[Bug 238733] kvm disk i/o extremely slow

bugzilla-noreply at freebsd.org bugzilla-noreply at freebsd.org
Fri Jun 21 07:17:47 UTC 2019


https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=238733

            Bug ID: 238733
           Summary: kvm disk i/o extremely slow
           Product: Base System
           Version: 12.0-RELEASE
          Hardware: Any
                OS: Any
            Status: New
          Severity: Affects Only Me
          Priority: ---
         Component: kern
          Assignee: bugs at FreeBSD.org
          Reporter: rainer at ultra-secure.de

Hi,

I've posted this on the forum:
https://forums.freebsd.org/threads/is-kvm-virtio-really-that-slow-on-freebsd.71186/#post-430366


But I'll reiterate here.

Basically, I'm running a FreeBSD 12 image created alongside this document:
https://docs.openstack.org/image-guide/freebsd-image.html
(plus swap)

On a KVM host that is, TTBOMK, Ubuntu 18. Storage is Ceph Blockstorage, using
pure NVMe SSDs - but I/O is capped at 2000 IOP/s.


Using CentOS as a KVM guest, I basically get this:

[root at centos ~]# dc3dd wipe=/dev/sdb

dc3dd 7.1.614 started at 2019-06-20 07:54:22 +0000
compiled options:
command line: dc3dd wipe=/dev/sdb
device size: 83886080 sectors (probed)
sector size: 512 bytes (probed)
42949672960 bytes (40 G) copied (100%), 342.37 s, 120 M/s                     

input results for pattern `00':
   83886080 sectors in

output results for device `/dev/sdb':
   83886080 sectors out

dc3dd completed at 2019-06-20 08:00:05 +0000


Using FreeBSD, I get this:


root at freebsd:~ # dc3dd wipe=/dev/vtbd2

dc3dd 7.2.646 started at 2019-06-20 09:37:10 +0200
compiled options:
command line: dc3dd wipe=/dev/vtbd2
device size: 83886080 sectors (probed),   42,949,672,960 bytes
sector size: 512 bytes (probed)
 42949672960 bytes ( 40 G ) copied ( 100% ), 4585 s, 8.9 M/s                  

input results for pattern `00':
   83886080 sectors in

output results for device `/dev/vtbd2':
   83886080 sectors out

dc3dd completed at 2019-06-20 10:53:35 +0200


The forum posting has fio-runs with the actual I/O numbers.


Are the drivers really that bad?

If this is the situation, FreeBSD is completely useless in the cloud. At least
for us.

-- 
You are receiving this mail because:
You are the assignee for the bug.


More information about the freebsd-bugs mailing list