Rare project developers

[Frédéric LOUI <>]

Hi Cristina,

Thanks for this preliminary report.

1) Can let you trace the packet path using bf_swithd console log (which is 
something you don’t have with the real p4 switch)
2) As we check together they were no Polka message observed on the P4 switch 
and therefore no entry in the Polka table
    This tests would let us verify if this behaviour is also applicable to 
the DPDK Dataplane. 
    If not, we can try to replay the tests by adding a DPDK node as Polka 
edge node in the Lab.

We can have certainly this VC Tuesday 27th during our RARE VC 4 PM Paris time.

Hopefully mc36 would be here, it would be a good Polka troubleshooting 
session !

All the best,
Frederic

> Le 19 sept. 2022 à 01:43, Cristina Klippel Dominicini 
> <> a écrit :
> 
> 
> Hi Csaba,
> 
> Thank you! It will be great to discuss with you and Frederic. Everson has 
> already started his 3-month internship at Caltech and he is helping to 
> configure their P4 testbed this week. Is it possible to schedule the VC for 
> the other week on September 27th?
> 
> In the meanwhile, I will report the preliminary results, but Everson can 
> give more detailed information. During the investigation with the Tofinos 
> in the RARE testbed (AMS&FRA edge-core + POZ&BUD core), when we use a PolKA 
> node as edge, it seems that the entries for encapsulating the traffic are 
> not installed in the tables of the fastpath. This may be the cause of the 
> performance degradation when we send traffic with iperf. Note that, in our 
> paper with ONDM, we were replaying packets that already contained the PolKA 
> header and no performance issue was detected, so we can infer that the 
> problem is in the edge function and not in the core.
> 
> Moreover, we have seen a huge amount of traffic going through the eth0 when 
> the traffic goes through a PolKA tunnel with PBR. Please see attached 
> printscreens. Note that the issue does not happen in Segment Routing 
> configuration, only in PolKA. One difference is that, with PolKA, we needed 
> to configure the vrf forwarding, as you indicated in a previous email. 
> Configuration can be seen here:
> wdv.rare.nmaas.eu/uploads/commits/master
> 
> We talked with Frederic and he suggested some tests in the RARE testbed:
> 1) Only PolKA (remove simultaneous SR configuration): AMS&FRA edge-core + 
> POZ&BUD core only
> Result: No difference, same issues happen
> 2) Only PolKA: AMS&FRA edge only + POZ&BUD core only
> Result: It is not possible to configure a PolKA node as edge only
> 
> It is possible that our configuration is not correct, so Frederic also 
> suggested to try to reproduce the issue in a virtual environment:
> 1)  each VM uses a P4 studio installation with bf_switchd to emulate the 
> RARE testbed
> 2)  each VM uses a DPDK installation with P4emu to emulate the RARE testbed
> Everson is working on these tasks.
> 
> Best regards,
> Cristina
> 
> 
> ________________________________________
> De:  <> em 
> nome de mc36 <>
> Enviado: quinta-feira, 15 de setembro de 2022 03:02
> Para: ; Cristina Klippel Dominicini; Frederic LOUI
> Cc: Cristina Klippel Dominicini; Everson Borges; Marcos Schwarz; rafaelsg; 
> Moises R. N. Ribeiro; Magnos Martinello; Domingos Paraiso
> Assunto: Re: [rare-dev] Support with PolKA/SR-MPLS experiment
> 
> hi,
> surely yess for the vc, will the next regular rare freertr session work for 
> you?
> thanks,
> cs
> 
> 
> On 9/1/22 21:34, Cristina Klippel Dominicini wrote:
>> Hi, Frederic and Csaba,
>> 
>> I hope this email finds you well :-)
>> 
>> Everson has been working in the PolKA/SR-MPLS tests in the RARE testbed. 
>> Following your last instructions, we are setting up the Tofino nodes at 
>> AMS and FRA as both PolKA (or
>> SR-MPLS) core and edge nodes, and the DTNs generate traffic (topology file 
>> attached).
>> 
>> However, we detected some inconsistencies in the results and have been 
>> facing some difficulties in debugging whether the problem is in our 
>> configuration (most probably) or in the
>> freeRtr implementation of PolKA when edge and nodes are in the same node 
>> (our previous emulated tests were using separated edge/core). I can send a 
>> more detailed description about
>> the debugging and, if possible, it would be very good if we could set up a 
>> VC with you to show what we have done. Is it possible please to use part 
>> of the slot of the freertr-dev
>> meeting on September 6th?
>> 
>> Everson has just arrived in Caltech to start a 3-month internship, with 
>> the goal of developing these PolKA/SRMPLS tests for the SuperComputing 
>> conference.
>> @Everson, when you settle down, could you please share with us the backup 
>> of the config files that you are using at the RARE testbed?
>> 
>> Thank you,
>> Cristina
>> ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>> 
>> Esta mensagem (incluindo anexos) cont  m informa    o confidencial 
>> destinada a um usu  rio espec  fico e seu conte  do    protegido por lei. 
>> Se voc   n  o    o destinat  rio correto deve
>> apagar esta mensagem.
>> 
>> O emitente desta mensagem    respons  vel por seu conte  do e endere  
>> amento.
>> Cabe ao destinat  rio cuidar quanto ao tratamento adequado. A divulga    
>> o, reprodu    o e/ou distribui    o sem a devida autoriza    o ou qualquer 
>> outra a    o sem conformidade com as normas
>> internas do Ifes s  o proibidas e pass  veis de san    o disciplinar, c  
>> vel e criminal.
> 
> 
> ________________________________
> 
> 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.
> <sr-tunnel.png><polka-tunnel.png>