Xen users - Apr 2008 - Release 0.8.9 of GPL PV drivers for Windows

Here''s the latest version. Took me ages to get to the bottom of what
turned out to be a pretty simple problem - windows can give us more
pages in a packet than Linux can handle but Linux (netback) doesn''t
complain about it, it just creates corrupt packets :(

Download from http://www.meadowcourt.org/WindowsXenPV-0.8.9.zip

I''ll probably do another release very shortly, mainly to reduce memory
consumption on the tx side so that more interfaces can run at once.
>From the testing I''ve done, on a UP windows DomU, with iperf
options ''-l
1M -w 1M'', with the iperf server running in Dom0, I get TX throughput
of
about 1.5Gbits/second and RX throughput of about 0.5Gbits/second. When I
tried it under SMP it worked, but the performance was horrible. Probably
best if you don''t run it under SMP for the moment :)

I did have a test performance of 2.5Gbits/second, but now that I have to
copy the windows buffers into my own buffers to reduce page usage, I
seem to only be able to get about 1.5Gbits/second out of it... This kind
of makes sense given that DomU to Dom0 network performance is going to
be CPU and Memory bandwidth bound.

James

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

Hi James, great news. As I''m in vacation, I won''t be able to
test it but
hope it works well.

Did you consider sending an email to the kernel dev guys to inform them
about this issue? maybe it''s by design, maybe there''s
something that can be
improved.

Emre

On Sat, Apr 26, 2008 at 9:49 AM, James Harper
<james.harper@bendigoit.com.au>
wrote:
> Here''s the latest version. Took me ages to get to the bottom of
what
> turned out to be a pretty simple problem - windows can give us more
> pages in a packet than Linux can handle but Linux (netback)
doesn''t
> complain about it, it just creates corrupt packets :(
>
> Download from http://www.meadowcourt.org/WindowsXenPV-0.8.9.zip
>
> I''ll probably do another release very shortly, mainly to reduce
memory
> consumption on the tx side so that more interfaces can run at once.
>
> >From the testing I''ve done, on a UP windows DomU, with iperf
options ''-l
> 1M -w 1M'', with the iperf server running in Dom0, I get TX
throughput of
> about 1.5Gbits/second and RX throughput of about 0.5Gbits/second. When I
> tried it under SMP it worked, but the performance was horrible. Probably
> best if you don''t run it under SMP for the moment :)
>
> I did have a test performance of 2.5Gbits/second, but now that I have to
> copy the windows buffers into my own buffers to reduce page usage, I
> seem to only be able to get about 1.5Gbits/second out of it... This kind
> of makes sense given that DomU to Dom0 network performance is going to
> be CPU and Memory bandwidth bound.
>
> James
>
> _______________________________________________
> Xen-users mailing list
> Xen-users@lists.xensource.com
> http://lists.xensource.com/xen-users
>


-- 
Emre Erenoglu
erenoglu@gmail.com


_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

> 
> Did you consider sending an email to the kernel dev guys to inform
them
> about this issue? maybe it''s by design, maybe there''s
something that
can
> be improved.
> 
Not really. Changing it would have a fairly high impact (eg Dom0 and all
Linux DomU''s on the system would have to support the higher number or
all sorts of strange things would happen).

The only thing you could thinking about defining as a bug is that
netback didn''t complain about it anywhere, it just pretended everything
was okay and put corrupt packets onto the bridge.

James

_______________________________________________
Xen-devel mailing list
Xen-devel@lists.xensource.com
http://lists.xensource.com/xen-devel

On Saturday April 26 2008 02:49:59 am James Harper
wrote:
> I did have a test performance of 2.5Gbits/second, but now that I have to
> copy the windows buffers into my own buffers to reduce page usage, I
> seem to only be able to get about 1.5Gbits/second out of it... This kind
> of makes sense given that DomU to Dom0 network performance is going to
> be CPU and Memory bandwidth bound.
What was the downside of the higher page usage?

Looking forward to testing it (or at least loading it :-).

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

Hello Jim,

jim burns schrieb:
> Now that ''PCI BUS'' still has the Microsoft driver, and
not xenhide, what is
> the recommended uninstall procedure, prior to installing a new version? 
> Thanx.
>   
I used the uninstall.bat and then the install.bat. Worked quite good on 
my system.
I don''t know if this is the recommended procedure.

Greetz Age_M

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Saturday April 26 2008 02:49:59 am James Harper
wrote:
> Here''s the latest version.
Now that ''PCI BUS'' still has the Microsoft driver, and not
xenhide, what is
the recommended uninstall procedure, prior to installing a new version? 
Thanx.

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Saturday April 26 2008 11:22:14 am Age_M wrote:
> I used the uninstall.bat and then the install.bat. Worked quite good on
> my system.
> I don''t know if this is the recommended procedure.
Doh! I didn''t notice that uninstall.bat is new in 0.8.9. Thanx.

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

The problem isn''t so much page usage as grant table usage. There is a
finite limit and it was being reached too quickly, eg with more than a few
network interfaces loaded.

-----Original Message-----
From: "jim burns"<jim_burn@bellsouth.net>
Sent: 27/04/08 1:22:53 AM
To:
"xen-users@lists.xensource.com"<xen-users@lists.xensource.com>
Subject: Re: [Xen-users] Release 0.8.9 of GPL PV drivers for Windows

On Saturday April 26 2008 02:49:59 am James Harper
wrote:
> I did have a test performance of 2.5Gbits/second, but now that I have to
> copy the windows buffers into my own buffers to reduce page usage, I
> seem to only be able to get about 1.5Gbits/second out of it... This kind
> of makes sense given that DomU to Dom0 network performance is going to
> be CPU and Memory bandwidth bound.
What was the downside of the higher page usage?

Looking forward to testing it (or at least loading it :-).

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

> jim burns schrieb:
> > Now that ''PCI BUS'' still has the Microsoft driver,
and not xenhide,
what
> is
> > the recommended uninstall procedure, prior to installing a new
version?
> > Thanx.
> >
> I used the uninstall.bat and then the install.bat. Worked quite good
on
> my system.
> I don''t know if this is the recommended procedure.
> 
Hmmm... uninstall.bat is a work in progress and probably shouldn''t have
been included. I suspect that if you rebooted after running
uninstall.bat but before running install.bat you would have had an
unbootable system.

The problem is the way I attach xenhide.sys to the pci bus, which is via
an AddReg in the .inf file. I think some simple .vbs would be able to
clear the registry key and the uninstall.bat file could just call that.

All in all though, I am still uncomfortable with the way I hide the qemu
network and ata drivers from windows, and I keep telling myself that
there must be a better way... everything I try works even worse than
before though :)

James

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Saturday April 26 2008 10:12:33 pm James Harper
wrote:
> Hmmm... uninstall.bat is a work in progress and probably shouldn''t
have
> been included. I suspect that if you rebooted after running
> uninstall.bat but before running install.bat you would have had an
> unbootable system.
Well, for the first time since 0.8.4, I got winxp to boot w/ /gplpv, using 
0.8.9. I booted w/o /gplpv and then ran uninstall.bat. Because of the warning 
above, I ran install.bat right away w/o a reboot. The final screen came up 
and said all drivers were updated except xennet, which was ''Ready to
Use''.
Then I rebooted w/ /gplpv, and the Find New Hardware Wizard came up 
automatically and guided me through installing xennet w/o a hitch.

I had previously disabled all the new features in Device Manager''s Xen
Net
Device Driver Properties'' Advanced Tab. (This was done by copying over
the
0.8.4 files so I could boot w/o a BSOD.) An ''iperf -c dom0-name -t
60'' came
up with 27.3 Mbits/s. I then proceeded to turn on each feature one at a time 
in the Advanced tab, and rebooting w/ /gplpv. Device Manager invariably hung 
after enabling each feature, which caused the reboot to hang. All 
measurements are with vcpus=2, unless noted otherwise.

After adding enabling Checksum Offload, iperf gave 30.5 Mb/s.

After adding setting 61440 for Large Send Offload, iperf gave 25.3 Mb/s.

After adding enabling Scatter/Gather, iperf gave 25.1 Mb/s.

So there are minor variations with and w/o the various options, but on the 
whole, much better than the last version I could test, 0.8.4.

From James'' original post:
> From the testing I''ve done, on a UP windows DomU, with iperf
options ''-l
> 1M -w 1M'', with the iperf server running in Dom0, I get TX
throughput of
> about 1.5Gbits/second and RX throughput of about 0.5Gbits/second. When I
> tried it under SMP it worked, but the performance was horrible. Probably
> best if you don''t run it under SMP for the moment :)
Doing ''iperf -c dom0-name -l 1M -w 1M'' gives 28.8 Mb/s, and
reversing the
direction (winxp as iperf server) gives 30.2 Mb/s.

Going down to vcpu=1, dom0 as server gives 27.1 Mb/s, and domu as server gives 
35.7 Mb/s, so there is not a lot of difference between 1 and 2 vcpus for me.

Nice improvements. I will test disk i/o w/ iometer later.

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

James Harper schrieb:
> Here''s the latest version. Took me ages to get to the bottom of
what
> turned out to be a pretty simple problem - windows can give us more
> pages in a packet than Linux can handle but Linux (netback)
doesn''t
> complain about it, it just creates corrupt packets :(
> 
> Download from http://www.meadowcourt.org/WindowsXenPV-0.8.9.zip
Sounds, nice, when i get the time these days, I''ll sign this stuff to 
make it acceptable for my 64Bit 2008. Then lets have a look.

Florian
> 
> I''ll probably do another release very shortly, mainly to reduce
memory
> consumption on the tx side so that more interfaces can run at once.
> 
>> >From the testing I''ve done, on a UP windows DomU, with
iperf options ''-l
> 1M -w 1M'', with the iperf server running in Dom0, I get TX
throughput of
> about 1.5Gbits/second and RX throughput of about 0.5Gbits/second. When I
> tried it under SMP it worked, but the performance was horrible. Probably
> best if you don''t run it under SMP for the moment :)
> 
> I did have a test performance of 2.5Gbits/second, but now that I have to
> copy the windows buffers into my own buffers to reduce page usage, I
> seem to only be able to get about 1.5Gbits/second out of it... This kind
> of makes sense given that DomU to Dom0 network performance is going to
> be CPU and Memory bandwidth bound.
> 
> James
> 
> _______________________________________________
> Xen-users mailing list
> Xen-users@lists.xensource.com
> http://lists.xensource.com/xen-users
> 


_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Sun, Apr 27, 2008 at 03:16:09AM -0400, jim burns
wrote:
> On Saturday April 26 2008 10:12:33 pm James Harper wrote:
> > Hmmm... uninstall.bat is a work in progress and probably
shouldn''t have
> > been included. I suspect that if you rebooted after running
> > uninstall.bat but before running install.bat you would have had an
> > unbootable system.
> 
> Well, for the first time since 0.8.4, I got winxp to boot w/ /gplpv, using 
> 0.8.9. I booted w/o /gplpv and then ran uninstall.bat. Because of the
warning
> above, I ran install.bat right away w/o a reboot. The final screen came up 
> and said all drivers were updated except xennet, which was ''Ready
to Use''.
> Then I rebooted w/ /gplpv, and the Find New Hardware Wizard came up 
> automatically and guided me through installing xennet w/o a hitch.
> 
> I had previously disabled all the new features in Device Manager''s
Xen Net
> Device Driver Properties'' Advanced Tab. (This was done by copying
over the
> 0.8.4 files so I could boot w/o a BSOD.) An ''iperf -c dom0-name -t
60'' came
> up with 27.3 Mbits/s. I then proceeded to turn on each feature one at a
time
> in the Advanced tab, and rebooting w/ /gplpv. Device Manager invariably
hung
> after enabling each feature, which caused the reboot to hang. All 
> measurements are with vcpus=2, unless noted otherwise.
> 
> After adding enabling Checksum Offload, iperf gave 30.5 Mb/s.
> 
> After adding setting 61440 for Large Send Offload, iperf gave 25.3 Mb/s.
> 
> After adding enabling Scatter/Gather, iperf gave 25.1 Mb/s.
> 
> So there are minor variations with and w/o the various options, but on the 
> whole, much better than the last version I could test, 0.8.4.
> 
Just to make sure, you mean megabytes (MB/sec) instead of megabits per
second (Mb/sec) ?

25 - 30 Mbit/sec would be really slow..
> From James'' original post:
> > From the testing I''ve done, on a UP windows DomU, with iperf
options ''-l
> > 1M -w 1M'', with the iperf server running in Dom0, I get TX
throughput of
> > about 1.5Gbits/second and RX throughput of about 0.5Gbits/second. When
I
> > tried it under SMP it worked, but the performance was horrible.
Probably
> > best if you don''t run it under SMP for the moment :)
> 
> Doing ''iperf -c dom0-name -l 1M -w 1M'' gives 28.8 Mb/s,
and reversing the
> direction (winxp as iperf server) gives 30.2 Mb/s.
> 
> Going down to vcpu=1, dom0 as server gives 27.1 Mb/s, and domu as server
gives
> 35.7 Mb/s, so there is not a lot of difference between 1 and 2 vcpus for
me.
> 
> Nice improvements. I will test disk i/o w/ iometer later.
> 
IOmeter results would be really nice!

-- Pasi

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

> On Saturday April 26 2008 10:12:33 pm James Harper wrote:
> > Hmmm... uninstall.bat is a work in progress and probably
shouldn''t
have
> > been included. I suspect that if you rebooted after running
> > uninstall.bat but before running install.bat you would have had an
> > unbootable system.
> 
> Well, for the first time since 0.8.4, I got winxp to boot w/ /gplpv,
using
> 0.8.9. I booted w/o /gplpv and then ran uninstall.bat. Because of the
> warning
> above, I ran install.bat right away w/o a reboot. The final screen
came up
> and said all drivers were updated except xennet, which was ''Ready
to
Use''.
> Then I rebooted w/ /gplpv, and the Find New Hardware Wizard came up
> automatically and guided me through installing xennet w/o a hitch.
> 
> I had previously disabled all the new features in Device Manager''s
Xen
Net
> Device Driver Properties'' Advanced Tab. (This was done by copying
over
the
> 0.8.4 files so I could boot w/o a BSOD.) An ''iperf -c dom0-name -t
60''
> came up with 27.3 Mbits/s. I then proceeded to turn on each feature
one at > a time in the Advanced tab, and rebooting w/ /gplpv. Device
Manager 
> invariably hung after enabling each feature, which caused the reboot
to
> hang.
Ouch. I was confident I''d fixed all of those hanging problems. I can
enable and disable those settings all I like and it never misses a beat
- the driver shuts itself down and restarts without a problem. Can you
tell me more about your test system?
> All measurements are with vcpus=2, unless noted otherwise.
> 
> After adding enabling Checksum Offload, iperf gave 30.5 Mb/s.
> 
> After adding setting 61440 for Large Send Offload, iperf gave 25.3
Mb/s.
> 
> After adding enabling Scatter/Gather, iperf gave 25.1 Mb/s.
Scatter/Gather disabled isn''t really a tested configuration, but the
others should work, although LSO disabled when Dom0 has it enabled could
be problematic...

What system are you running? My main test machine is a 1.8GHz dual core
AMD 1210 system, and I get 1GBit/sec TX and 200MBit/sec RX with all
features enabled. So your figures are lower than I''d expect, given that
any system that supports HVM should be at least comparable to my test
machine in terms of horsepower...

With LSO disabled my testing drops to 283 TX and 6 RX...
> So there are minor variations with and w/o the various options, but on
the
> whole, much better than the last version I could test, 0.8.4. 
Well... that''s something :)
> 
> Doing ''iperf -c dom0-name -l 1M -w 1M'' gives 28.8 Mb/s,
and reversing
the
> direction (winxp as iperf server) gives 30.2 Mb/s.
You can achieve the same results with -r to do a TX test followed by an
RX (eg client/server roles reverse) or -d to do the TX and RX tests
concurrently.
> Going down to vcpu=1, dom0 as server gives 27.1 Mb/s, and domu as
server
> gives
> 35.7 Mb/s, so there is not a lot of difference between 1 and 2 vcpus
for
> me.
With all features enabled?
> Nice improvements. I will test disk i/o w/ iometer later.
I''ll be interested in the results, but the disk stuff hasn''t
changed in
a while.

Thanks for the feedback!

James

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

> On Sun, Apr 27, 2008 at 03:16:09AM -0400, jim burns wrote:
> > On Saturday April 26 2008 10:12:33 pm James Harper wrote:
> > > Hmmm... uninstall.bat is a work in progress and probably
shouldn''t
> have
> > > been included. I suspect that if you rebooted after running
> > > uninstall.bat but before running install.bat you would have had
an
> > > unbootable system.
> >
> > Well, for the first time since 0.8.4, I got winxp to boot w/ /gplpv,
> using
> > 0.8.9. I booted w/o /gplpv and then ran uninstall.bat. Because of the
> warning
> > above, I ran install.bat right away w/o a reboot. The final screen
came
> up
> > and said all drivers were updated except xennet, which was
''Ready to
> Use''.
> > Then I rebooted w/ /gplpv, and the Find New Hardware Wizard came up
> > automatically and guided me through installing xennet w/o a hitch.
> >
> > I had previously disabled all the new features in Device
Manager''s Xen
> Net
> > Device Driver Properties'' Advanced Tab. (This was done by
copying over
> the
> > 0.8.4 files so I could boot w/o a BSOD.) An ''iperf -c
dom0-name -t 60''
> came
> > up with 27.3 Mbits/s. I then proceeded to turn on each feature one at
a
> time
> > in the Advanced tab, and rebooting w/ /gplpv. Device Manager
invariably
> hung
> > after enabling each feature, which caused the reboot to hang. All
> > measurements are with vcpus=2, unless noted otherwise.
> >
> > After adding enabling Checksum Offload, iperf gave 30.5 Mb/s.
> >
> > After adding setting 61440 for Large Send Offload, iperf gave 25.3
Mb/s.
> >
> > After adding enabling Scatter/Gather, iperf gave 25.1 Mb/s.
> >
> > So there are minor variations with and w/o the various options, but on
> the
> > whole, much better than the last version I could test, 0.8.4.
> >
> 
> Just to make sure, you mean megabytes (MB/sec) instead of megabits per
> second (Mb/sec) ?
> 
> 25 - 30 Mbit/sec would be really slow..
I hadn''t thought of that when I wrote my previous email. If he did mean
Mbytes as opposed to the Mbits that I assumed then it makes a lot more sense.
Iperf gives the results in Mbits so that is what I assumed...

James

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Sat, 26 Apr 2008 16:49:59 +1000, "James Harper"
<james.harper@bendigoit.com.au> wrote:
> Here''s the latest version. Took me ages to get to the bottom of
what
> turned out to be a pretty simple problem - windows can give us more
> pages in a packet than Linux can handle but Linux (netback)
doesn''t
> complain about it, it just creates corrupt packets :(
> 
> Download from http://www.meadowcourt.org/WindowsXenPV-0.8.9.zip
> 
> I''ll probably do another release very shortly, mainly to reduce
memory
> consumption on the tx side so that more interfaces can run at once.
> 
>>From the testing I''ve done, on a UP windows DomU, with iperf
options ''-l
> 1M -w 1M'', with the iperf server running in Dom0, I get TX
throughput of
> about 1.5Gbits/second and RX throughput of about 0.5Gbits/second. When I
> tried it under SMP it worked, but the performance was horrible. Probably
> best if you don''t run it under SMP for the moment :)
> 
> I did have a test performance of 2.5Gbits/second, but now that I have to
> copy the windows buffers into my own buffers to reduce page usage, I
> seem to only be able to get about 1.5Gbits/second out of it... This kind
> of makes sense given that DomU to Dom0 network performance is going to
> be CPU and Memory bandwidth bound.
> 
> James
Hi James

I''ve tested this release on my system (fresh install) and I''m
still getting
the duplicate disk problem when I boot with the /gplpv option.

There has been some talk lately that this may be a fault of the Red Hat
kernel.  FWIW I''m running CentOS 5.1 64bit, kernel is
2.6.18-53.1.14.el5xen.


Cheers
Scott



_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Sunday April 27 2008 07:36:46 am James Harper wrote:
> Ouch. I was confident I''d fixed all of those hanging problems. I
can
> enable and disable those settings all I like and it never misses a beat
> - the driver shuts itself down and restarts without a problem. Can you
> tell me more about your test system?
From a previous benchmark post, ''Release 0.8.0 of GPL PV Drivers for
Windows'':

Equipment: core duo 2300, 1.66ghz each, 2M, sata drive configured for UDMA/100
System: fc8 32bit pae, xen 3.1.2, xen.gz 3.1.3 [updated from 3.1.0-rc7], dom0 
2.6.21
Tested hvm: XP Pro SP2, 2002 w/512M

Method:

The version tested was 1.7.0, to avoid having to apply the kernel patch that 
comes with 2.0.2. The binaries downloaded were from the project homepage 
http://dast.nlanr.net/Projects/Iperf/#download. For linux, I chose the
''Linux
libc 2.3 binary and (on fc8 at least) I still had to install the 
compat-libstdc++-33 package to get it to run.
> Scatter/Gather disabled isn''t really a tested configuration, but
the
> others should work, although LSO disabled when Dom0 has it enabled could
> be problematic...
It isn''t:
[822] > ethtool -k peth0
Offload parameters for peth0:
Cannot get device rx csum settings: Operation not supported
Cannot get device tx csum settings: Operation not supported
Cannot get device scatter-gather settings: Operation not supported
Cannot get device tcp segmentation offload settings: Operation not supported
Cannot get device udp large send offload settings: Operation not supported
rx-checksumming: off
tx-checksumming: off
scatter-gather: off
tcp segmentation offload: off
udp fragmentation offload: off
generic segmentation offload: off
jimb@Insp6400 04/27/08 11:15AM:~
[823] > lspci|grep Broad
03:00.0 Ethernet controller: Broadcom Corporation BCM4401-B0 100Base-TX (rev 
02)

 and thus the b44 driver.
>> Doing ''iperf -c dom0-name -l 1M -w 1M'' gives 28.8
Mb/s, and reversing
the
>> direction (winxp as iperf server) gives 30.2 Mb/s.
> You can achieve the same results with -r to do a TX test followed by an
> RX (eg client/server roles reverse) or -d to do the TX and RX tests
> concurrently.
Ok:
iperf -c dom0-name -l 1M -w 1M -r gave me a BSOD:
DRIVER_IRQL_NOT_LESS_OR_EQUAL
[...]
*** STOP: 0x000000D1 (0x8945365C,0x00000002,0x00000001,0xF87EE6B4)

***    xennet.sys - Address F87EE6B4 base at F87EA000, DateStamp 4812c73b

Rebooting and trying again gave me minor variations in the STOP line:
*** STOP: 0x000000D1 (0x895D973C,0x00000002,0x00000001,0xF87FE6B4)

The dom0 output was:
[501] > iperf -s
------------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 85.3 KByte (default)
------------------------------------------------------------
[  4] local 192.168.1.100 port 5001 connected with 192.168.1.102 port 1038
[ ID] Interval       Transfer     Bandwidth
[  4]  0.0-10.2 sec  30.7 MBytes  25.3 Mbits/sec
------------------------------------------------------------
Client connecting to 192.168.1.102, TCP port 5001
TCP window size:  256 KByte (WARNING: requested 1.00 MByte)
------------------------------------------------------------
[  4] local 192.168.1.100 port 33125 connected with 192.168.1.102 port 5001
Waiting for server threads to complete. Interrupt again to force quit.

Running higher window size on the server also gave:
[502] > iperf -s -l 1M -w 1M
------------------------------------------------------------
Server listening on TCP port 5001
TCP window size:  256 KByte (WARNING: requested 1.00 MByte)
------------------------------------------------------------

and so was not pursued further, ''iperf -c dom0-name -l 256k -w 256k
-r'' gave
on dom0:
[505] > iperf -s
------------------------------------------------------------
Server listening on TCP port 5001
TCP window size: 85.3 KByte (default)
------------------------------------------------------------
[  4] local 192.168.1.100 port 5001 connected with 192.168.1.102 port 1037
[ ID] Interval       Transfer     Bandwidth
[  4]  0.0-10.1 sec  41.5 MBytes  34.6 Mbits/sec
------------------------------------------------------------
Client connecting to 192.168.1.102, TCP port 5001
TCP window size:  256 KByte (WARNING: requested  256 KByte)
------------------------------------------------------------
[  4] local 192.168.1.100 port 53348 connected with 192.168.1.102 port 5001
[ ID] Interval       Transfer     Bandwidth
[  4]  0.0-10.1 sec  38.8 MBytes  32.3 Mbits/sec

Isn''t debugging fun :-)
>> Going down to vcpu=1, dom0 as server gives 27.1 Mb/s, and domu as
server
>> gives
>> 35.7 Mb/s, so there is not a lot of difference between 1 and 2 vcpus
for
>> me.
> With all features enabled?
Yes, I left all features enabled after getting them where a new install would 
put them.
>> Nice improvements. I will test disk i/o w/ iometer later.
> I''ll be interested in the results, but the disk stuff
hasn''t changed in
a while.

I kind of thought so, since you haven''t mentioned it, but I''ll
post the
results later anyway.

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Sunday April 27 2008 01:09:57 pm you wrote:
> That''s really slow.. how much do you get between pv linux domU and
dom?
Currently 580 Mb/s. I really haven''t done anything to my Fedora pv domu
that
would affect connectivity since the 0.8.0 benchmarks, where I reported 1563 
Mb/s. Yeah, I know - a step backwards, but still more than I need. (The only 
Fedora pv change was updating to Rawhide. Dom0 was updated from xen.gz 
3.1.0-rc7 to 3.1.3. Xen is still 3.1.2.)

We already covered this ground on my system''s performance in that same 
thread ''Release 0.8.0 of GPL PV Drivers for Windows'', where
you had me test a
bunch of settings.

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Sun, Apr 27, 2008 at 01:44:04PM -0400, jim burns
wrote:
> On Sunday April 27 2008 01:09:57 pm you wrote:
> > That''s really slow.. how much do you get between pv linux
domU and dom?
> 
> Currently 580 Mb/s. I really haven''t done anything to my Fedora pv
domu that
> would affect connectivity since the 0.8.0 benchmarks, where I reported 1563
> Mb/s. Yeah, I know - a step backwards, but still more than I need. (The
only
> Fedora pv change was updating to Rawhide. Dom0 was updated from xen.gz 
> 3.1.0-rc7 to 3.1.3. Xen is still 3.1.2.)
> 
> We already covered this ground on my system''s performance in that
same
> thread ''Release 0.8.0 of GPL PV Drivers for Windows'',
where you had me test a
> bunch of settings.
Hmm.. so hvm winxp with pv drivers is around 20x slower for you than pv
linux domU.. 

btw did you have other domains running while your tests were running? 

Have you tried VMware btw? Would be nice to see results with other
virtualization software too..

-- Pasi

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

Scott McKenzie-3 wrote:
> 
> 
> I''ve tested this release on my system (fresh install) and
I''m still
> getting
> the duplicate disk problem when I boot with the /gplpv option.
> 
> There has been some talk lately that this may be a fault of the Red Hat
> kernel.  FWIW I''m running CentOS 5.1 64bit, kernel is
> 2.6.18-53.1.14.el5xen.
> 
> 
Scott,

I havent tried this latest version, but its good (in a bad way!) to hear
some else having the same duplication issues i''ve been having.
I''m running
rhel51 which is identical? to centos51..

So sounds like a versioning issue with xen/dom0 kernel in rhel51

When I get the time, I might attempt using rhel beta 5.2 as the dom0

cheers,

James

-- 
View this message in context:
http://www.nabble.com/Release-0.8.9-of-GPL-PV-drivers-for-Windows-tp16910023p16930327.html
Sent from the Xen - User mailing list archive at Nabble.com.


_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

> Scott McKenzie-3 wrote:
> >
> >
> > I''ve tested this release on my system (fresh install) and
I''m still
> > getting
> > the duplicate disk problem when I boot with the /gplpv option.
> >
> > There has been some talk lately that this may be a fault of the Red
Hat
> > kernel.  FWIW I''m running CentOS 5.1 64bit, kernel is
> > 2.6.18-53.1.14.el5xen.
> >
> >
> 
> Scott,
> 
> I havent tried this latest version, but its good (in a bad way!) to
hear
> some else having the same duplication issues i''ve been having.
I''m
running
> rhel51 which is identical? to centos51..
> 
> So sounds like a versioning issue with xen/dom0 kernel in rhel51
> 
> When I get the time, I might attempt using rhel beta 5.2 as the dom0
> 
> cheers,
> 
Scott/James,

Would it be possible for me to get remote desktop (port 3389) access to
one of your systems? I would like to check a few things.

Thanks

James


_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

James,

Nope, there are at least 4 firewalls that will stop you..

I can however  do some debugging if you like?

cheers

James
=======================================================================Electricity
Networks Corporation, trading as Western Power
ABN: 18 540 492 861

TO THE ADDRESSEE - this email is for the intended addressee only and may contain
information that is confidential.
If you have received this email in error, please notify us immediately by return
email or by telephone.
Please also destroy this message and any electronic or hard copies of this
message.

Any claim to confidentiality is not waived or lost by reason of mistaken
transmission of this email.

Unencrypted email is not secure and may not be authentic.  Western Power cannot
guarantee the accuracy, reliability,
completeness or confidentiality of this email and any attachments.

VIRUSES - Western Power scans all outgoing emails and attachments for viruses,
however it is the recipient''s responsibility
to ensure this email is free of viruses.
=======================================================================

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

> James,
> 
> Nope, there are at least 4 firewalls that will stop you..
:)
> I can however  do some debugging if you like?
> 
I''ll send you a list of the stuff I want to know.

Thanks

James

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Monday April 28 2008 12:42:57 am Pasi Kärkkäinen
wrote:
> Hmm.. so hvm winxp with pv drivers is around 20x slower for you than pv
> linux domU..
Hence my interest in the performance of the PV drivers. I''m consigned
to
*some* slow down in an hvm domain where every paging request is a software 
trap. Actually, my biggest complaint is the speed of the video
''adapter''.
Anything higher than DSL speeds on the net ''adapter'' is
acceptable, and James
is already well past that. A little less sluggishness would be nice on 
the ''disk'' also, though I realize there are limits to file
backed vbds.
> btw did you have other domains running while your tests were running?
Just an idle Fedora vm. Dom0 wasn''t doing anything in particular also.
> Have you tried VMware btw? Would be nice to see results with other
> virtualization software too..
I played with Vmware Player before Vmware Workstation was free. Was frustrated 
by the lack of Windows appliances that only ran on Windows. My whole interest 
in virtualization is so I don''t have to keep ''hitting the
reset button'', and
I have that now with Xen. Kvm also didn''t support Windows at the time.
I
played with Uml at the beginning, but SuSE was starting to phase it out, and 
I could never get it to work. Vmware is probably an acceptable alternative 
nowadays, as is qemu and variants, but I''ve chosen Xen. This gives me
more
than enough topics to learn about - iscsi, HA, PV drivers, PCI passthrough, 
etc., and I probably don''t have the time to devote to another product
now.

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Mon, Apr 28, 2008 at 02:18:00AM -0400, jim burns
wrote:
> On Monday April 28 2008 12:42:57 am Pasi Kärkkäinen wrote:
> > Hmm.. so hvm winxp with pv drivers is around 20x slower for you than
pv
> > linux domU..
> 
> Hence my interest in the performance of the PV drivers. I''m
consigned to
> *some* slow down in an hvm domain where every paging request is a software 
> trap. Actually, my biggest complaint is the speed of the video
''adapter''.
> Anything higher than DSL speeds on the net ''adapter'' is
acceptable, and James
> is already well past that. A little less sluggishness would be nice on 
> the ''disk'' also, though I realize there are limits to
file backed vbds.
> 
> > btw did you have other domains running while your tests were running?
> 
> Just an idle Fedora vm. Dom0 wasn''t doing anything in particular
also.
>
OK.
 
> > Have you tried VMware btw? Would be nice to see results with other
> > virtualization software too..
> 
> I played with Vmware Player before Vmware Workstation was free. Was
frustrated
> by the lack of Windows appliances that only ran on Windows. My whole
interest
> in virtualization is so I don''t have to keep ''hitting the
reset button'', and
> I have that now with Xen. Kvm also didn''t support Windows at the
time. I
> played with Uml at the beginning, but SuSE was starting to phase it out,
and
> I could never get it to work. Vmware is probably an acceptable alternative 
> nowadays, as is qemu and variants, but I''ve chosen Xen. This gives
me more
> than enough topics to learn about - iscsi, HA, PV drivers, PCI passthrough,
> etc., and I probably don''t have the time to devote to another
product now.
> 
Yeah.. I was just wondering that it would be nice to get some performance
numbers on the exact same hardware using vmware accelerated drivers.. to
compare with xen windows gplpv drivers.

-- Pasi

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

Hi guys,
I''ve done a Windos ITA fresh install under Xen in a non production
envirement and installed the GPL PV drivers. Now I have duplicated
(working) entries in my hardware list, but I think it''s not correct..

If you are interested I can tell what I''ve done:

1. Install a Debian Etch + xen stable + quemu (I don''t think
it''s really
relevant)
2. Install Windows XP SP2 ITA under xen without (in)security updates
from Bill at the moment
3. Downloaded and extracted 0.8.9 GPL PV drivers
4. Launched install.bat (I didn''t build anything, is it correct?)
4a. It showed me a console which tells: Windows XP detected... Install..
    and a ShotdownMon.exe error message:
    Application error 0xc0000135 -> OK to close.
    Then all goes as expected.
5. Finally I''ve rebooted non gplpv, founded new hardware, changed
boot.ini, rebooted gplpv, and founded duplicated entries..

Now I have two CDs, one QEMU and one XEN PV Scsi,  both working, two
network cards, same as above, two HD controllers but the HD is on the
old QEMU''s one..
I''ve disabled the Realtek NIC by hand and still have networking throw
XEN PV.

If there is something more I can tell you, ask.

Mic
	


_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

Michele Castigliego schrieb:
> Hi guys,
> I''ve done a Windos ITA fresh install under Xen in a non production
> envirement and installed the GPL PV drivers. Now I have duplicated
> (working) entries in my hardware list, but I think it''s not
correct..
> 
> If you are interested I can tell what I''ve done:
> 
> 1. Install a Debian Etch + xen stable + quemu (I don''t think
it''s really
> relevant)
> 2. Install Windows XP SP2 ITA under xen without (in)security updates
> from Bill at the moment
> 3. Downloaded and extracted 0.8.9 GPL PV drivers
> 4. Launched install.bat (I didn''t build anything, is it correct?)
> 4a. It showed me a console which tells: Windows XP detected... Install..
>     and a ShotdownMon.exe error message:
>     Application error 0xc0000135 -> OK to close.
This error occurs if the Microsoft.NET Framework 2.0 is not installed.
shutdownmon needs this framework, the drivers itself should work without
it.
>     Then all goes as expected.
> 5. Finally I''ve rebooted non gplpv, founded new hardware, changed
> boot.ini, rebooted gplpv, and founded duplicated entries..
> 
> Now I have two CDs, one QEMU and one XEN PV Scsi,  both working, two
> network cards, same as above, two HD controllers but the HD is on the
> old QEMU''s one..
This is an often reported bug, AFAIK there is no real solution. In this
situation you''re at high risk of corrupting your domU filesystem. You
shouldn''t boot /gplpv if this occurs. I think it has to do with some
xen / qemu-dm device namespace as it depends which version of xen
you''re
running. It looks like internationalized versions of windows are hit more
often by this bug.
> I''ve disabled the Realtek NIC by hand and still have networking
throw
> XEN PV.
To use xen pv nic''s, I''ve always changed the vif =[...] from
type=ioemu
to type=netfront and changed the mac address as i found windows getting
confused by different nic''s with the same mac.
> 
> If there is something more I can tell you, ask.
> 
> Mic
> 	
> 
> 
> _______________________________________________
> Xen-users mailing list
> Xen-users@lists.xensource.com
> http://lists.xensource.com/xen-users

-- 
Stephan Seitz
Senior System Administrator

*netz-haut* e.K.
multimediale kommunikation

zweierweg 22
97074 würzburg

fon: +49 931 2876247
fax: +49 931 2876248

web: www.netz-haut.de <http://www.netz-haut.de/>

registriergericht: amtsgericht würzburg, hra 5054



_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

Scott McKenzie wrote:
> On Sat, 26 Apr 2008 16:49:59 +1000, "James Harper"
> <james.harper@bendigoit.com.au> wrote:
>   
>> Here''s the latest version. Took me ages to get to the bottom
of what
>> turned out to be a pretty simple problem - windows can give us more
>> pages in a packet than Linux can handle but Linux (netback)
doesn''t
>> complain about it, it just creates corrupt packets :(
>>
>> Download from http://www.meadowcourt.org/WindowsXenPV-0.8.9.zip
>>
>> I''ll probably do another release very shortly, mainly to
reduce memory
>> consumption on the tx side so that more interfaces can run at once.
>>
>> >From the testing I''ve done, on a UP windows DomU, with
iperf options ''-l
>> 1M -w 1M'', with the iperf server running in Dom0, I get TX
throughput of
>> about 1.5Gbits/second and RX throughput of about 0.5Gbits/second. When
I
>> tried it under SMP it worked, but the performance was horrible.
Probably
>> best if you don''t run it under SMP for the moment :)
>>
>> I did have a test performance of 2.5Gbits/second, but now that I have
to
>> copy the windows buffers into my own buffers to reduce page usage, I
>> seem to only be able to get about 1.5Gbits/second out of it... This
kind
>> of makes sense given that DomU to Dom0 network performance is going to
>> be CPU and Memory bandwidth bound.
>>
>> James
>>     
>
> Hi James
>
> I''ve tested this release on my system (fresh install) and
I''m still getting
> the duplicate disk problem when I boot with the /gplpv option.
>
> There has been some talk lately that this may be a fault of the Red Hat
> kernel.  FWIW I''m running CentOS 5.1 64bit, kernel is
> 2.6.18-53.1.14.el5xen.
>
>
> Cheers
> Scott
>
>   
I''ve just installed openSUSE on my system to test the drivers with
their
kernel and Xen version.  I took a copy of my Windows HVM, booted it, 
installed 0.8.9, rebooted without /gplpv, rebooted with /gplpv and I had 
two disk devices appearing in device manager.  So it doesn''t look like 
it''s the dom0 kernel that''s causing this problem.



_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Sunday April 27 2008 12:10:35 pm Jim Burns wrote:
> >> Nice improvements. I will test disk i/o w/ iometer later.
> >
> > I''ll be interested in the results, but the disk stuff
hasn''t changed in
>
> a while.
>
> I kind of thought so, since you haven''t mentioned it, but
I''ll post the
> results later anyway.
And here we go:

Equipment: core duo 2300, 1.66ghz each, 2M, sata drive configured for UDMA/100
System: fc8 32bit pae, xen 3.1.2, xen.gz 3.1.3, dom0 2.6.21
Tested hvm: XP Pro SP2, 2002 w/512M, file backed vbd on local disk, tested w/ 
iometer 2006-07-27 (1Gb \iobw.tst, 5min run) & iperf 1.7.0 (1 min run)

Previous iometer results with 0.8.4 (from the ''New binary release of
GPL PV
drivers for Windows'' thread on 02/24, which was using xen.gz
3.1.0-rc7):

pattern 4k, 50% read, 0% random

(dynamo is the windows or linux client doing the actual work)
dynamo on?  |   io/s   |  MB/s  | Avg. i/o time(ms} | max i/o time(ms) | %CPU
domu w/gplpv|   273.0  |   1.07 |      431.52       |       0          | 32.27
domu w/qemu |   251.6  |   0.98 |       10.05       |       0          | 28.44
dom0 w/4Gb  |  1040.1  |   4.06 |        0.96       |      395.4       |  0
dom0 w/4Gb  |   808.1  |   3.16 |        1.24       |      977.1       |  0
(2nd dom0 numbers from when booted w/o /gplpv)

pattern 32k, 50% read, 0% random

domu w/gplpv|   161.6  |   5.05 |       -3.32       |        0         | 19.80
domu w/qemu |   109.0  |   3.41 |       -8.93       |        0         | 25.35
dom0 w/4Gb  |   140.7  |   4.40 |        7.10       |      467.6       |  0
dom0 w/4Gb  |   159.3  |   4.98 |        6.28       |      270.1       |  0


And now the 0.8.9 results:

pattern 4k, 50% read, 0% random

dynamo on?  |   io/s   |  MB/s  | Avg. i/o time(ms} | max i/o time(ms) | %CPU
domu w/gplpv|   331.5  |   1.29 |      232.29       |       0          | 35.63
domu w/qemu |   166.1  |   0.65 |        9.67       |       0          | 35.09
dom0 w/4Gb  |  1088.3  |   4.25 |        0.92       |      487.4       |  0
dom0 w/4Gb  |  1118.0  |   4.37 |        0.89       |      181.3       |  0
(2nd dom0 numbers from when booted w/o /pv)

pattern 32k, 50% read, 0% random

domu w/gplpv|   166.0  |   5.19 |        7.98       |        0         | 29.85
domu w/qemu |   100.4  |   3.14 |       21.09       |        0         | 35.93
dom0 w/4Gb  |    61.8  |   1.93 |       16.14       |     1492.3       |  0
dom0 w/4Gb  |   104.9  |   3.28 |        9.54       |      906.6       |  0

Despite some odd anomalies in the 0.8.9 dom0 32k pattern results, the general 
results are that 0.8.9 is marginally faster than 0.8.4. Domu 32k patterns are 
closer to dom0 performance than 4k patterns. 0.8.9 is much faster than qemu, 
compared to the 0.8.4 vs. qemu numbers, mostly because today''s qemu
numbers
were slower than the previous ones.

And now for something totally different: I just upgraded my processor from an 
Intel Core Duo 2300, 1.66Mhz to a Core 2 Duo 5600, 1.86 Mhz. Here''s
some new
iometer results:

pattern 4k, 50% read, 0% random

dynamo on?  |   io/s   |  MB/s  | Avg. i/o time(ms} | max i/o time(ms) | %CPU
domu w/gplpv|   501.7  |   1.96 |        2.90       |       0          | 31.68
domu w/qemu |   187.5  |   0.73 |        5.87       |       0          | 29.89
dom0 w/4Gb  |  1102.3  |   4.31 |        0.91       |      445.5       |  0
dom0 w/4Gb  |  1125.8  |   4.40 |        0.89       |      332.1       |  0
(2nd dom0 numbers from when booted w/o /pv)

pattern 32k, 50% read, 0% random

domu w/gplpv|   238.3  |   7.45 |        4.09       |        0         | 22.48
domu w/qemu |   157.4  |   4.92 |        6.35       |        0         | 20.51
dom0 w/4Gb  |    52.5  |   1.64 |       19.05       |     1590.0       |  0
dom0 w/4Gb  |    87.8  |   2.74 |       11.39       |     1286.4       |  0

So, between the two processors, the new one gives qemu and dom0 numbers that 
are modestly faster, and gplpv numbers that are 50% greater.

As far as iperf goes, ''iperf -c dom0-name -t 60'' gives 10
Mbits/s w/o /gplpv,
and 32.1 Mbits/s w/ /gplpv. This was with all advanced options turned on for 
the PV nic. My previous number for 0.8.9 w/ the old processor was 25 Mbits/s.

And I haven''t even enables 64 bits yet!

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Mon, May 05, 2008 at 12:47:39AM -0400, jim burns
wrote:
> On Sunday April 27 2008 12:10:35 pm Jim Burns wrote:
> > >> Nice improvements. I will test disk i/o w/ iometer later.
> > >
> > > I''ll be interested in the results, but the disk stuff
hasn''t changed in
> >
> > a while.
> >
> > I kind of thought so, since you haven''t mentioned it, but
I''ll post the
> > results later anyway.
> 
> And here we go:
> 
> Equipment: core duo 2300, 1.66ghz each, 2M, sata drive configured for
UDMA/100
> System: fc8 32bit pae, xen 3.1.2, xen.gz 3.1.3, dom0 2.6.21
> Tested hvm: XP Pro SP2, 2002 w/512M, file backed vbd on local disk, tested
w/
> iometer 2006-07-27 (1Gb \iobw.tst, 5min run) & iperf 1.7.0 (1 min run)
>
Hmm.. have you tried LVM backed devices for HVM guest? Or raw devices.. 

Could you try iometer on dom0 to see what kind of performance you get
there.. or on linux pv domU? 

And one more thing.. was your XP HVM single vcpu or more? Did you try binding
both dom0 and hvm domU to their own dedicated cpu cores?  

-- Pasi

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Monday May 05 2008 05:00:07 am Pasi Kärkkäinen wrote:
> Hmm.. have you tried LVM backed devices for HVM guest? Or raw devices..
From the previous post:
> And now for something totally different: I just upgraded my processor from
> an  Intel Core Duo 2300, 1.66Mhz to a Core 2 Duo 5600, 1.83 Mhz.
Here''s some
> new  iometer results:
pattern 4k, 50% read, 0% random

dynamo on?  |   io/s   |  MB/s  | Avg. i/o time(ms} | max i/o time(ms) | %CPU
domu w/gplpv|   501.7  |   1.96 |        2.90       |       0          | 31.68
domu w/qemu |   187.5  |   0.73 |        5.87       |       0          | 29.89
dom0 w/4Gb  |  1102.3  |   4.31 |        0.91       |      445.5       |  0
dom0 w/4Gb  |  1125.8  |   4.40 |        0.89       |      332.1       |  0
(2nd dom0 numbers from when booted w/o /gplpv)

pattern 32k, 50% read, 0% random

domu w/gplpv|   238.3  |   7.45 |        4.09       |        0         | 22.48
domu w/qemu |   157.4  |   4.92 |        6.35       |        0         | 20.51
dom0 w/4Gb  |    52.5  |   1.64 |       19.05       |     1590.0       |  0
dom0 w/4Gb  |    87.8  |   2.74 |       11.39       |     1286.4       |  0
> So, between the two processors, the new one gives qemu and dom0 numbers
that
> are modestly faster, and gplpv numbers that are 50% greater.
I never claimed to have the slickest hardware setup on the block. When I do 
benchmarks, it''s the relative differences I''m stressing, eg -

- qemu vs. gplpv: I obviously expect gplpv to be faster

- one version of gplpv vs. the next: the trend has been that each version of 
gplpv is faster than the previous, especially for iperf, where Realtek gets 
10Mbits/s, 0.8.4 got 19.8 Mb/s, and 0.8.9 is getting 32.1 Mb/s. (But that 
last number is w/ the new processor - it was 25, but that''s still
better.)

- dom0 vs. domu: obviously, the standard to match is dom0 performance. (I 
suspect, tho'', that non-xen kernel performance would be even better.)
Looking
at the 4k pattern numbers above, hvm severely lags dom0. Interestingly 
enough, for the 32k pattern, hvm is doing better than dom0.

That having been said, sure, my hardware setup could be better. My (Fedora) 
xen server''s physical volume spans all of the physical disk, and I have
no
room left on that system for anything but my every day domus, leaving a few 
gig left over for kernel compiles. I currently store my backup & test domus 
on a SuSE system which does have lots of room. Currently, if I want to fire 
one of them up, I access it over samba (ouch!). I eventually plan to convert 
the SuSE box to an iscsi server, serving up lvm slices. As with the processor 
upgrade above, any change in my configuration will be benchmarked as well.

The iscsi conversion is on the back burner for now, tho'', in favor of
flashing
my Fedora bios to support 64 bits, and then loading a 64bit dom0. I suspect 
that will get even better results than iscsi. As always, any significant 
changes will be posted.
> Could you try iometer on dom0 to see what kind of performance you get
> there.. or on linux pv domU?
As you can see above, I did do dom0. I could do a linux pv, but your next idea 
interests me more.
> And one more thing.. was your XP HVM single vcpu or more? Did you try
> binding both dom0 and hvm domU to their own dedicated cpu cores?
It was vcpu=2.

root@Insp6400 05/05/08  6:32PM:~
[977] > xm vcpu-pin Domain-0 0 0
root@Insp6400 05/05/08  6:33PM:~
[978] > xm vcpu-pin Domain-0 1 0
root@Insp6400 05/05/08  6:33PM:~
[979] > xm vcpu-pin fedora 0 0
root@Insp6400 05/05/08  6:33PM:~
[980] > xm vcpu-pin winxp 0 1
root@Insp6400 05/05/08  6:34PM:~
[981] > xm vcpu-pin winxp 1 1
root@Insp6400 05/05/08  6:34PM:~
[982] > xm vcpu-list
Name                                ID  VCPU   CPU State   Time(s) CPU 
Affinity
Domain-0                             0     0     0   r--    5548.2 0
Domain-0                             0     1     0   ---    3392.1 0
fedora                               3     0     0   -b-    1444.0 0
winxp                                2     0     1   r--   14713.3 1
winxp                                2     1     1   ---   15013.8 1

The idle fedora domain shares the same pcpu as dom0. This unfortunately 
results in a very sluggish and useless domu & its desktop. It even took 10 
times as long to reboot. Dom0 seems unaffected. Restoring two pcpus to the 
domain eliminated the sluggishness. Next I tried booting with vcpus=1, with 
it pinned to pcpu 1. Now the new iometer results (qemu (booting w/o /gplpv) 
not tested):

pattern 4k, 50% read, 0% random

dynamo on?  |   io/s   |  MB/s  | Avg. i/o time(ms} | max i/o time(ms) | %CPU
domu w/gplpv|   115.6  |   0.45 |        8.65       |     1836.7       | 67.63
dom0 w/4Gb  |   501.2  |   1.96 |        1.99       |      739.2       |  0
(2nd dom0 numbers from when booted w/o /gplpv)

pattern 32k, 50% read, 0% random

domu w/gplpv|   115.3  |   3.65 |        8.67       |     1735.2       | 54.41
dom0 w/4Gb  |    53.0  |   1.66 |       18.86       |     1751.3       |  0

Yeeaaahh - everything tanked! MB/s down, Cpu % up, etc. Console was still a 
little sluggish. (I suppose pinning cpus might work better with more than one 
socket on the mobo.) I won''t be trying that config again ;-)

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Mon, May 05, 2008 at 08:32:46PM -0400, jim burns wrote:
> - dom0 vs. domu: obviously, the standard to match is dom0 performance. (I 
> suspect, tho'', that non-xen kernel performance would be even
better.) Looking
> at the 4k pattern numbers above, hvm severely lags dom0. Interestingly 
> enough, for the 32k pattern, hvm is doing better than dom0.
>
domU doing better than dom0 usually happens when you use file backed disks on
dom0.. then the memory cache of dom0 will affect the domU results. 
 
> > Could you try iometer on dom0 to see what kind of performance you get
> > there.. or on linux pv domU?
> 
> As you can see above, I did do dom0. I could do a linux pv, but your next
idea
> interests me more.
> 
OK. I think measuring pv domU is worth trying too :)
> > And one more thing.. was your XP HVM single vcpu or more? Did you try
> > binding both dom0 and hvm domU to their own dedicated cpu cores?
> 
> It was vcpu=2.
> 
I think you should re-test with vcpu=1. 

Configure dom0 for 1 vcpu and domU for 1 vcpu and pin the domains to have a 
dedicated core. This way you''re not sharing any pcpu''s between
the domains.
I think this is the "recommended" setup from xen developers for
getting
maximum performance. 

I think the performance will be worse when you have more vcpus in use than
your actual pcpu count.. 
> 
> Yeeaaahh - everything tanked! MB/s down, Cpu % up, etc. Console was still a
> little sluggish. (I suppose pinning cpus might work better with more than
one
> socket on the mobo.) I won''t be trying that config again ;-)
> 
Hmm.. interesting. Maybe it was because of the shared pcpu''s.. 

-- Pasi


_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Tuesday May 06 2008 03:21:29 am Pasi Kärkkäinen
wrote:
> > - dom0 vs. domu: obviously, the standard to match is dom0 performance.
(I
> > suspect, tho'', that non-xen kernel performance would be even
better.)
> > Looking at the 4k pattern numbers above, hvm severely lags dom0.
> > Interestingly enough, for the 32k pattern, hvm is doing better than
dom0.
>
> domU doing better than dom0 usually happens when you use file backed disks
> on dom0.. then the memory cache of dom0 will affect the domU results.
Interesting that that didn''t happen with the 4k pattern numbers,
tho''.
> I think you should re-test with vcpu=1.
>
> Configure dom0 for 1 vcpu and domU for 1 vcpu and pin the domains to have a
> dedicated core. This way you''re not sharing any pcpu''s
between the domains.
> I think this is the "recommended" setup from xen developers for
getting
> maximum performance.
>
> I think the performance will be worse when you have more vcpus in use than
> your actual pcpu count..
Will try that later, after I''ve tested out a new (non-xen) kernel
update.
Having more vcpus than pcpus would be very easy tho'', if you have many 
domains. I can try this with just the hvm domain running.

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Tuesday May 06 2008 03:21:29 am Pasi Kärkkäinen
wrote:
> OK. I think measuring pv domU is worth trying too :)
Ok, let''s try a few things. Repeating my original 0.8.9 numbers, with
the new
processor:

pattern 4k, 50% read, 0% random

dynamo on?  |   io/s   |  MB/s  | Avg. i/o time(ms} | max i/o time(ms) | %CPU
domu w/gplpv|   501.7  |   1.96 |        2.90       |       0          | 31.68
domu w/qemu |   187.5  |   0.73 |        5.87       |       0          | 29.89
dom0 w/4Gb  |  1102.3  |   4.31 |        0.91       |      445.5       |  0
dom0 w/4Gb  |  1125.8  |   4.40 |        0.89       |      332.1       |  0
(2nd dom0 numbers from when booted w/o /gplpv)

pattern 32k, 50% read, 0% random

domu w/gplpv|   238.3  |   7.45 |        4.09       |        0         | 22.48
domu w/qemu |   157.4  |   4.92 |        6.35       |        0         | 20.51
dom0 w/4Gb  |    52.5  |   1.64 |       19.05       |     1590.0       |  0
dom0 w/4Gb  |    87.8  |   2.74 |       11.39       |     1286.4       |  0

Now, that was with all workers running on domu and dom0 simultaneously.
Let''s
try one at a time. On hvm w/gplpv, first the 4k pattern, then later the 32k 
pattern, with dom0 using the ''idle'' task:

 4k pattern |  1026.6  |   4.01 |       39.37       |       0          | 49.70
32k pattern |   311.1  |   9.72 |       45.33       |       0          | 26.21

Now test dom0, with the hvm running the ''idle'' task:

 4k pattern |  1376.7  |   5.38 |        0.73       |      365.7       |  0
32k pattern |   165.9  |   5.19 |        6.02       |      226.6       |  0

As expected, all numbers are significantly faster. Compare this to
''dd''
creating the 4GB /iobw.tst file on dom0 at a 22MB/s rate.

Now, to test a fedora pv, since space is tight on my fedora xen server, I 
just ''xm block-attach''-ed dom0''s /iobw.tst as a new
partition on the domu,
and in the domu, did mkfs, mount, and created a new /iobw.tst on that 
partition. Results:

 4k pattern |  1160.5  |   4.53 |        0.86       |      247.1       |  0
32k pattern |   284.1  |   8.88 |        3.52       |      326.4       |  0

The numbers are very similar to the hvm, including the 32k pattern being 
faster than dom0, which you pointed out is due to caching. This compares 
to ''dd'' creating the 3.7GB iobw.tst on the mounted new
partition at an 18MB/s
rate.
> Configure dom0 for 1 vcpu and domU for 1 vcpu and pin the domains to have a
> dedicated core. This way you''re not sharing any pcpu''s
between the domains.
> I think this is the "recommended" setup from xen developers for
getting
> maximum performance.
>
> I think the performance will be worse when you have more vcpus in use than
> your actual pcpu count..
Now I rebooted dom0, after editing xend-config.sxp to include
''(dom0-cpus 1)'',
and then did the following pins:

[576] > xm create winxp
Using config file "/etc/xen/winxp".
Started domain winxp
root@Insp6400 05/06/08 10:32PM:~
[577] > xm vcpu-pin 0 all 0
root@Insp6400 05/06/08 10:32PM:~
[578] > xm vcpu-pin winxp all 1
root@Insp6400 05/06/08 10:32PM:~
[579] > xm vcpu-list
Name                               ID  VCPU   CPU State   Time(s) CPU Affinity
Domain-0                             0     0     0   r--     228.7 0
Domain-0                             0     1     -   --p      16.0 0
winxp                                5     0     1   r--      36.4 1

Note I also had to set vcpus=1, because with two, I was again getting that 
extremely sluggish response in my hvm.

Going back to simultaneous execution of all workers, to compare against the 
numbers at the top of this post, I got:

pattern 4k, 50% read, 0% random

dynamo on?  |   io/s   |  MB/s  | Avg. i/o time(ms} | max i/o time(ms) | %CPU
domu w/gplpv|   286.4  |   1.12 |        3.49       |      564.9       | 36.97
dom0 w/4Gb  |  1173.9  |   4.59 |        0.85       |      507.3       |  0

pattern 32k, 50% read, 0% random

domu w/gplpv|   217.9  |   6.81 |        4.57       |     1633.5       | 22.93
dom0 w/4Gb  |    63.3  |   1.97 |       15.85       |     1266.5       |  0

which is somewhat slower. Recommendations of the xen developers aside, my 
experience is that allowing xen to schedule any vcpu on any pcpu is most 
efficient.

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Tue, May 6, 2008 at 11:05 PM, jim burns <jim_burn@bellsouth.net>
wrote:
> On Tuesday May 06 2008 03:21:29 am Pasi Kärkkäinen wrote:
>
> > OK. I think measuring pv domU is worth trying too :)
>
>  Ok, let''s try a few things. Repeating my original 0.8.9 numbers,
with the new
>  processor:
>
>
>  pattern 4k, 50% read, 0% random
>
>  dynamo on?  |   io/s   |  MB/s  | Avg. i/o time(ms} | max i/o time(ms) |
%CPU
>  domu w/gplpv|   501.7  |   1.96 |        2.90       |       0          |
31.68
>  domu w/qemu |   187.5  |   0.73 |        5.87       |       0          |
29.89
>  dom0 w/4Gb  |  1102.3  |   4.31 |        0.91       |      445.5       | 
0
>  dom0 w/4Gb  |  1125.8  |   4.40 |        0.89       |      332.1       | 
0
>  (2nd dom0 numbers from when booted w/o /gplpv)
>
>  pattern 32k, 50% read, 0% random
>
>  domu w/gplpv|   238.3  |   7.45 |        4.09       |        0         |
22.48
>  domu w/qemu |   157.4  |   4.92 |        6.35       |        0         |
20.51
>  dom0 w/4Gb  |    52.5  |   1.64 |       19.05       |     1590.0       | 
0
>  dom0 w/4Gb  |    87.8  |   2.74 |       11.39       |     1286.4       | 
0
>
>  Now, that was with all workers running on domu and dom0 simultaneously.
Let''s
>  try one at a time. On hvm w/gplpv, first the 4k pattern, then later the
32k
>  pattern, with dom0 using the ''idle'' task:
>
>   4k pattern |  1026.6  |   4.01 |       39.37       |       0          |
49.70
>  32k pattern |   311.1  |   9.72 |       45.33       |       0          |
26.21
>
>  Now test dom0, with the hvm running the ''idle'' task:
>
>   4k pattern |  1376.7  |   5.38 |        0.73       |      365.7       | 
0
>  32k pattern |   165.9  |   5.19 |        6.02       |      226.6       | 
0
>
>  As expected, all numbers are significantly faster. Compare this to
''dd''
>  creating the 4GB /iobw.tst file on dom0 at a 22MB/s rate.
>
>  Now, to test a fedora pv, since space is tight on my fedora xen server, I
>  just ''xm block-attach''-ed dom0''s /iobw.tst as a
new partition on the domu,
>  and in the domu, did mkfs, mount, and created a new /iobw.tst on that
>  partition. Results:
>
>   4k pattern |  1160.5  |   4.53 |        0.86       |      247.1       | 
0
>  32k pattern |   284.1  |   8.88 |        3.52       |      326.4       | 
0
>
>  The numbers are very similar to the hvm, including the 32k pattern being
>  faster than dom0, which you pointed out is due to caching. This compares
>  to ''dd'' creating the 3.7GB iobw.tst on the mounted new
partition at an 18MB/s
>  rate.
>
>
>  > Configure dom0 for 1 vcpu and domU for 1 vcpu and pin the domains to
have a
>  > dedicated core. This way you''re not sharing any
pcpu''s between the domains.
>  > I think this is the "recommended" setup from xen developers
for getting
>  > maximum performance.
>  >
>  > I think the performance will be worse when you have more vcpus in use
than
>  > your actual pcpu count..
>
>  Now I rebooted dom0, after editing xend-config.sxp to include
''(dom0-cpus 1)'',
>  and then did the following pins:
>
>  [576] > xm create winxp
>  Using config file "/etc/xen/winxp".
>  Started domain winxp
>  root@Insp6400 05/06/08 10:32PM:~
>  [577] > xm vcpu-pin 0 all 0
>  root@Insp6400 05/06/08 10:32PM:~
>  [578] > xm vcpu-pin winxp all 1
>  root@Insp6400 05/06/08 10:32PM:~
>  [579] > xm vcpu-list
>
> Name                               ID  VCPU   CPU State   Time(s) CPU
Affinity
>  Domain-0                             0     0     0   r--     228.7 0
>  Domain-0                             0     1     -   --p      16.0 0
>  winxp                                5     0     1   r--      36.4 1
>
>  Note I also had to set vcpus=1, because with two, I was again getting that
>  extremely sluggish response in my hvm.
>
>  Going back to simultaneous execution of all workers, to compare against
the
>  numbers at the top of this post, I got:
>
>
>  pattern 4k, 50% read, 0% random
>
>  dynamo on?  |   io/s   |  MB/s  | Avg. i/o time(ms} | max i/o time(ms) |
%CPU
>  domu w/gplpv|   286.4  |   1.12 |        3.49       |      564.9       |
36.97
>  dom0 w/4Gb  |  1173.9  |   4.59 |        0.85       |      507.3       | 
0
>
>
>  pattern 32k, 50% read, 0% random
>
>  domu w/gplpv|   217.9  |   6.81 |        4.57       |     1633.5       |
22.93
>  dom0 w/4Gb  |    63.3  |   1.97 |       15.85       |     1266.5       | 
0
>
>  which is somewhat slower. Recommendations of the xen developers aside, my
>  experience is that allowing xen to schedule any vcpu on any pcpu is most
>  efficient.
>
I think that your experience (allowing Xen to do the scheduling itself is most
efficient and only try to tweak the scheduling in very special cases and/or you
really know what you are doing) should be considered conventional wisdom.

Can you refresh me on the recommendations of the Xen developers that you
are referring to?

Thanks,
Todd
>
>
>  _______________________________________________
>  Xen-users mailing list
>  Xen-users@lists.xensource.com
>  http://lists.xensource.com/xen-users
>
_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Tuesday May 06 2008 11:13:36 pm Todd Deshane wrote:
> I think that your experience (allowing Xen to do the scheduling itself is
> most efficient and only try to tweak the scheduling in very special cases
> and/or you really know what you are doing) should be considered
> conventional wisdom.
>
> Can you refresh me on the recommendations of the Xen developers that you
> are referring to?
I was responding to Pasi''s comment:

On Tuesday May 06 2008 03:21:29 am Pasi Kärkkäinen
wrote:
> I think you should re-test with vcpu=1.
>
> Configure dom0 for 1 vcpu and domU for 1 vcpu and pin the domains to have a
> dedicated core. This way you''re not sharing any pcpu''s
between the domains.
> I think this is the "recommended" setup from xen developers for
getting
> maximum performance.
>
> I think the performance will be worse when you have more vcpus in use than
> your actual pcpu count..


_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Tue, May 6, 2008 at 11:53 PM, jim burns <jim_burn@bellsouth.net>
wrote:
> On Tuesday May 06 2008 11:13:36 pm Todd Deshane wrote:
>  > I think that your experience (allowing Xen to do the scheduling
itself is
>  > most efficient and only try to tweak the scheduling in very special
cases
>  > and/or you really know what you are doing) should be considered
>  > conventional wisdom.
>  >
>  > Can you refresh me on the recommendations of the Xen developers that
you
>  > are referring to?
>
>  I was responding to Pasi''s comment:
>
>
>  On Tuesday May 06 2008 03:21:29 am Pasi Kärkkäinen wrote:
>
> > I think you should re-test with vcpu=1.
>  >
>
> > Configure dom0 for 1 vcpu and domU for 1 vcpu and pin the domains to
have a
>  > dedicated core. This way you''re not sharing any
pcpu''s between the domains.
>  > I think this is the "recommended" setup from xen developers
for getting
>  > maximum performance.
>  >
>  > I think the performance will be worse when you have more vcpus in use
than
>  > your actual pcpu count..
>
>
In our Running Xen book [1] chapter 12 "Managing Guest Resources" it
says:

"If you are planning on having a substantial number of guests running, we
recommend sticking with the default VCPUs. The only place that VCPU
pinning may be advantageous is to restrict a CPU to run only for Domain0.
Each guest relies on the services Domain0 offers ..."

It goes on to talk about especially heavy I/O etc. etc.

I bring this up since a lot of thought went into the details of the book. Not
that we will always be right, but continuing to sharpen our knowledge and
working through the details can only help future versions etc.

Cheers,
Todd

[1] http://runningxen.com
>
>
>
> _______________________________________________
>  Xen-users mailing list
>  Xen-users@lists.xensource.com
>  http://lists.xensource.com/xen-users
>
_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Wed, May 07, 2008 at 12:06:41AM -0400, Todd Deshane
wrote:
> On Tue, May 6, 2008 at 11:53 PM, jim burns <jim_burn@bellsouth.net>
wrote:
> > On Tuesday May 06 2008 11:13:36 pm Todd Deshane wrote:
> >  > I think that your experience (allowing Xen to do the scheduling
itself is
> >  > most efficient and only try to tweak the scheduling in very
special cases
> >  > and/or you really know what you are doing) should be considered
> >  > conventional wisdom.
> >  >
> >  > Can you refresh me on the recommendations of the Xen developers
that you
> >  > are referring to?
> >
> >  I was responding to Pasi''s comment:
> >
> >
> >  On Tuesday May 06 2008 03:21:29 am Pasi Kärkkäinen wrote:
> >
> > > I think you should re-test with vcpu=1.
> >  >
> >
And more thing for the number of vcpu''s for the HVM guest.. I think
Windows
installs UNI or SMP HAL during the install time.. I don''t know what
kind of
effect there is if you run SMP HAL with only a single (v)CPU ?
> > > Configure dom0 for 1 vcpu and domU for 1 vcpu and pin the domains
to have a
> >  > dedicated core. This way you''re not sharing any
pcpu''s between the domains.
> >  > I think this is the "recommended" setup from xen
developers for getting
> >  > maximum performance.
> >  >
> >  > I think the performance will be worse when you have more vcpus
in use than
> >  > your actual pcpu count..
> >
> >
> In our Running Xen book [1] chapter 12 "Managing Guest Resources"
it says:
> 
> "If you are planning on having a substantial number of guests running,
we
> recommend sticking with the default VCPUs. The only place that VCPU
> pinning may be advantageous is to restrict a CPU to run only for Domain0.
> Each guest relies on the services Domain0 offers ..."
> 
> It goes on to talk about especially heavy I/O etc. etc.
> 
> I bring this up since a lot of thought went into the details of the book.
Not
> that we will always be right, but continuing to sharpen our knowledge and
> working through the details can only help future versions etc.
> 
Yep. What I meant was if dom0 can''t get enough CPU time it will impact
all
vm''s.. so at least in some cases it will help to dedicate a pcpu for
dom0.

-- Pasi

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

> 
> And more thing for the number of vcpu''s for the HVM guest.. I
think
> Windows
> installs UNI or SMP HAL during the install time.. I don''t know
what kind
> of
> effect there is if you run SMP HAL with only a single (v)CPU ?
> 
The first time you start a previously UP windows DomU with more than a single
vcpu, you get a message like ''your system must be rebooted for your new
hardware to complete installation'', and sure enough you don''t
see the additional CPU(s) in process monitor until you reboot.

Subsequent changes to and from SMP don''t do the same thing, so it is
possible that the first time you boot with >1 cpu an SMP hal gets installed,
and that it doesn''t get uninstalled afterwards... In a physical as
opposed to virtual environment this kind of makes sense - how often are you
going to take an SMP machine back to a single processor unless you have a
hardware failure or something, which will probably only be temporary anyway
until the replacement cpu arrives?

James

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Wed, May 07, 2008 at 04:58:43PM +1000, James Harper
wrote:
> > 
> > And more thing for the number of vcpu''s for the HVM guest.. I
think
> > Windows
> > installs UNI or SMP HAL during the install time.. I don''t
know what kind
> > of
> > effect there is if you run SMP HAL with only a single (v)CPU ?
> > 
> 
> The first time you start a previously UP windows DomU with more than a
single vcpu,
> you get a message like ''your system must be rebooted for your new
hardware to
> complete installation'', and sure enough you don''t see the
additional CPU(s) in
> process monitor until you reboot.
> 
> Subsequent changes to and from SMP don''t do the same thing, so it
is possible
> that the first time you boot with >1 cpu an SMP hal gets installed, and
that it
> doesn''t get uninstalled afterwards... In a physical as opposed to
virtual environment
> this kind of makes sense - how often are you going to take an SMP machine
back to
> a single processor unless you have a hardware failure or something, which
will
> probably only be temporary anyway until the replacement cpu arrives?
> 
Yep.

Do you know if running the SMP HAL with just a single vcpu will cause any
performance degrations? 

-- Pasi

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

> 
> Yep.
> 
> Do you know if running the SMP HAL with just a single vcpu will cause any
> performance degrations?
> 
Presumably it would... spinlock''s effectively become no-op''s
on UP.

James

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

James,

I''ve recently got the time to use the 0.8.9 drivers on a fresh and
completely upgraded (SP3)
Version of XP Pro 32bit (german) and on a W2K3 SBS SP2 32bit (german) with
former 0.7.something PV
drivers. Host System is 3.2.0 64bit.

I didn''t find any problems even on heavy disk and net IO. I''ll
report ioperf results in the next
few days.

One thing that isn''t clear to me are the domU vif configuration. For
former driver versions, I
used type=netfront for the PV drivers, the recent drivers seems to work only
with the type=ioemu.
I''m not sure how to configure the vifs.

For blockdevices I continue using xvdN.

Thanks for this release!

Cheers,

Stephan






James Harper schrieb:
>> James,
>>
>> Nope, there are at least 4 firewalls that will stop you..
> 
> :)
> 
>> I can however  do some debugging if you like?
>>
> 
> I''ll send you a list of the stuff I want to know.
> 
> Thanks
> 
> James
> 
> _______________________________________________
> Xen-users mailing list
> Xen-users@lists.xensource.com
> http://lists.xensource.com/xen-users

-- 
Stephan Seitz
Senior System Administrator

*netz-haut* e.K.
multimediale kommunikation

zweierweg 22
97074 würzburg

fon: +49 931 2876247
fax: +49 931 2876248

web: www.netz-haut.de <http://www.netz-haut.de/>

registriergericht: amtsgericht würzburg, hra 5054



_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

On Saturday May 10 2008 07:43:08 am Stephan Seitz wrote:
> One thing that isn''t clear to me are the domU vif configuration.
For former
> driver versions, I used type=netfront for the PV drivers, the recent
> drivers seems to work only with the type=ioemu. I''m not sure how
to
> configure the vifs.
A previous suggestion on this list said to change the mac address when you 
change between netfront and ioemu. Does that help?

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

Scott/James,


Scott McKenzie-3 wrote:
> 
> Scott McKenzie wrote:
> 
> I''ve just installed openSUSE on my system to test the drivers with
their
> kernel and Xen version.  I took a copy of my Windows HVM, booted it, 
> installed 0.8.9, rebooted without /gplpv, rebooted with /gplpv and I had 
> two disk devices appearing in device manager.  So it doesn''t look
like
> it''s the dom0 kernel that''s causing this problem.
> 
I''ve just done an upgrade to rhel5.2 beta, and supposable allot of
kernel
2.6.24/xen 3.1.2 stuff has been backported..

However I''m still have the same issue.

James, can I ask what platform (versions) you are using to test these gplpv
drivers for windows?

Cheers,

James
-- 
View this message in context:
http://www.nabble.com/Release-0.8.9-of-GPL-PV-drivers-for-Windows-tp16910023p17224706.html
Sent from the Xen - User mailing list archive at Nabble.com.


_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

> 
> James, can I ask what platform (versions) you are using to test these
> gplpv drivers for windows?
Debian Etch, with the versions of xen from backports (3.1 and just
recently 3.2).

James

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

James,

Do you have any other recommendations
on what to try to get this issue fixed.

I suppose I could try the ubuntu server
installation, to see ifs something related to the xen dom0 software versioning?

This would issolate possible hardware
issue.

What do you think?

Cheers,

James Shirley

=======================================================================
Electricity Networks Corporation, trading as Western Power

ABN: 18 540 492 861

TO THE ADDRESSEE - this email is for the intended addressee only and may contain information that is confidential. 

If you have received this email in error, please notify us immediately by return email or by telephone. 

Please also destroy this message and any electronic or hard copies of this message.

Any claim to confidentiality is not waived or lost by reason of mistaken transmission of this email.

Unencrypted email is not secure and may not be authentic.  Western Power cannot guarantee the accuracy, reliability,

completeness or confidentiality of this email and any attachments.

VIRUSES - Western Power scans all outgoing emails and attachments for viruses, however it is the recipient's responsibility 

to ensure this email is free of viruses.

=======================================================================

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

> 
> Do you have any other recommendations on what to try to get this issue
> fixed.
> 
> I suppose I could try the ubuntu server installation, to see ifs
something
> related to the xen dom0 software versioning?
> 
> This would issolate possible hardware issue.
> 
> What do you think?
> 
If you can wait a bit the next release should resolve these problems.

Thanks

James

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users

All right, I just got finished upgrading my i386 Fedora 8 system to x86_64, 
thanx to the new Core 2 duo processor. (Actually, it was a ''new
install'',
since apparently there is no clean upgrade path from one arch to another.) 
After restoring my old configuration, I was pleased to see that my system 
behaved exactly the same as before, no extra quirks. If it wasn''t for
''yum
update''s offering both arches, I wouldn''t be able to tell the
difference,
tho'' I haven''t explored multimedia much, yet. Even my kernel
compiles are
simpler, not having to explicitly specify ''rpmbuild
--target=i686'', since
there are no subarches to worry about.

So let''s see whether it''s any faster. Note - I''m only
testing the same 0.8.9
gplpv version, just before and after the processor & software upgrade. 
Current configuration:

Equipment: core 2 duo 5600, 1.83ghz each, 2M, sata drive configured for
UDMA/100 System: fc8 64bit, xen 3.1.2, xen.gz 3.1.3, dom0 2.6.21
Tested hvm: XP Pro SP2, 2002 32bit  w/512M, file backed vbd on local disk, 
tested w/ iometer 2006-07-27 (1Gb \iobw.tst, 5min run) & iperf 2.0.2 (1 min 
run)

Note, I''m no longer using iperf 1.7.0, since I discovered that iperf
2.0.2
comes with Fedora 8. First the old iometer numbers, from the old 32bit 
processor, both domu & dom0 threads running at the same time:

pattern 4k, 50% read, 0% random

dynamo on?  |   io/s   |  MB/s  | Avg. i/o time(ms} | max i/o time(ms) | %CPU
domu w/gplpv|   331.5  |   1.29 |      232.29       |       0          | 35.63
domu w/qemu |   166.1  |   0.65 |        9.67       |       0          | 35.09
dom0 w/4Gb  |  1088.3  |   4.25 |        0.92       |      487.4       |  0
dom0 w/4Gb  |  1118.0  |   4.37 |        0.89       |      181.3       |  0
(2nd dom0 numbers from when booted w/o /gplpv)

pattern 32k, 50% read, 0% random

domu w/gplpv|   166.0  |   5.19 |        7.98       |        0         | 29.85
domu w/qemu |   100.4  |   3.14 |       21.09       |        0         | 35.93
dom0 w/4Gb  |    61.8  |   1.93 |       16.14       |     1492.3       |  0
dom0 w/4Gb  |   104.9  |   3.28 |        9.54       |      906.6       |  0

And now the new numbers:

pattern 4k, 50% read, 0% random

dynamo on?  |   io/s   |  MB/s  | Avg. i/o time(ms} | max i/o time(ms) | %CPU
domu w/gplpv|   417.5  |   1.63 |        7.39       |       0          | 27.29
domu w/qemu |   155.4  |   0.60 |       -4.60       |       0          | 29.23
dom0 w/2Gb  |   891.6  |   3.48 |        1.12       |      574.4       |  0
dom0 w/2Gb  |  1033.1  |   4.04 |        0.97       |      242.4       |  0
(2nd dom0 numbers from when booted w/o /gplpv)

pattern 32k, 50% read, 0% random

domu w/gplpv|   228.6  |   7.15 |       -4.65       |        0         | 21.64
domu w/qemu |   120.4  |   3.76 |       83.63       |        0         | 28.50
dom0 w/2Gb  |    42.0  |   1.31 |       23.80       |     2084.7       |  0
dom0 w/2Gb  |    88.3  |   2.76 |       11.32       |     1267.3       |  0

There are significant improvements in gplpv io/s. MB/s, avg. i/o time. 
and %cpu. There are modest decreases in dom0 performance, and modest 
improvements in qemu.

Now running one domain thread at a time, with any other domains running 
the ''idle'' task. First the old numbers (with the new
processor, but 32bit
dom0):

gplpv 0.8.9:
 4k pattern |  1026.6  |   4.01 |       39.37       |       0          | 49.70
32k pattern |   311.1  |   9.72 |       45.33       |       0          | 26.21

dom0:
 4k pattern |  1376.7  |   5.38 |        0.73       |      365.7       |  0
32k pattern |   165.9  |   5.19 |        6.02       |      226.6       |  0

and now the new:

gplpv 0.8.9:
 4k pattern |  1170.0  |   4.57 |        7.16       |       0          | 41.34
32k pattern |   287.0  |   8.97 |      -30.85       |       0          | 23.39

dom0:
 4k pattern |  1376.7  |   5.38 |        0.73       |      365.7       |  0
32k pattern |  1484.3  |   5.80 |        0.67       |      314.4       |  0

The differences are insignificant for single thread execution. Since the 
underlying disk has not changed, just the processor and software, this is not 
unexpected. However, it was nice to see multi-thread performance improve 
(which is more software dependent), even if it was just on gplpv.

As far as ''iperf -c dom0-name -t 60'' goes, the old numbers
(for 1.7.0) are:
realtek: 10Mb/s
gplpv (old processor): 25 Mb/s
gplpv (new processor): 32 Mb/s

and the new numbers (for 2.0.2) are:
realtek: .5 Mb/s
gplpv (new processor): 4Mb/s

Huh?!? Ok, let''s try iperf 1.7.0 again:
realtek: 9.1 Mb/s
gplpv (new processor): 33.6 Mb/s

That''s interesting - guess I''ll be sticking with 1.7.0 after
all! (Btw, by
adding the -r option, I get nearly identical write speeds for dom0 to gplpv 
domu, but 2-6x faster for qemu.)

I''ll look at 0.9.0 later, and if there are significant differences from
0.8.9,
I''ll report to the list.

_______________________________________________
Xen-users mailing list
Xen-users@lists.xensource.com
http://lists.xensource.com/xen-users