On Wed, Mar 28, 2012 at 07:00:22PM -0700, hung -cuncon wrote: > # dd if=/dev/zero of=/mnt/test bs=2M count=7862
You need oflag=direct so the I/O will be performed to the underlying device instead of just the page cache. This will make the comparison fairer since the guest has 3 GB RAM and the host has 8 GB RAM (more page cache). It's best to eliminate page cache if you are trying to look at actual I/O performance rather than overall system performance. Please post your kvm command-line so we can see the guest's configuration in detail. If you are interested in understanding the difference between host I/O and guest I/O patterns, it's worth capturing a tcpdump of the iSCSI session so that you can see what the I/O pattern boils down to: request sizes, flush commands, request timing, etc. Stefan -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
