DASH Behavioral Model WG, 29 Mar 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: DASH Behavioral Model Weekly 20220329 (March 29, 2022)

Description

No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).

A

Or the sonic using the new ipdk layer that you mentioned and it uses the dpdk pipeline, which is p for programmable, hence p4dpdk, so um yeah. That work is ongoing. I think uh in in some time I won't be able to tell exactly when I would say in q2, early q2, probably mid q2 we'll be able to demo, and definitely this work will be open source. uh It will be in the pna architecture.

A

um You know a lot of things that will be there in the beginning uh will be l, layer, 3 and and other basic features, but if we, um but in the p40 pdk itself, we are already adding the overlay connection tracking as well uh and that needs uh you know uh that can be used directly with the p4 dpdk.

B

Thank you, russia, I actually I I was referring specifically to support for the adonis and modifiable table entries. uh I guess in order to do connection tracking you, you need to have that yes, but but is it something that's already available today I mean you spoke about two different things. If I understood correctly, one is the integration with sonic and the other one is the the implementation of the overlay and the support for these stateful functionalities.

B

So is the second one already available, or also that one is going to be available um at a later date.

A

um I uh in just a couple of weeks, probably in our next call, I will be able to update on exactly when um the p40 pdk, with a hit on I hit on miss add-on, miss my goodness. Sorry about that. I have lots of things. Coming up um will be available right. The connection tracking, I think it's in uh full progress either it is already demoed or will be coming soon.

A

I will update in the next next call: okay,.

C

Just does that answer your question.

C

Sorry, I only have one screen here so.

A

Yeah, mario, if there are any questions, please uh send them over uh um yeah. I think that work is in progress. I can update, uh probably even sooner than the next call, if it is uh urgent to make any decisions- or you know I'll I'll.

C

Update us on a second next call is tomorrow, so I think I think we could wait till.

A

Tomorrow, I see okay.

C

B

A

About the behavioral model, uh you know volunteer sink call, but that's uh I will try my best to see if I have some answer by tomorrow. Thanks.

B

Yeah, I mean, I think, it's more more up to marion, because I mean they are working on the behavioral model, so they might want to. um You know, decide whether to invest on that one or just use the you know the p4 pdk. I don't know so. I was asking more for their for their planning. Yeah.

D

Right now we are considering both because we don't know where- uh which one would be better. So as soon as we will have both actually available. Maybe then we can think about what to choose right now. We don't even know uh where effort will be bigger, where the integration is uh more difficult to do um so yeah. So we are working on bmv2 and we would be glad to see what's the status with dpdke as soon as it is available. So, let's, let's maybe wait for that.

C

I'm just making notes to myself here: okay,.

C

And I I did see something in my email about actually running the pipeline and marion had made a suggestion. um I can't dig in my email right now. uh Well, maybe I can here um muhammad muhammad. So uh do you do you guys want to let everyone know what you were talking about for that particular issue.

C

Is muhammad on the line, I guess.

E

C

I'll, let him explain.

F

So, actually, as part of the wrap up, we were actually trying to run this previous part on the bmp model and uh yeah. That's what people doing and we spaced that out.

C

So trying to run a serious pipeline, blah blah, ran into an error and.

C

Changed so yeah and then and then marion um replied here.

C

So if you change these here, um it will work with bmv2. Is that clear to everybody.

F

G

Can you send that this email to the group uh christian.

C

Yeah I'll clean it up and send it out sure I will make a note to do that great and so overall, um for these pieces right here in the backlog, can we move forward with bmv2 without these marion.

C

Like udp support, for example, I'm not thinking so.

D

I don't understand the question: what do you mean? Can we move on with bmv too.

C

Well, in the in the development we said, we need all of these features um and it doesn't look like we have resourcing to work on these. So in order for the whole model to work, we need these things right.

D

Yeah, okay, all right.

C

So I'll bring it up again in the community meeting and see um tomorrow and see if um anyone can free up to look at these.

G

I have a question on the psy api marian. I I remember.

B

G

Pull request or some code change to do the psi ap generation from behavioral model right.

G

Yes, so if that's the case like if you write the dot p4 file and then if you use that generator, we will get this, I apis right. I mean automatically right what activity is this? I don't understand the the iap for connection tracking, so you you are expecting um basically some.

B

Changes in the.

G

Iap generator um or like.

D

um So for connection tracking, we haven't finalized that yet uh I don't know if this is going to be auto generated, or maybe it will be a fixed ati. We don't.

G

Know, okay, so I'll say it to me christian on that one I will work on it.

C

Okay and I'm flipping between screens, that was.

G

uh Psi api for connection tracking.

C

G

C

Yeah right, I got that it was vent m yeah.

G

E

So yeah, I have a question uh um guys you know so on these ones. I was looking at all the the backlog part, and then I was thinking about whether I could you know um sign up for say, ipv6 support, so my question for marianne is before I basically do. That is that, if, if there is a you know equivalent, um you know, support provided for ipv4.

E

um Is that, along the same lines that we have that we have to provide the ipv6 support as well? So if there is already an example of you know, ipv4 support that is available. That would really help to.

B

E

Okay, now there is an equivalent support that is there and then can. If you can point me to that, then then then I can, you know they'll become a much more. uh You know smoother transition for me to really take over and then you know say: okay, if I can basically add the ipv6 support for it.

D

uh Okay, okay, uh so basically all of the much action tables that are working with ipv4 but to require ipv6 support. I I can get you that.

E

Yeah, if you can get me that so so chris uh christina, if you could, you know, sign me up for ipvc support, then then I'll go ahead and basically you know um work on it.

C

E

Thank you marian and thank you christina appreciate. It.

C

No, no, I thank you.

C

I'm still kind of learning how to work with these um great. So that's really all I had for today guys uh marion did you have anything else.

D

That's actually good, because I have a heart: stop in 10 minutes.

C

Okay, maybe we could are all of these prize zero. I guess that would be the question. Do all of these need to be done right away then, or do these kidneys have any kind of priority.

D

Oh priority: it's a good question. Actually I would need to go over all the lists to understand. I.

C

D

I cannot say right now, which one is more uh critical.

C

Okay, great well, then you and I can work on that offline um throughout the week and come up with come up with the priority. So um anything else from the rest of you guys on the phone I feel like marion, and I and a few people have been talking a lot. So is there anything else you want to bring up before we go.

G

Yes, you can say first, then I will say yeah, okay, so, regarding the priority marian um see in order to make any p4c or bmv2 changes, we really need the p4 file right. Then only you can see what changes are required from compiler or bmv to simulator perspective. Right I mean you mentioned about you're. Making some changes in the bmv too. uh Is that, uh regarding what like, I, I didn't get quite get that part.

G

Which item are we talking about the bmv two? You mentioned right marian in the beginning, um you are working on the simulator part right.

D

Yeah we are working on supporting the epna add-on mess in bmv2.

G

Okay, an import okay,.

H

Thank you yes, so I have one question from maria: can I ask.

C

Yep he has a hard stop and uh eight minutes to go ahead.

H

Yes, uh yes, so it was mentioned as a prerequisite for that contribution that we need to have a skill set of compiler background and bmw code. So how much compiler background is needed for modifying the p4 uh p4c code? So the effort that are needed in these projects uh are more related to bmv2 modification itself or it is like mix like we have to know the background of bmv2 plus p4c, because we are doing internal ramp up.

H

So we are just asking that from which area we focus more like bmv, 2 code base bc, or we should see e4 c code or both are needed.

D

Oh so, actually, if you look at the comment section, there is a scope for each item, so it's as if it's either the changing the p4 code itself or it's changing the p4c code. So from from that scope you can figure out uh which area it needs uh more attention and more background in so if it says p4c, then you need to be. You need to get familiarized with p4c code. If it's bmv2, um I think bmv2 items are mostly assigned to to nvidia already so.

H

D

The p4, like our uh our dash pipeline code, um which requires only getting familiar with the p4 language itself, or there are some items that need you to to get into p4c. So in the scope it says about every item, what you need to be familiar with.

H

Okay, so it is required that you share the link of this project in chat. Maybe so we.

B

Can see this book.

H

C

Hey here we go.

C

Hey hi anyone else.

F

uh Yeah, can you hear me yeah, uh yeah, so uh christina as part of the email? I also wrote that um I shared my.

C

F

Account so that you may have into the project as well.

C

Yeah, I will. I will definitely do that after this meeting mohamed okay, thank you. I will do that and then what will happen is you'll receive an invitation and then it may ask for two-factor auth, which then got m told me you might fail a couple of times, but eventually it'll work and then um I'll have you in there.

C

Thank you. Okay, all right.

E

Just uh just a quick question on this project view um christina as you assign the assignees to the items yeah.

C

E

See that you know assignees are assigned, but the status continues to remain. Backlog needs in order.

C

E

C

E

It has to be automatic, oh so it's not automatic. It needs to be.

C

That was me sorry about that. Okay,.

E

C

There we go now, it moved you're right. You are right.

E

C

E

Oh, what what happened to ipv6, I think that disappears right.

C

E

Here, oh down here, okay, once.

C

I changed the tag it went to the right place and then it'll reflect over here, so it'll it automatically syncs over to here. So.

E

C

You for that, thank you for that. Okay, great all right.

F

C

Right: well, thanks everyone, if there's no more questions um muhammad I'll, get to yours, you know in the next couple minutes here and thank you for everyone. You know offering up and volunteering, and um we don't have to make this a weekly meeting unless you guys feel it needs to be, um or would you we could check in over email as well, uh which do you guys prefer.

B

As far as I'm concerned, I don't mind to check in briefly, we don't have anything to discuss with gap and we gain.

C

B

But that's just my opinion.

E

Yeah, we can do it on a neat basis. I think that that would be.

C

E

C

Okay, that's a great idea because then we don't have you know, standing meetings on everyone's calendars because I feel like I'm in front of the laptop 12 hours a day, so yeah seriously. Okay! Well thanks! You guys um have a have a good rest of the week.

E