kernel level virtualisation requirements.

[Marko Zec]

On Wednesday 17 October 2007 00:54:21 you wrote:
> Andrea Campi wrote:
> > In para-virtualization you modify the kernel source in such a way
> > that accesses to the hardware are instead translated into calls to
> > the hypervisor. This means you could simply write device drivers
> > for a "virtual network adapter", "virtual disk" etc. What this buys
> > you is that you can have a full kernel (say 6.x) running as a
> > hypervisor, and trimmed down kernels (say 7.x and several 6.x
> > versions), compiled with only the virtual device drivers, running
> > as additional VMs.
> >
> > WDYT?
>
> Well Xen does paravirtulization like you described (and I agree
> something like that is more flexible then jails, if supported by
> other operating systems).

Actually, resource virtualization done at kernel level could offer great 
degree of flexiblity.  Ideally, a modular virtualization framework 
would allow one to virtualize only the resources one needs, for example 
having a single process talking to several isolated networking domains, 
or having several processes bound to the same slot in a proportional 
share CPU scheduler, sharing or not sharing the same filesystem 
hierarchy etc.  I think the thrust of this thread was in tackling 
people's imagination on how such a modular virtualization framework 
should look like, and which capabilities it should offer and which not.  
I.e. not get carried away in comparing kernel-level virtualization in 
general against Xen and alike, which are undoubtably very useful tools 
which have secured their place under the sun...

Cheers,

Marko


> DragonflyBSD has its own flavor of 
> virtualization similar to user mode Linux, but it has greatly
> diverged from FreeBSD so it't probably not trivially portable.
>
> Or do you mean something like this:
> http://feanor.sssup.it/~fabio/freebsd/lkvm/ ?