Hi! I'm experimenting with KVM on my modest home server, and I ran into a peculiar issue that I can't explain, so I'm now looking for some insight. The problem I see is that raw sequential disk read speed from within a guest is somehow limited. I did some measurements and I don't really understand the result. My setup is as follows: Two SATA HDDs are set up with full disk encryption (LUKS) and used as physical volumes in an LVM volume group. On this VG I create a striped LV, and when measuring raw read speed on this I get about 113MB/s. The decryption is handled by two threads, each consuming about 50% of one CPU core. The raw disk speed for one disk is around 60-70MB/s, so this is as expected. When I instead connect these encrypted devices to a virtual machine using virtio, and doing the same measurements, I get the same raw disk speed when measuring only one disk, but when I measure on the striped volume, I only get about half the speed I expect. I can still see that the two kcryptd processes are running, but they now only consume about 25% CPU time each, and nothing else seem to be maxed out. The 'dd' process on the guest is consuming 9% CPU, and the kvm process on the host is using about 30% of one core. All timing is measured like this: time dd </dev/vgfs/striped bs=256k count=10000 >/dev/null I'd be grateful for any insight, since it might help me understand other bottlenecks I might stumble upon in the future. This disk speed issue is not much of a problem really, but I think it's odd. Just tell me if you want more details about the case. // Anders -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
