Announcement: Genode OS Framework version 14.02 released

[Norman Feske]

Hi Lonnie,

> In my case, I am currently focused on having NOVA+Genode+VirtualBox in a
> type of updated version of the NOVA v0.3 demo, but to spawn off VirtualBox
> instances and have a graphical Tutor section for a Type-1 Hypervisor that
> "might" be able to compete with the other existing Type-1 hypervisors like
> Xen, Hyper-V, and VMWare as I think that the NOVA+Genode+VirtualBox (once
> the networking and SMP areas are working) can be a great alternative to the
> other hypervisors while also allowing Genode/NOVA to move slowly into the
> competition space for hypervisors. 

that sounds good! Btw, networking and SMP should work fine with NOVA in
general. Of course, VirtualBox is still in flux. For example, there is
no support for using the network from a guest OS yet. But that will
follow for sure.

> This is the project that I am trying to ramp up and slowly get started.
> Additionally, my goal will be, beyond the initial Type-1 hypervisor system,
> will be to further modify VirtualBox (in a special fork) to allow for a
> "distributed SMP and distributed memory" VMM so that I will be able to run
> it on commodity desktops with one instance of VB running the desktop OS
> (like Windows 7) and another instance of VB running the distributed VMM and
> peer connecting to other distributed VMM's on the LAN to make a hugely
> distributed SMP machine (ie.. cluster with a single OS) and hundreds of
> nodes. I also want to keep the build as minimum in size as is possible.
> 
> In my day job, I would (as a Physicist) at a research facility where such a
> system would be very useful for running various modelers and simulators that
> we run and which has prompted my strong interest in NOVA+Genode+VirtualBox.

This is much more involved. To me, it ultimately does not look like a
SMP system but a NUMA system. An OS that would span all the VMMs would
need to account for this structure to get good performance. By modeling
a system like this, you will certainly need to implement a distributed
shared memory mechanism. NOVA, with handling page faults at the user
level, is generally suitable for that. But the topic is still extremely
complex.

Best regards
Norman

-- 
Dr.-Ing. Norman Feske
Genode Labs

http://www.genode-labs.com · http://genode.org

Genode Labs GmbH · Amtsgericht Dresden · HRB 28424 · Sitz Dresden
Geschäftsführer: Dr.-Ing. Norman Feske, Christian Helmuth