Install paravirtualized drivers after booting for high-performance IO

Install paravirtualized drivers after booting for high-performance IO

[Arlen.D.Wang]

Hi!

    What is "high-performance IO" ?
    What meaning for "Install paravirtualized drivers after booting for 
high-performance IO"?

    Thanks,

-- 
DengYu  Wang

Star SoftComm(China) Ltd
www.starsoftcomm.com

Phone: +86-10-82893388 Ext 825
Mobile: +86-13520799946 Mail: dengyu@starsoftcomm.com
Address: 3rd Floor Xinxi Road,Shangdi,Haidian District, Beijing 100085,PRC 

Re: Install paravirtualized drivers after booting for high-performance IO

[Mark Williamson]

Hi there,

I expect you've seen a reference to the Windows support using Vanderpool 
Technology?

The current VT implementation provides an emulation of real hardware devices 
(e.g. an NE2000 ethernet card) to the guest OS.  This has the advantage that 
you can boot an unmodified OS and it'll recognise the emulated devices and 
Just Work.  This will allow you to install the OS using its usual installer 
and get things up and running.

However, you pay a performance price for emulating a device like this.  The 
most straightforward way to solve this is to replace the emulated device 
drivers with Xen-aware drivers (like the ones used by the Xen-native Linux 
port).  Once you've installed the OS, you can upgrade to the Xen device 
drivers.

This should give better bandwidth for things like disk and network IO.

Does that help?

Cheers,
Mark

>     What is "high-performance IO" ?
>     What meaning for "Install paravirtualized drivers after booting for
> high-performance IO"?
>
>     Thanks,

Re: Install paravirtualized drivers after booting forhigh-performance IO

[Arlen.D.Wang]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1430 bytes --]

Yes, Thank you .

Where is get  a reference to the Windows support using Vanderpool Technology 
?
Who is doing that Windows support using Vanderpool Technology ?

If Xen can run unmodified Linux , then Xen can run unmodified Windows also , 
right ?

DengYu  Wang

"Mark Williamson" <mark.williamson@cl.cam.ac.uk> дÈëÏûÏ¢ÐÂÎÅ:200505091257.04034.mark.williamson@cl.cam.ac.uk...
> Hi there,
>
> I expect you've seen a reference to the Windows support using Vanderpool
> Technology?
>
> The current VT implementation provides an emulation of real hardware 
> devices
> (e.g. an NE2000 ethernet card) to the guest OS.  This has the advantage 
> that
> you can boot an unmodified OS and it'll recognise the emulated devices and
> Just Work.  This will allow you to install the OS using its usual 
> installer
> and get things up and running.
>
> However, you pay a performance price for emulating a device like this. 
> The
> most straightforward way to solve this is to replace the emulated device
> drivers with Xen-aware drivers (like the ones used by the Xen-native Linux
> port).  Once you've installed the OS, you can upgrade to the Xen device
> drivers.
>
> This should give better bandwidth for things like disk and network IO.
>
> Does that help?
>
> Cheers,
> Mark
>
>>     What is "high-performance IO" ?
>>     What meaning for "Install paravirtualized drivers after booting for
>> high-performance IO"?
>>
>>     Thanks, 

Re: Re: Install paravirtualized drivers after booting forhigh-performance IO

[Mark Williamson]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain; charset="utf-8", Size: 2265 bytes --]

> Where is get  a reference to the Windows support using Vanderpool
> Technology ?

The code in the unstable tree is the best reference (if you feel brave).  
There are the slides for a couple of talks about VT at the Xen Summit webpage 
(http://summit.xensource.com/presentations/).

> Who is doing that Windows support using Vanderpool Technology ?

This implementation work has been spearheaded by Intel, although I think there 
are also some people at IBM also working on Xen/VT (is that right?)

> If Xen can run unmodified Linux , then Xen can run unmodified Windows also
> , right ?

Not yet: running Windows requires emulating extra 16 bit x86 stuff that Linux 
doesn't need in order to run.  There is a design for supporting this and it 
might even be under development right now.

HTH,
Mark

> DengYu  Wang
>
> "Mark Williamson" <mark.williamson@cl.cam.ac.uk>
> дÈëÏûÏ¢ÐÂÎÅ:200505091257.04034.mark.williamson@cl.cam.ac.uk...
>
> > Hi there,
> >
> > I expect you've seen a reference to the Windows support using Vanderpool
> > Technology?
> >
> > The current VT implementation provides an emulation of real hardware
> > devices
> > (e.g. an NE2000 ethernet card) to the guest OS.  This has the advantage
> > that
> > you can boot an unmodified OS and it'll recognise the emulated devices
> > and Just Work.  This will allow you to install the OS using its usual
> > installer
> > and get things up and running.
> >
> > However, you pay a performance price for emulating a device like this.
> > The
> > most straightforward way to solve this is to replace the emulated device
> > drivers with Xen-aware drivers (like the ones used by the Xen-native
> > Linux port).  Once you've installed the OS, you can upgrade to the Xen
> > device drivers.
> >
> > This should give better bandwidth for things like disk and network IO.
> >
> > Does that help?
> >
> > Cheers,
> > Mark
> >
> >>     What is "high-performance IO" ?
> >>     What meaning for "Install paravirtualized drivers after booting for
> >> high-performance IO"?
> >>
> >>     Thanks,
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel

Re: Re: Install paravirtualized drivers after bootingforhigh-performance IO

[Arlen.D.Wang]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 2271 bytes --]

What is the difference between paravirtualized and virtualized?


DengYu  Wang



> Where is get  a reference to the Windows support using Vanderpool
> Technology ?

The code in the unstable tree is the best reference (if you feel brave).
There are the slides for a couple of talks about VT at the Xen Summit 
webpage
(http://summit.xensource.com/presentations/).

> Who is doing that Windows support using Vanderpool Technology ?

This implementation work has been spearheaded by Intel, although I think 
there
are also some people at IBM also working on Xen/VT (is that right?)

> If Xen can run unmodified Linux , then Xen can run unmodified Windows also
> , right ?

Not yet: running Windows requires emulating extra 16 bit x86 stuff that 
Linux
doesn't need in order to run.  There is a design for supporting this and it
might even be under development right now.

HTH,
Mark

> DengYu  Wang
>
> "Mark Williamson" <mark.williamson@cl.cam.ac.uk>
> §Õ?:200505091257.04034.mark.williamson@cl.cam.ac.uk...
>
> > Hi there,
> >
> > I expect you've seen a reference to the Windows support using Vanderpool
> > Technology?
> >
> > The current VT implementation provides an emulation of real hardware
> > devices
> > (e.g. an NE2000 ethernet card) to the guest OS.  This has the advantage
> > that
> > you can boot an unmodified OS and it'll recognise the emulated devices
> > and Just Work.  This will allow you to install the OS using its usual
> > installer
> > and get things up and running.
> >
> > However, you pay a performance price for emulating a device like this.
> > The
> > most straightforward way to solve this is to replace the emulated device
> > drivers with Xen-aware drivers (like the ones used by the Xen-native
> > Linux port).  Once you've installed the OS, you can upgrade to the Xen
> > device drivers.
> >
> > This should give better bandwidth for things like disk and network IO.
> >
> > Does that help?
> >
> > Cheers,
> > Mark
> >
> >>     What is "high-performance IO" ?
> >>     What meaning for "Install paravirtualized drivers after booting for
> >> high-performance IO"?
> >>
> >>     Thanks,
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel 

Re: Re: Re: Install paravirtualized drivers after bootingforhigh-performance IO

[Miao Feng]

Hi:
    As i recognized, paravirtulization is an implemention of high
performence virtulization in non-VT box, pure virtulization needs no
modification on guest os, but paravirtulization do needs. Is that right?
    If so, guest os on VT cpu will provide no paravirtulization?

    Any suggestions?

Miao
StarSoftComm.com


Arlen.D.Wang wrote:
>What is the difference between paravirtualized and virtualized?
>
>
>DengYu  Wang
>
>
>
>  
>>Where is get  a reference to the Windows support using Vanderpool
>>Technology ?
>>    
>
>The code in the unstable tree is the best reference (if you feel brave).
>There are the slides for a couple of talks about VT at the Xen Summit 
>webpage
>(http://summit.xensource.com/presentations/).
>
>  
>>Who is doing that Windows support using Vanderpool Technology ?
>>    
>
>This implementation work has been spearheaded by Intel, although I think 
>there
>are also some people at IBM also working on Xen/VT (is that right?)
>
>  
>>If Xen can run unmodified Linux , then Xen can run unmodified Windows also
>>, right ?
>>    
>
>Not yet: running Windows requires emulating extra 16 bit x86 stuff that 
>Linux
>doesn't need in order to run.  There is a design for supporting this and it
>might even be under development right now.
>
>HTH,
>Mark
>
>  
>>DengYu  Wang
>>
>>"Mark Williamson" <mark.williamson@cl.cam.ac.uk>
>>§Õ?:200505091257.04034.mark.williamson@cl.cam.ac.uk...
>>
>>    
>>>Hi there,
>>>
>>>I expect you've seen a reference to the Windows support using Vanderpool
>>>Technology?
>>>
>>>The current VT implementation provides an emulation of real hardware
>>>devices
>>>(e.g. an NE2000 ethernet card) to the guest OS.  This has the advantage
>>>that
>>>you can boot an unmodified OS and it'll recognise the emulated devices
>>>and Just Work.  This will allow you to install the OS using its usual
>>>installer
>>>and get things up and running.
>>>
>>>However, you pay a performance price for emulating a device like this.
>>>The
>>>most straightforward way to solve this is to replace the emulated device
>>>drivers with Xen-aware drivers (like the ones used by the Xen-native
>>>Linux port).  Once you've installed the OS, you can upgrade to the Xen
>>>device drivers.
>>>
>>>This should give better bandwidth for things like disk and network IO.
>>>
>>>Does that help?
>>>
>>>Cheers,
>>>Mark
>>>
>>>      
>>>>    What is "high-performance IO" ?
>>>>    What meaning for "Install paravirtualized drivers after booting for
>>>>high-performance IO"?
>>>>
>>>>    Thanks,
>>>>        
>>_______________________________________________
>>Xen-devel mailing list
>>Xen-devel@lists.xensource.com
>>http://lists.xensource.com/xen-devel 
>>    
>
>
>
>
>_______________________________________________
>Xen-devel mailing list
>Xen-devel@lists.xensource.com
>http://lists.xensource.com/xen-devel
>
>
>  

Re: Re: Re: Install paravirtualized drivers after bootingforhigh-performance IO

[Mark Williamson]

>     As i recognized, paravirtulization is an implemention of high
> performence virtulization in non-VT box, pure virtulization needs no
> modification on guest os, but paravirtulization do needs. Is that right?

Yes, that's all correct.

>     If so, guest os on VT cpu will provide no paravirtulization?

A fully virtualised environment will be available for guests on a VT CPU, so 
they won't *need* any modifications in order to run.  However, for good IO 
performance, you're likely to want to install Xen-aware device drivers which 
will be compatible with the existing frontend/backend scheme.

Other fully virtualised environments (e.g. VMWare) also sometimes provide 
higher-performance drivers that are aware of the environment.  In Xen's case 
we have Xen-aware drivers already but they need to be adapted to run in a VT 
virtual machine

HTH,
Mark

>     Any suggestions?
>
> Miao
> StarSoftComm.com
>
> Arlen.D.Wang wrote:
> >What is the difference between paravirtualized and virtualized?
> >
> >
> >DengYu  Wang
> >
> >>Where is get  a reference to the Windows support using Vanderpool
> >>Technology ?
> >
> >The code in the unstable tree is the best reference (if you feel brave).
> >There are the slides for a couple of talks about VT at the Xen Summit
> >webpage
> >(http://summit.xensource.com/presentations/).
> >
> >>Who is doing that Windows support using Vanderpool Technology ?
> >
> >This implementation work has been spearheaded by Intel, although I think
> >there
> >are also some people at IBM also working on Xen/VT (is that right?)
> >
> >>If Xen can run unmodified Linux , then Xen can run unmodified Windows
> >> also , right ?
> >
> >Not yet: running Windows requires emulating extra 16 bit x86 stuff that
> >Linux
> >doesn't need in order to run.  There is a design for supporting this and
> > it might even be under development right now.
> >
> >HTH,
> >Mark
> >
> >>DengYu  Wang
> >>
> >>"Mark Williamson" <mark.williamson@cl.cam.ac.uk>
> >>§Õ?:200505091257.04034.mark.williamson@cl.cam.ac.uk...
> >>
> >>>Hi there,
> >>>
> >>>I expect you've seen a reference to the Windows support using Vanderpool
> >>>Technology?
> >>>
> >>>The current VT implementation provides an emulation of real hardware
> >>>devices
> >>>(e.g. an NE2000 ethernet card) to the guest OS.  This has the advantage
> >>>that
> >>>you can boot an unmodified OS and it'll recognise the emulated devices
> >>>and Just Work.  This will allow you to install the OS using its usual
> >>>installer
> >>>and get things up and running.
> >>>
> >>>However, you pay a performance price for emulating a device like this.
> >>>The
> >>>most straightforward way to solve this is to replace the emulated device
> >>>drivers with Xen-aware drivers (like the ones used by the Xen-native
> >>>Linux port).  Once you've installed the OS, you can upgrade to the Xen
> >>>device drivers.
> >>>
> >>>This should give better bandwidth for things like disk and network IO.
> >>>
> >>>Does that help?
> >>>
> >>>Cheers,
> >>>Mark
> >>>
> >>>>    What is "high-performance IO" ?
> >>>>    What meaning for "Install paravirtualized drivers after booting for
> >>>>high-performance IO"?
> >>>>
> >>>>    Thanks,
> >>
> >>_______________________________________________
> >>Xen-devel mailing list
> >>Xen-devel@lists.xensource.com
> >>http://lists.xensource.com/xen-devel
> >
> >_______________________________________________
> >Xen-devel mailing list
> >Xen-devel@lists.xensource.com
> >http://lists.xensource.com/xen-devel
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel

Re: Re: Install paravirtualized drivers after bootingforhigh-performance IO

[miaofeng]

[-- Attachment #1: Type: text/plain, Size: 1628 bytes --]

Thanx alot, Mark.

I am wondering if device management, like vga, would share the same architecture as block device?
if that is the case, would vga be accessed by guest os directly?
if that is not the case, shall we have to base it on X11 system?

if all above is wrong, will u please give me some outline of the device achitecture?

Miao

StarSoftComm.com


----- Original Message ----- 
From: "Mark Williamson" <mark.williamson@cl.cam.ac.uk>
To: <xen-devel@lists.xensource.com>
Cc: "Miao Feng" <miaofeng@starsoftcomm.com>
Sent: Friday, May 13, 2005 11:33 PM
Subject: Re: [Xen-devel] Re: Re: Install paravirtualized drivers after bootingforhigh-performance IO


>     As i recognized, paravirtulization is an implemention of high
> performence virtulization in non-VT box, pure virtulization needs no
> modification on guest os, but paravirtulization do needs. Is that right?

Yes, that's all correct.

>     If so, guest os on VT cpu will provide no paravirtulization?

A fully virtualised environment will be available for guests on a VT CPU, so 
they won't *need* any modifications in order to run.  However, for good IO 
performance, you're likely to want to install Xen-aware device drivers which 
will be compatible with the existing frontend/backend scheme.

Other fully virtualised environments (e.g. VMWare) also sometimes provide 
higher-performance drivers that are aware of the environment.  In Xen's case 
we have Xen-aware drivers already but they need to be adapted to run in a VT 
virtual machine

HTH,
Mark

>     Any suggestions?
>
> Miao
> StarSoftComm.com
>


[-- Attachment #2: Type: text/plain, Size: 138 bytes --]

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

Re: Re: Install paravirtualized drivers after bootingforhigh-performance IO

[Mark Williamson]

> I am wondering if device management, like vga, would share the same
> architecture as block device? if that is the case, would vga be accessed by
> guest os directly?
> if that is not the case, shall we have to base it on X11 system?

Right now, a VGA video card is emulated by the ioemu device model in dom0, 
which displays the current screen state in an X11 window.  The guest is not 
aware of this.

It may well be that initially the block and net drivers are made Xen-aware but 
the VGA device continues to be an emulated PCI device.

At some stage in the future, there may be a "virtual framebuffer" device that 
will be Xen-aware and use shared memory like the block and net drivers.  This 
may give better performance and thus be preferable to the fully emulated 
graphics device for desktop systems.  This is likely to be some time after 
the Xen-aware block / net devices.

Even further into the future, we might see a high performance Xen virtual 3D 
device.  This is some way off, though.

HTH,
Mark

> if all above is wrong, will u please give me some outline of the device
> achitecture?
>
> Miao
>
> StarSoftComm.com
>
>
> ----- Original Message -----
> From: "Mark Williamson" <mark.williamson@cl.cam.ac.uk>
> To: <xen-devel@lists.xensource.com>
> Cc: "Miao Feng" <miaofeng@starsoftcomm.com>
> Sent: Friday, May 13, 2005 11:33 PM
> Subject: Re: [Xen-devel] Re: Re: Install paravirtualized drivers after
> bootingforhigh-performance IO
>
> >     As i recognized, paravirtulization is an implemention of high
> > performence virtulization in non-VT box, pure virtulization needs no
> > modification on guest os, but paravirtulization do needs. Is that right?
>
> Yes, that's all correct.
>
> >     If so, guest os on VT cpu will provide no paravirtulization?
>
> A fully virtualised environment will be available for guests on a VT CPU,
> so they won't *need* any modifications in order to run.  However, for good
> IO performance, you're likely to want to install Xen-aware device drivers
> which will be compatible with the existing frontend/backend scheme.
>
> Other fully virtualised environments (e.g. VMWare) also sometimes provide
> higher-performance drivers that are aware of the environment.  In Xen's
> case we have Xen-aware drivers already but they need to be adapted to run
> in a VT virtual machine
>
> HTH,
> Mark
>
> >     Any suggestions?
> >
> > Miao
> > StarSoftComm.com

Re: Re: Install paravirtualized drivers after booting forhigh-performance IO

[Mark Williamson]

> What is the difference between paravirtualized and virtualized?

Paravirtualisation is providing virtual machines to operating systems that 
have been ported to run on Xen.

Full virtualisation is providing virtual machines to unmodified operating 
systems.

A compromise for when you can't properly modify an OS is to full virtualise 
the CPU using VT but install device drivers that are Xen-aware.

Cheers,
Mark

>
>
> ----- Original Message -----
> From: "Mark Williamson" <mark.williamson@cl.cam.ac.uk>
> To: <xen-devel@lists.xensource.com>
> Cc: "Arlen.D.Wang" <dengyu@starsoftcomm.com>
> Sent: Tuesday, May 10, 2005 10:24 AM
> Subject: Re: [Xen-devel] Re: Install paravirtualized drivers after booting
> forhigh-performance IO
>
> > Where is get  a reference to the Windows support using Vanderpool
> > Technology ?
>
> The code in the unstable tree is the best reference (if you feel brave).
> There are the slides for a couple of talks about VT at the Xen Summit
> webpage (http://summit.xensource.com/presentations/).
>
> > Who is doing that Windows support using Vanderpool Technology ?
>
> This implementation work has been spearheaded by Intel, although I think
> there are also some people at IBM also working on Xen/VT (is that right?)
>
> > If Xen can run unmodified Linux , then Xen can run unmodified Windows
> > also , right ?
>
> Not yet: running Windows requires emulating extra 16 bit x86 stuff that
> Linux doesn't need in order to run.  There is a design for supporting this
> and it might even be under development right now.
>
> HTH,
> Mark
>
> > DengYu  Wang
> >
> > "Mark Williamson" <mark.williamson@cl.cam.ac.uk>
> > дϢ:200505091257.04034.mark.williamson@cl.cam.ac.uk...
> >
> > > Hi there,
> > >
> > > I expect you've seen a reference to the Windows support using
> > > Vanderpool Technology?
> > >
> > > The current VT implementation provides an emulation of real hardware
> > > devices
> > > (e.g. an NE2000 ethernet card) to the guest OS.  This has the advantage
> > > that
> > > you can boot an unmodified OS and it'll recognise the emulated devices
> > > and Just Work.  This will allow you to install the OS using its usual
> > > installer
> > > and get things up and running.
> > >
> > > However, you pay a performance price for emulating a device like this.
> > > The
> > > most straightforward way to solve this is to replace the emulated
> > > device drivers with Xen-aware drivers (like the ones used by the
> > > Xen-native Linux port).  Once you've installed the OS, you can upgrade
> > > to the Xen device drivers.
> > >
> > > This should give better bandwidth for things like disk and network IO.
> > >
> > > Does that help?
> > >
> > > Cheers,
> > > Mark
> > >
> > >>     What is "high-performance IO" ?
> > >>     What meaning for "Install paravirtualized drivers after booting
> > >> for high-performance IO"?
> > >>
> > >>     Thanks,
> >
> > _______________________________________________
> > Xen-devel mailing list
> > Xen-devel@lists.xensource.com
> > http://lists.xensource.com/xen-devel