marionnet-dev team mailing list archive

["Jean-Vincent Loddo" <Jean-vincent.Loddo@xxxxxxxxxxxxxxxxxxxx>]

Hey Alexey,

I didn't know GNS3, thank you for the link. Yes, the goal is very close to 
Marionnet and your analysis about technical differences seems correct.

We have been planning for some months to study an integration of VirtualBox as a 
further engine for Marionnet's virtual machines. Not only that, we have been 
planning also a Dynamips integration, in order to simulate more closely CISCO 
devices.

About the VirtualBox integration, it seems to us, with our limited knowledge of 
VirtualBox, that there could be a problem. Actually, Marionnet takes advantages 
of the UML ability to deal with cow filesystems. When a Marionnet virtual 
machine is created, its virtual disk is a couple (B,F) where B is a (big) cow 
"Backend" and F is a (small, sparse file) cow "Frontend". In this way, when the 
user defines a project with some machines M1,..,Mn, we store in the project file 
(which is a tar archive) only the set of frontends F1,..,Fn. The project can be 
sent and re-used in another host only if the set of backends B1,..,Bn are also 
installed in the new host. Backends are a sort of "prerequisite" to run a 
project in a host.

VirtualBox provides a snapshot mechanism that seems equivalent: as far as I can 
see, the backend is the virtual disk (.vdi) and the frontend is the snapshot 
(again a .vdi). But what about the disk space? Could a snapshot be a sparse 
file, i.e. a small piece of data that we could put in the project file, as we do 
with cow files? This is our main doubt about the introduction of VirtualBox in 
Marionnet. Maybe, you will dissipate it immediately...

Thank's in advance,
Jean-Vincent Loddo

Le dimanche 15 mai 2011 20:54:22, Alexey Eromenko a écrit :
> Hello,
> 
> I am Alexey "Technologov", Open-Source community member, and
> contributor to GNS3 graphical network simulator and VirtualBox
> projects.
> 
> GNS3 is a sister Open-Source project, also under GPL, with a similar
> educational goal to Marionnet.
> http://www.gns3.net/
> 
> Recently I have ported GNS3 to VirtualBox engine:
> http://www.gns3.net/phpBB/topic3262.html?sid=9c8d0c8210c585dab6594e8a2536e8
> 49
> 
> I have looked at Marionnet, and it looks interesting and promising project.
> I would like to share different ideas about network simulations with you.
> 
> One thing, that I would like to see at Marionnet, is a user-forum,
> where people can exchange ideas.
> 
> Technical comparison:
> GNS3 is based on 3 underlying engines, Dynamips, Qemu and VirtualBox,
> which allows us to be cross-platform solution. Dynamic topology layer
> taken from 'Dynagen' project. GNS3 is also multi-host capable -- i.e.
> it has distributed architecture. -- it focuses on Windows, Linux and
> Cisco.
> Marionnet uses 2 engines: VDE and UML. (Dynamic topology layer
> developed from scratch.) - with focus on Linux.
> 
> I have read the paper about whole design of Marionnet in half-year. Good
> job. You're encouraged to take a look at GNS3.
> I hope for productive cooperation and discussion of different ideas in
> future.
> 
> Best wishes,