I've been using sqlio.exe to benchmark a new ssd drive and I get much higher results running it in a Hyper-V VM compared to the physical host, much higher.
For example, on the raw host I get 2000 IOPS, on the VM I get 5000 etc.
When I compare the physical disk stats in perfmon when the test is running it appears that the VM is writing at a rate of about 250MB/sec, while the host is only writing about 60MB/sec.
I'm using a 20GB datafile in the test to ensure nothing gets cached in memory etc and I know its writing to disk based on the host's perfmon counters.
Any ideas why I'm getting artificially high results in the VM?
View Complete Post