Rare project developers

[Cristina Klippel Dominicini <>]

Hi Csaba,

Thank you very much for your support. We are very happy with the fast 
progress of the PolKA integration into freeRouter. We are now trying to test 
and exploring your modifications in the freeRouter code to understand how to 
contribute.
@Frederic, is tomorrow freeRouter's meeting confirmed? I think it would be a 
good opportunity to wrap up our current status and next steps.

Regarding the nodeIDs, in the Chinese Remainder Theorem, it is sufficient if 
the nodeID polynomials are pairwise relatively prime (Section 17.4 of 
https://shoup.net/ntb/ntb-v2.pdf). For simplicity purposes, we use 
irreducible polynomials to assure this condition for any case (similar to 
prime numbers in the integer arithmetic). Please, find attached an example 
table of irreducible polynomials. I refactored the code and have just 
submitted a new commit in the shared repository to include the computation of 
nodeID polynomials with some examples (a random test and also a test using 
the values from the ONDM paper). I added a simple code that tests all the bit 
possibilities to generate a list of n irreducible polynomials of a specific 
degree (for real scenarios, I believe the best approach is to lookup a 
precomputed list of nodeids). Please let me know if this code helps. We still 
have to produce more mature control plane functionalities that explore these 
basic functions.

Regarding the design choices, the discussions with you, Frederic and Jordi, 
were very helpful to tune our arguments about the potential benefits of PolKA:
- Tables: In the paper, one argument was that PolKA reduces the number of 
states by not relying on tables. However, I agree with you that the use of an 
almost static table to map the index of the next hops does not discredit the 
original argument. Thus, your suggestion to use BGP/IGP for node discovery 
and reuse the Segment Routing identifiers seems to be a very good solution. 
We still reduce the management burden of maintaining routing tables when 
compared with traditional approaches.
- Header rewrite: In the paper, one argument was that PolKA does not need to 
rewrite the routeid in the packet in each hop. However, I agree with you that 
we need to rewrite at least a TTL field and ethernet address to prevent loops 
and allow PolKA packets to traverse bridged networks, respectively. Thus, I 
think we shouldn't focus on this argument.
- Fixed header size: In the paper, we briefly show that PolKA's pipeline is 
simpler to implement when compared to a list based approach, because it does 
not need to operate over variable headers. As you said, this can be a strong 
benefit considering the current hardware limitations in the maximum number of 
hops for var header operations, as used in segment routing.

From your previous email:
>>some more progess here, as you're carrying ethertype within polka header, i 
>>added some mpls over polka test cases... so one can tunnel mpls vpns (both 
>>layer2 and layer3) through a polka domain... 
>>https://github.com/mc36/freeRouter/commit/aa261c3e3a5bf33f85e8068cf7555653f24c4e85
Nice! We used the Ethernet header as a Proof of Concept, but we could 
encapsulate the PolKA header in other protocols if you think it is more 
adequate.

Best Regards,
Cristina

________________________________________
De: mc36 <>
Enviado: sábado, 30 de outubro de 2021 06:24
Para: ; Frédéric LOUI
Cc: Cristina Klippel Dominicini; Rafael Silva Guimarães; Moises R. N. 
Ribeiro; Magnos Martinello
Assunto: Re: [rare-dev] polka is approaching rare... was: Re: custom crc32 
library in freerouter

>
> So basically Polka should be live and kicking in P4 LAB using RARE/freeRtr ?
>
> PS: Please let me know if you need P4 dataplane and bf_forwarder.py to be 
> upgraded.
>

so yeahhh, the freerouter upgrade does not carry (at the moment:) the 
tofino.bin....
so well the lab need a fresh bfrouter and bfforwarder... also, some features 
needed
to be taken out here because we ran out of stages... at my case it was 
multicast,
as it's a core (P) router here so no need for raw ip multicast, bier is 
enough 4 me...
attaching the constants table used to configure my stordis...

regards,
cs

________________________________

Esta mensagem (incluindo anexos) contém informação confidencial destinada a 
um usuário específico e seu conteúdo é protegido por lei. Se você não é o 
destinatário correto deve apagar esta mensagem.

O emitente desta mensagem é responsável por seu conteúdo e endereçamento.
Cabe ao destinatário cuidar quanto ao tratamento adequado. A divulgação, 
reprodução e/ou distribuição sem a devida autorização ou qualquer outra ação 
sem conformidade com as normas internas do Ifes são proibidas e passíveis de 
sanção disciplinar, cível e criminal.

Attachment: poly.pdf
Description: poly.pdf