Skip site navigation (1)Skip section navigation (2)
Date:      Sun, 19 Jan 2014 02:10:12 +0100
From:      Sydney Meyer <meyer.sydney@googlemail.com>
To:        "freebsd-virtualization@freebsd.org" <freebsd-virtualization@freebsd.org>
Subject:   Re: Xen PVHVM with FreeBSD10 Guest
Message-ID:  <76E20919-2457-407F-9564-7A032658C515@googlemail.com>
In-Reply-To: <52D94E74.8000401@citrix.com>
References:  <9DF57091-9957-452D-8A15-C2267F66ABEC@googlemail.com> <52D81009.6050603@citrix.com> <51F93577-E5A2-4237-9EDD-A89DDA5FC428@gmail.com> <F672F9F6-7F85-4315-AFA0-EA18527A1893@googlemail.com> <52D8F301.2080701@citrix.com> <A5ECBAEA-8E55-4D95-87D0-5FFC1D559BA3@googlemail.com> <52D94E74.8000401@citrix.com>

next in thread | previous in thread | raw e-mail | index | archive | help
Hello Roger,

here are the results. I=92ve tested FreeBSD 9.2, 10.0-RC5 in conjunction =
with XEN PVHVM,XEN-QEMU-DM, KVM-VirtIO and the bare metal performance of =
the hosts, every run respectively with file IO via raw images and block =
IO via LVM volumes. Each VM ran with 1GB of memory and 1 VCPU. The =
results are the average of 3 consecutive runs with bonnie, writing each =
time the double size of the hosts memory. Xen 4.1 with Linux 3.2 Dom0 =
and QEMU 0.10.2 on Intel i3 4310T / 8 GB, KVM with Linux 3.11 on Intel =
i7 860 / 12 GB.

Please keep in mind that these results are more of an =93subjective=94 =
impression rather a scientific comparison held under the exact same =
conditions, etc.

http://pastebin.com/pUZfXda7

+----------+-----+-------+--------+------------+-----------+
|    OS    | HV  | Disk  |   IO   | Write MB/s | Read MB/s |
+----------+-----+-------+--------+------------+-----------+
| 9.2      | Xen | File  | QEMU   | 60,5       | 111,2     |
| 9.2      | Xen | File  | PV     | 43,2       | 106,5     |
| 9.2      | Xen | Block | QEMU   | 49,4       | 91,0      |
| 9.2      | Xen | Block | PV     | 72,9       | 100,4     |
| 10.0     | Xen | File  | PV     | 38,6       | 107,2     |
| 10.0     | Xen | Block | PV     | 72,6       | 99,8      |
| 9.2      | KVM | File  | VirtIO | 40,6       | 138,1     |
| 9.2      | KVM | Block | VirtIO | 44,5       | 131,6     |
| 10.0     | KVM | File  | VirtIO | 38,1       | 131,4     |
| 10.0     | KVM | Block | VirtIO | 43,2       | 134,7     |
| Xen Host |     |       |        | 90,1       | 109,0     |
| KVM Host |     |       |        | 93,4       | 126,1     |
+----------+-----+-------+--------+------------+-----------+

As you can see, there are no leads to my inital observation (Emulated =
Block IO =93faster=94 than PVHVM Block IO). I will take a closer look =
what went wrong first time, probably some sort of caching issue.=20

Anyhow, i=92ve read that you have been working together with others to =
bring PVHVM support into GENERIC, so i wanted to thank you and all the =
other developers for making FreeBSD 10.0 another great release of this =
rock solid, awesome Operating System.

Native Xen PV- and therefore freebsd-update support are a very big plus =
in our evergoing fight against management over the choice of weapons.

Cheers,
S.

On 17.01.2014, at 16:38, Roger Pau Monn=E9 <roger.pau@citrix.com> wrote:

> On 17/01/14 10:17, Sydney Meyer wrote:
>> I=92m doing some benchmarks with bonnie and dd on the Variations =
9.2/10.0;PVHVM/VirtIO;fileio/blockio. I will post the results here to =
this thread.
>=20
> By VirtIO I guess you mean emulated IO? That sounds great, I'm eager =
to
> see the results :)
>=20
> Roger.




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?76E20919-2457-407F-9564-7A032658C515>