Xen devel - Sep 2004 - two host interfaces to same virtual lan?

Thankyou for previous responses about my compiler question, I appreciate it!

I have another quick question.  I want to create a virtual LAN and have
domain0 be a NAT gateway to NATd VMs.  Is it possible for domain0 to
have two network interfaces on this LAN?

Thanks!


-------------------------------------------------------
This SF.Net email is sponsored by BEA Weblogic Workshop
FREE Java Enterprise J2EE developer tools!
Get your free copy of BEA WebLogic Workshop 8.1 today.
http://ads.osdn.com/?ad_id=5047&alloc_id=10808&op=click
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

> Thankyou for previous responses about my compiler question, I appreciate
it!
> 
> I have another quick question.  I want to create a virtual LAN and have
> domain0 be a NAT gateway to NATd VMs.  Is it possible for domain0 to
> have two network interfaces on this LAN?
I''m not sure I understand what you''re trying to do. The bottom
line is that you can think of the vifX.Y interface that appears
in domain 0 as being connected to the domain''s ethY interface by
a virtual piece of wire. 

You can create a domain with multiple eth interfaces, which will
thus appear as multiple vif''s in domain 0. How you choose to
interconnect these vif''s is entirely up to you. You can use any
of Linux''s networking capabilities such as bridging, routeing,
firewalling, NAT, etc.  When thinking about how to configure
domain 0, just think about how you''d do the same thing if instead
of using virtual machines, the other domains were actually other
physical machines connected via ethernet cables to dedicated
ethernet cards within the domain 0 machine. You configure
networking in Xen in exactly the same way.

Ian


-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM. 
Deadline: Sept. 13. Go here: http://sf.net/ppc_contest.php
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

I see..  I was thinking of the ''wire'' you''re speaking
of as a switch
that other domains (including dom0) could ''plug into'' somehow,
not a
crossover cable, in which case I would have to do it via Linux.  I
don''t
have any Xen installs right now, so I didn''t get a chance to experiment
(mailing list faux pas, but thankyou for answering!).

On Thu, 09 Sep 2004 21:36:47 +0100
Ian Pratt <Ian.Pratt@cl.cam.ac.uk> wrote:
> > Thankyou for previous responses about my compiler question, I
appreciate it!
> > 
> > I have another quick question.  I want to create a virtual LAN and
have
> > domain0 be a NAT gateway to NATd VMs.  Is it possible for domain0 to
> > have two network interfaces on this LAN?
> 
> I''m not sure I understand what you''re trying to do. The
bottom
> line is that you can think of the vifX.Y interface that appears
> in domain 0 as being connected to the domain''s ethY interface by
> a virtual piece of wire. 
> 
> You can create a domain with multiple eth interfaces, which will
> thus appear as multiple vif''s in domain 0. How you choose to
> interconnect these vif''s is entirely up to you. You can use any
> of Linux''s networking capabilities such as bridging, routeing,
> firewalling, NAT, etc.  When thinking about how to configure
> domain 0, just think about how you''d do the same thing if instead
> of using virtual machines, the other domains were actually other
> physical machines connected via ethernet cables to dedicated
> ethernet cards within the domain 0 machine. You configure
> networking in Xen in exactly the same way.
> 
> Ian
> 
> 
> -------------------------------------------------------
> This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
> Project Admins to receive an Apple iPod Mini FREE for your judgement on
> who ports your project to Linux PPC the best. Sponsored by IBM. 
> Deadline: Sept. 13. Go here: http://sf.net/ppc_contest.php
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/xen-devel
> 

-- 


-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM. 
Deadline: Sept. 13. Go here: http://sf.net/ppc_contest.php
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

The default behavior is a dedicated virtual "wire" (implemented as
shared
memory) from each domain to a virtual nic in dom0 - dom0 then does bridging / 
switching with the standard Linux bridge code.

You could equally well have dom0 do routing between the vifs, rather than 
bridging (in fact, we''d like to have scripts for doing this at some
stage).

HTH,
Mark



-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM. 
Deadline: Sept. 13. Go here: http://sf.net/ppc_contest.php
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

On Thu, 9 Sep 2004, Mark Williamson wrote:
> The default behavior is a dedicated virtual "wire" (implemented
as shared
> memory) from each domain to a virtual nic in dom0 - dom0 then does bridging
/
> switching with the standard Linux bridge code.
>
> You could equally well have dom0 do routing between the vifs, rather than
> bridging (in fact, we''d like to have scripts for doing this at
some stage).
You''ll get them at some point.

We use completely diskless machines.  Nfsroot and all.  And, the bridge code
turns off the primary nic momentarily, which then of course means / is no
longer accessible.


-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM. 
Deadline: Sept. 13. Go here: http://sf.net/ppc_contest.php
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel

> I see..  I was thinking of the ''wire'' you''re
speaking of as a switch
> that other domains (including dom0) could ''plug into''
somehow, not a
> crossover cable, in which case I would have to do it via
> Linux. 
In Xen 1.2, Xen implemented a virtual switch that interconnected
the other domains.

In 2.0, we just implement the ''crossover wire'' into domain 0,
allowing you to create one or more bridges and interconnect the
other domains however you wish.

It''s all zero-copy, so performance is good. We''re not quite as
CPU-efficient as Xen 1.2, but Xen 2.0 will still saturate gigabit
links with ease. The extra flexibility of being able to do
firewalling/NAT/routeing/bridging etc is worth the CPU overhead
for most users.

Have a look at the ''Restructuring IO'' paper on the Xen site
for
more details.

Ian




-------------------------------------------------------
This SF.Net email is sponsored by: YOU BE THE JUDGE. Be one of 170
Project Admins to receive an Apple iPod Mini FREE for your judgement on
who ports your project to Linux PPC the best. Sponsored by IBM. 
Deadline: Sept. 13. Go here: http://sf.net/ppc_contest.php
_______________________________________________
Xen-devel mailing list
Xen-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/xen-devel