diff options
Diffstat (limited to 'results/classifier/zero-shot/118/performance/1139')
| -rw-r--r-- | results/classifier/zero-shot/118/performance/1139 | 108 |
1 files changed, 108 insertions, 0 deletions
diff --git a/results/classifier/zero-shot/118/performance/1139 b/results/classifier/zero-shot/118/performance/1139 new file mode 100644 index 000000000..93e9f514a --- /dev/null +++ b/results/classifier/zero-shot/118/performance/1139 @@ -0,0 +1,108 @@ +performance: 0.959 +network: 0.927 +device: 0.909 +ppc: 0.884 +files: 0.853 +graphic: 0.852 +socket: 0.845 +mistranslation: 0.842 +peripherals: 0.813 +hypervisor: 0.810 +virtual: 0.807 +architecture: 0.804 +kernel: 0.797 +register: 0.793 +PID: 0.784 +VMM: 0.783 +vnc: 0.777 +boot: 0.774 +assembly: 0.770 +semantic: 0.765 +x86: 0.749 +debug: 0.745 +permissions: 0.734 +user-level: 0.715 +arm: 0.678 +risc-v: 0.676 +i386: 0.661 +KVM: 0.630 +TCG: 0.613 + +block/nbd.c and drive backup to a remote nbd server +Description of problem: +Good afternoon! + +I trying to copy attached drive content to remote NBD server via drive-backup QMP method. I'he tested two very similar ways but with very different performance. First is a backuping to exported NBD at another server. Second way is a backuping to same server but with connecting to /dev/nbd*. + +Exporting qcow2 via nbd: +``` +(nbd) ~ # qemu-nbd -p 12345 -x backup --cache=none --aio=native --persistent -f qcow2 backup.qcow2 + +(qemu) ~ # qemu-img info nbd://10.0.0.1:12345/backup +image: nbd://10.0.0.1:12345/backup +file format: raw +virtual size: 10 GiB (10737418240 bytes) +disk size: unavailable +``` + +Starting drive backuping via QMP: + +``` +{ + "execute": "drive-backup", + "arguments": { + "device": "disk", + "sync": "full", + "target": "nbd://10.0.0.1:12345/backup", + "mode": "existing" + } +} +``` + +With process starting qemu notifying about warning: + +> warning: The target block device doesn't provide information about the block size and it doesn't have a backing file. The default block size of 65536 bytes is used. If the actual block size of the target exceeds this default, the backup may be unusable + +And backup process is limited by speed around 30MBps, watched by iotop + + +Second way to creating backup + +Exporting qcow2 via nbd: +``` +(nbd) ~ # qemu-nbd -p 12345 -x backup --cache=none --aio=native --persistent -f qcow2 backup.qcow2 +``` + +``` +(qemu) ~ # qemu-img info nbd://10.0.0.1:12345/backup +image: nbd://10.0.0.1:12345/backup +file format: raw +virtual size: 10 GiB (10737418240 bytes) +disk size: unavailable +(qemu) ~ # qemu-nbd -c /dev/nbd0 nbd://10.0.0.1:12345/backup +(qemu) ~ # qemu-img info /dev/nbd0 +image: /dev/nbd0 +file format: raw +virtual size: 10 GiB (10737418240 bytes) +disk size: 0 B +``` + +Starting drive backuping via QMP to local nbd device: + +``` +{ + "execute": "drive-backup", + "arguments": { + "device": "disk", + "sync": "full", + "target": "/dev/nbd0", + "mode": "existing" + } +} +``` + +Backup process started without previous warning, and speed limited around 100MBps (network limit) + +So I have question: how I can get same performance without connection network device to local block nbd device at the qemu host? + +Kind regards |