DASH Behavioral Model WG, 12 May 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: DASH Behavioral Model WG 20220512 (May 12, 2022)

Description

May 12, 2022
Discussed updates to work items, new CI/CD resource, IPv6 PR approval, test infrastructure, and new content in pipeline ReadMe.md file

A

One I wanted to let you know that I received an update from vincat this morning and fincat was working on the udp support, and so he was letting me know that he's presenting at ocp tech talk today, so he'll miss this meeting, but with he feels with respect to udp support, he feels he's done with the code, so he's waiting for test infrastructure to commit the code.

A

um He said he was going to wait a week or else he'll go ahead and commit, and then he also said he's still working on fragmentation, support and he'll be able to give us an eta at the end of this week.

A

So that's where venkat is on on his items.

B

Okay- and I also provided my update right earlier this week, uh christina.

A

B

uh Yeah so I incorporated you know in the last meeting we went through the review of the pr and, uh as per um the feedback that was received, I made the the requested changes and or required.

A

B

Yeah exactly and then then I have basically, you know, um you know, pushed the changes uh into the same pr and uh just you know now waiting for the feedback. So hopefully you know, if marion, can um you know he's okay with it? Then we can merge the change. We can merge. The changes then.

A

Yeah, let me get to that one here: honey. There we go okay and.

A

I need to look up the pr, I'm still not super well versed um number 88.

B

No, I I think it's 97 right, isn't it oh.

A

97, let's do nice, unbelievable, okay and my my dog just escaped outside. I have to run get him, but I'm gonna pull this up. Okay, okay, here we go.

B

Yeah, that's it.

A

Yep yep, thank you so joint table our conversation last week and then requested a review for maureen and prince. So you guys go ahead and talk for a second. While I grab my dog.

B

Yeah so uh marianne, um you know I'm not sure whether you get a chance to take a look at the uh the recent changes. So if you could, please, you know, um give your feedback on it and then we can proceed with you know the next step.

C

Yes, so two things about that. First, I remember it was the only thing that um I requested to change, uh so I believe I can approve that I I will review it of course, but I uh I think that it will get an approval right away, but another thing is: I was working on my testing infrastructure. There were a few things that went into the behavioral model just to make it compile right away.

C

So the thing is that, uh since the testing infrastructure is a bigger, bigger, pull request, I would like to merge this first and then proceed with the next ones. That will be great.

B

They'll help me with basically not do the rebase again and then do it so yeah appreciate it. Thank you.

B

A

I'm gonna grab my mouse okay, great, thank thanks. Honey.

B

Yeah, thank you.

A

Okay- and I believe we were- we had a date may 19th so may 19th. Is this connection tracking and simulator? um Do we have something to talk about today, marion on the simulator.

C

uh Not yet worked still in progress, we submitted a proposal to the pna community that will help us support connection tracking, but that's it uh we'll wait for the feedback. As soon as it's there, uh we will publish our uh connection tracking based on pna.

A

I actually attended the p4 meeting this last monday and I put keshav's question in front of andy and the p4 did anyone else go to that.

C

No, I didn't have an invitation by the way.

A

C

Slight misalignment: that's.

A

Okay, it comes from andy and I'm trying to remember exactly what they said. Let me maybe I made notes sorry.

D

uh Christina I was there, I did.

A

Make notes oh, go ahead.

D

Yeah so marion, uh I think the conversation was around.

D

You know we have a flow table where the learning happens and the connection rate supported by the hardware may be different than the line rate of the traffic, and the discussion was when the learn happens as part of the learn policy, if you would like to apply a policer for a given vm traffic, and that is to achieve the fairness across the game, traffic, okay.

D

So so what was the fairness as a connection rate and other was fairness as the available resource? So that is the max number of entries. uh Vm can occupy. You know, so that was the discussion. I think there was a suggestion by andy that we will have uh uh some n number of bits for a policy profile that can be derived uh for a given uh flue learning connection and based on that policy profile.

D

I think that's the semantics of it in the in the p4 program is still being worked out, but but that was the gist of the discussion.

D

Back to you, christina.

C

A

Well, I'm just trying to put it together. um You know I took a a lot of notes here, and this was more of the conversation about well. We could do this or we could do that or we could do this other thing um and then alan lowe had something more questions. You know, and so is andy here today um anyway, um we can thank you for the summary there and then I guess we can reach out to andy or wait for before pna. So thank you and I hope this answers.

A

Kashav's questions he's in india, so.

A

So he might not be able to attend in person here.

D

By the way, sorry, what's up, this is john hi.

A

Oh sorry, you guys have the a similar sounding voice and I need to put up two screens so that I can see who's talking. Let me do that. Real quick.

A

Okay, I'll put this screen on I'm running a second monitor today, so I'll do that? Thank you, okay, so so um I'll put I'll go ahead and put vincott's updates in there and is there something else you guys want to talk about today.

C

So I want to ask everyone uh to review the testing infrastructure, pull request, the one that introduces psi library implementation, so we can configure the simulator.

C

There is a lot of changes, but I think, like the best way to do it is just to switch to that branch. Just clone it and try it out yourself.

C

There are steps in readme described what needs to be done.

A

C

A

Me and the behavioral model read me or.

C

um So let me let me send you a link one second, okay,.

A

Or you can present if you like, or I can.

C

I'll just quickly send them.

A

C

Yeah, pull request is 108, and this is the readme that I'm talking about.

A

C

No, no, I just posted the link in the chat.

C

A

Readme got it right.

C

Yeah, so this is just a few steps uh and it's supposed to work without any issues. I want to make sure that this is compatible across different environments. I saw a few people complaining that the simulator couldn't add the interfaces.

C

That's because docker was changed to run from the same user and the simulator needs to be still run from sudo. So there is a slight difference, but I fixed that so you shouldn't encounter any issues at this point. So here is just make sure that you will rebuild the docker. That's the first step. Then you can build all of the pipeline related stuff like the p4 and the psi, and also build the test, and then there are steps to run them.

A

I think, in the notes, last time I had tried to capture what you said about you know: rebuilding docker make docker if you've run it before you'll need to.

C

A

C

A

C

Has all the steps.

A

C

So if anyone can try this out and give me a feedback.

A

Sounds good, um pretty quiet so did anyone else have q a questions and answers or anything.

A

So it sounds like everyone's on mute. We don't have any questions or answers um anything else for today then guys.

C

I think we're good. I will review pull requests by muhammad and uh venkat.

A

Okay sounds good.

D

I was just gonna.

A

Look in my I was just gonna look in the pr's that I had. um Let's see james okay, I think that's what we need to do so.

C

A

Okay, oh it looks like chris is in the lobby. Again, I don't know why that.

C

Maybe he dropped yep.

A

Okay, I'm gonna stop the recording now real, quick.

A