When I run iotop on a kvm server I see the qemu processes for KVM being listed as between 1MB/s and 8MB/s for writes, the aggregate of those processes is about 10MB/s. This isn't an impossible number as the image files for KVM are stored on a RAID-1 array of SSDs. Reads tend to be about zero because there is a lot of cache and the system has been running for a while. When I run iostat on the block devices for the RAID-1 I see the average write speed reported at something that's always below 2MB/s and which is often below 1MB/s for a 10 second reporting period. Could iotop be considering writes to /dev/kvm or /dev/net/tun as disk IO? What else could be the explanation for this? How could I work this out? -- My Main Blog http://etbe.coker.com.au/ My Documents Blog http://doc.coker.com.au/