[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [Xen-users] VT/Pacifica - Paravirtualization


  • To: "MaTT" <mrbarletta@xxxxxxxxxxxx>, xen-users@xxxxxxxxxxxxxxxxxxx
  • From: "Petersson, Mats" <Mats.Petersson@xxxxxxx>
  • Date: Mon, 4 Sep 2006 13:27:07 +0200
  • Delivery-date: Mon, 04 Sep 2006 04:28:51 -0700
  • List-id: Xen user discussion <xen-users.lists.xensource.com>
  • Thread-index: AcbQBm2e+xmre46qQnu1m+hZGfonqAADU4nA
  • Thread-topic: [Xen-users] VT/Pacifica - Paravirtualization

 

> -----Original Message-----
> From: xen-users-bounces@xxxxxxxxxxxxxxxxxxx 
> [mailto:xen-users-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of MaTT
> Sent: 04 September 2006 10:48
> To: xen-users@xxxxxxxxxxxxxxxxxxx
> Subject: [Xen-users] VT/Pacifica - Paravirtualization
> 
> Hei all!.
> 
> I was reading about paravirtualization outperforming hardware enabled 
> virtualization, and virtual iron's posts about 
> paravirtualization as a 
> dead end.

The big problem with para-virtualization is that the OS source code
needs to be modified. This is both time-consuming and "delaying" (there
will always be a lag behind the official release, unless the OS has
para-virtualization support from the original source code). Since each
new release of the base-OS needs an update of the para-virt work, it
will require work every time a new release is made - even if the release
is only a minor upgrade some work is probably necessary.

> 
> Regarding performance and reliability, is paravirtualization still 
> better solutions than HW assisted virtualization?

Reliability in this case is very much on the hypervisor components, and
there shouldn't be much difference, except perhaps the maturity of the
code [para-virt being more mature - but it's countered by the full
virtualization solution having slightly less code to deal with...]

> 
>  Would HW reach a point where paravirtualization won't longer 
> be necessary?

Possibly, but almost certainly it would require a new set of drivers,
and it would take some time before those are available for the more
common OS's, and the less common (legacy) OS's will probably never have
them (but of coruse, they will not have Paravirtualization either,
unless they are Open Source and someone decides that they want to put
the effort in to redo the OS for Para-virt usage). 

There is work already in progress to create para-virtual drivers for Xen
+ Linux, so that a unmodified Linux kernel can be used to run with
para-virtual drivers. As it's drivers that cause the biggest performance
loss when running full virtualization, this is going to be a great
feature. When there will be para-virtual Windows drivers is another
interesting question... 

Note that para-virtual drivers will currently not work on "bare-metal",
and each VMM/Hypervisor would have slightly different interfaces for
their para-virt drivers, so a VMWare para-virt driver for Windows would
be very unlikely to work with Xen, etc. 

--
Mats
> 
> 
> What are your thoughts about this?
> 
> Thanks !
> 
> MRB
> 
> _______________________________________________
> Xen-users mailing list
> Xen-users@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-users
> 
> 
> 



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


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.