Re: dragonfly bsd and vkernels

["Samuel J. Greear" <sjg@xxxxxxxxxxxx>]

On Thu, Dec 8, 2011 at 1:51 AM, Joe Gain <joe.gain@gmail.com> wrote:
> Hi everyone,
>
> this is a really general question, but I'm just starting my bachelor thesis,
> which is going to have something to do with virtualization (probably network
> virtualization) and I was just wondering how dragonfly bsd's kernel
> virtualization fit's into the virtualization scene? How would vkernels fit
> into categories like full-virtualization (Qemu), paravirtualization (Xen),
> and O/S virtualization (Jails)? Also, are vkernels mainly for kernel
> development (and sandboxing), or do people also use them to create virtual
> machines for running servers, routers, clustering etc.?
>
> The Wikipedia article about User-mode Linux says:
>
>> User-mode Linux is generally considered to have lower performance than
>> some competing technologies, such as Xen and OpenVZ[citation needed]. Future
>> work in adding support for x86 virtualization to UML may reduce this
>> disadvantage.
>>
>> Often cited as a strength of Xen (a competing technology) is support for
>> Thread Local Storage (TLS). This is now also supported in the latest UML
>> kernels. Xen concentrates on virtualizing the whole machine, and thus all
>> systems running on a Xen machine are really virtual machines. In UML, the
>> host machine is not virtualized in any way, and only guest systems are true
>> virtual machines.
>
>
> And the Wiki article about dragonfly bsd says:
>>
>> The virtual kernel is run in completely isolated environment with emulated
>> network and storage interfaces, thus simplifying testing the drivers, kernel
>> subsystems and clustering features.
>
>
> And in an article from 2007, on the Kernel Trap site, one of the major goals
> of dragonfly bsd is said to be:
>>
>> My primary goal is to eventually have a fully cross-machine coherent and
>> transparent cluster OS capable of migrating processes (and thus the work
>> load) on the fly. Doing this properly requires direct, integrated support in
>> the kernel. We are probably two years away from accomplishing this goal.
>
>
> Can anyone say how this part of the project is developing? What are the
> competing products and how would dragonfly bsd like to fit into this market?
>
> Thanks,
>
> Joe
>
> PS. I'd be happy to update the wiki page with any relevant new information!
>
> --
> joe gain
>
> jacob-burckhardt-str. 16
> 78464 konstanz
> germany
>
> +49 (0)7531 60389
>
> (...otherwise in ???)

These two articles on LWN cover the vkernel architecture in detail and
will probably answer a lot of your questions.

http://lwn.net/Articles/228404/
http://lwn.net/Articles/230658/

In terms of where it sits compared to other technologies, it is most
similar to UML, but also Xen PVM, minus all the actual hardware
emulation aspects of Xen PVM.

Vkernels are mostly used for kernel development and testing, but also
certainly for isolation.

SSI aka "fully cross-machine coherent and transparent clustering" is
not being actively pursued, but filesystem clustering is. There is no
ETA. Competing products? I do not believe the project as a whole has
any commercial aspirations, we're just trying to build a system that
is stable, useful and highly performing. A clustering file-system is
believed to fill the useful aspect.

Sam