On Sun, Apr 15, 2007 at 06:28:10PM -0700, Curtis Doty wrote:
8:33pm Daniel P. Berrange said:
The pure-compute results look reasonable, but I don't believe those results for Disk I/O in fully-virt for a second - Section 6, Disk I/O Unixbench is the exact opposite of expectation - were results flipped?. Based knowledge of the architecture
And is it even fair to compare tap:aio: to file: disks without even a mention of the technical difference?
It is as fair as you'll get for file-backed VMs. 'file:' for fully-virt doesn't have the same disadvantages / problems that 'file:' does for paravirt, since its actually better to compare 'tap:aio:' for PV with 'file:' for FV that to compare 'file:' with PV with 'file:' for FV.
It would certainly be worth doing 2 sets of disk I/O tests though, one with file backed disks, and the other with raw partition / LVM backed disks.
Dan.