[Olsr-users] Virtual Machine testing net

[Benjamin Henrion]

On Fri, Apr 10, 2009 at 12:44 PM, Henning Rogge <(spam-protected)> wrote:
> On Freitag 10 April 2009 09:43:21 Krzysztof Szuster wrote:
>> On Wednesday 08 of April 2009 12:00:07 Henning Rogge <(spam-protected)> wrote:
>> > (...)
>> > The first step is just a compile test... then we do some VM tests in
>> > small scenarios (maybe 4-8 nodes) to see if something is horrible wrong.
>> > (...)
>> > Henning
>>
>> I'm really interested in testing OLSR network on some kind of VM network.
>> I don't want to build a virtual network based on VmWare, qemu or like. A
>> network containing 10-20 nodes would consume a lot of memory on host system
>> and a LOT of work to configure it. The right soution would be creating
>> multiple instances of ONE virtual machine with some script tool to
>> configure and interconnect nodes (maybe UML?). There is also very promising
>> olsr_switch tool in olsrd project. This is pretty smart approach. As I
>> understand, it's a virtual "network" or switch connected to a set of
>> olsrd's on socket level.  However olsr_switch is a bit buggy, perhaps not
>> maintained any more? What are your experience Henning? I'd be grateful for
>> any suggestions.
> I don't think the OLSR-switch is really a great idea, because it does not use
> the normal routing code and have some other special codepaths. Just set up a
> number of VMs with OpenWRT on it, so you need only 8-16 megabytes of RAM for
> each VM. This way you can easily emulate 100+ VMs without loosing the tests on
> a real linux system.

OpenVZ is nice for that. But I still have to find time to port the
Openwrt images as containers.

If you have a machine supporting hardware acceleration (Intel VT), KVM
might be better.

Unfortunately, I don't have a machine which supports this here.

--
Benjamin Henrion <bhenrion at ffii.org>
FFII Brussels - +32-484-566109 - +32-2-4148403