RARE user and assistance email list

[mc36 <>]

hi,

On 4/20/22 16:59, Moises R. N. Ribeiro wrote:
> looking into the future (not for this proposal) we can use (and actually reuse for the sake of scalability) multiple polynomials IDs at a node for also identifying a 
> partition (slice) and QoS policy in PolKA. Is that correct/feasible Csaba? 
that is already the case, that's why you configure polka enable under every 
interface...

> And in that case, how the queueing policy would be implemented: user-level or data plane? I am not aware 
> how P4 would handle that (translation to data plane and queuing policy itself?
for queueing to happen, you need ram near the asic, and that's not the case 
with the current tofinos...

> Another point is that the integration with MPLS is a hot topic here 
> https://www.segment-routing.net/srv6-conferences 
> <https://www.segment-routing.net/srv6-conferences>
> In your opinion, shall we look at that with PolKA too, or we rather try to 
> push PolKA toward Ethernet-based modern solutions (like TSN) for its framing 
> solution?
polka carries ethertype and the payload, which means it can carry mpls 
packets...
i already test againts that here:
l2 mpls vpn: http://sources.freertr.net/cfg/mpls-polka11.tst
l3 mpls vpn: http://sources.freertr.net/cfg/mpls-polka12.tst

br,
cs