RARE user and assistance email list

[mc36 <>]

hi,

@frederic, please extend my responses with some shows,
you have a live wg to nmaas as far as i know?

On 4/14/22 13:34, Moises R. N. Ribeiro wrote:
> Provided that we're going for the "overlay" architecture, forwarding could be a bottleneck as Freerouter sits at application layer. Efficient data plane mapping would be key to 
> unleash performance (in particular: throughput. Please don't try to fool us with your ping's RTT with !!!!! :-) ) . Any particular comment/results/post/report/publication on that 
> performance issue?
so true, that's why we need the dataplanes... :)
since they do the forwarding and only the exception packets
are handled in freerouter, there is no bottleneck at all...


> My questions are:
> 1) Is user-level Freerouter's port mapping done/translated at data plane?
yess, when you execute the 'externalize forwarding' part from freertr.net,
it'll convert your ethX interfaces to sdnX interfaces, and the counters will
have sw+hw notation afterwards...

> 2) Is wireguard cryptography done at user-level or it is offloaded to data 
> plane yet?
the key exchange is done in freerouter and the actual packet processing is 
done at the dataplane...
but see the list of hardware capabilities at rare.freertr.net, the list 
contains a lot of crypto stuff also...

> 3) Besides being the "protocoland" it is, for the sake of user's programmability, is there room for Freerouter supporting kind of user level "high level" scripts/templates that 
> would be translated to P4 at data planes, such as:
> https://arxiv.org/pdf/2204.02739.pdf <https://arxiv.org/pdf/2204.02739.pdf>
yess, as it's foss, anybody can hack whatever he wants on top of the existing 
codebase...

br,
cs