 9000 MTU:
Same though... 54.3MB/s and the same access times of 7.5ms
So my earlier post of MTU 9000 being slower while possibly true outside of the guests but it is not true for inside the guests where things really matter. So upgrade to 9000 MTU and call it a day and enjoy your newly configured network without the worry about performance or CPU impacts.
Bookmark/Search this post with:
