Cloud Native Computing Foundation Technical Oversight Committee, 27 Apr 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: CNCF TOC Meeting 2021-04-27

Description

CNCF TOC Meeting 2021-04-27

A

All right, we are currently.

B

Rolling, please go ahead all right so uh starting at the top with vineyard, and I think this was the one where we just wanted to check whether they truly believed that cncf was the best home for them uh harry. Did you have any conversation with them about that.

C

Yeah, so I have reached out to the team and they also have another presentation: uh there's a seek storage. I think they already got some feedback from sick storage, so my personal view is that this project uh has an independent uh reason and field, but it actually lied with the clone native ecosystem. Very well, just like uh the kubi flow, like the other ei infrastructure.

C

So it's complementary to the existing tool state and I also think the feedback from the sig storage is same with me and will, since they also view this project as a complementary to these currency, unsafe landscapes, I think it's. It's actually has a good alignment with over a reason. um Yeah I'd like to have other feedback. If I don't, I don't know, I'm not sure if you stick start with the focus, is here and maybe can have more, you put them there.

B

Maybe sad or aaron have any thoughts about sig storage's opinion on this.

B

Sorry, what was the.

D

B

D

uh I unfortunately don't have any context on that. One erin, do you.

A

Aaron might have stepped away for a bit, so um maybe we revisit this one.

B

Okay, yeah, we can come back fine, so uh the next one. Don't let us forget that one, the next one was ssv and I, I think, ssvm we were slightly worried about the.

A

Name and then we have put in a new name as well and drop that one into email um right just bring it out over here. They would like to be able to change towards uh wasm edge time say that again tonight in here there you go. What's image run time. What's.

B

Image run time, put it in the chain.

A

ah um I should beg your pardon aaron is now available. We can move back to vineyard, bouncing back and forth this morning. All right.

E

I had to I've been continuous meetings and lots of coffee, so we know what that means.

A

Question is vineyard.

B

All right, let's jump back to vineyard yeah, we heard that they've presented to sig storage and and wondered if that yeah.

E

It's I don't know our initial thoughts, both alex and I felt like it was more something that would be under kind of the umbrella of like an apache project for like analytics. There isn't really a good um pillar within kubernetes. That is, you know very data centric and big data related.

E

um But I don't know it'd be interesting to see. I guess where it goes from, um because it's just sandbox right if it can get traction and evolve into something else it is, it does run in cube. It's it's utilizes, a lot of the functions. It's different, I mean that's, I think, has it going for it. It's not yet another project of the same flavor.

E

um We just don't really seem to have a very data, centric sig, I would say in cube. Yet today.

F

Yeah, but I I don't see why we shouldn't have going forward, and I mean maybe we do need to have a a working group on that um or something for a sake I mean I, I think there was another project as well in that space as well, that we come into them. Yeah.

E

And I agree I I I truly believe like that's the next evolution of kubernetes. Is you know, data governance, data security, um where it came from how it's being managed, how it's being archived and all that done automatically? I think there could be a whole new flavor of projects it just where does that belong today? Does it belong in storage or does like you said? Is there another working group that maybe should be more focused on that cornelia? Did you want to ask something to add? No, no.

F

Yeah I mean this project is not a very storagey project. Given it's about not not storing things. It's about serialization, it fits more with grpc in a sense, although not really either but yeah. I wonder whether.

B

We should have some kind of machine learning, kind of sig or something I.

F

Think it would be a good idea to.

C

Yeah, I also mentioned that before that we are missing a machine learning working group, but this is actually a very important actually uh field in cloud native ecosystem. uh We know that we missed the kobe flow, somehow it's fine, but I think there will be more projects coming. uh I also talked with sleek around time. Folks there are talking with death breaks.

C

I think they are also have some candidate project there, including of course the weingarten and more, I would say we should have something like a machine learning working group to handle this project in the future.

B

Yeah dave's uh put in chat, uh we didn't hear you yeah. I.

D

Guess you're a.

B

Mute or something um the data in ai foundation- I I mean, I think it's fine for us to overlap with them. uh You know in the same way that we have some overlap with the cdf.

C

Yeah, it's the same issue yeah, but I don't think as we should. You know, stop doing uh right things and just because they're overlapping, just like cdxd animation, we have a lot of ddos project in cncf. We have c-gap delivery, but you know there are overlaps if they see it, but they still need to do the right thing like this. That's my opinion. I don't know if coco corey has an idea of that.

B

I'm going to say actually conversation with harry did kind of slightly focus my mind a bit on the fact that a lot of machine learning is being done. You know all the model training is being done in the cloud. You know it's like an ideal application for using cloud resources to deal with these giant amounts of data for a short period of time. So it seems like a very natural cloud native application, so yeah, I would lean towards saying we should have a sake to kind of, or probably say to look in that area.

G

So we start with this project and then see if the leads here can form the sig. Is that what you're thinking less.

B

There might be other people in the community as well yeah.

H

Is that a first project in the sandbox of this kind? Do we have anything that belongs to the similar species.

F

I think there might be yeah, I think we might have something.

B

Or something and I think there's another.

F

One in is there another one in this spreadsheet as well.

H

I'm looking at yesterday, remember you, you did that you did it.

F

Yeah yeah vega is a little bit. It's not it's a little bit. Yeah there's some overlap. It's yeah, it's a sparkish thing. So it's in that it's in that kind of space, yeah.

H

Yeah, maybe it would be nice to do the yeah analysis of all the sandbox projects that we have to see what belongs in the space but yeah. I don't think we should stop uh wineyard from from applying just because it overlaps with other foundations.

B

All right should we do votes for.

B

B

Okay and moving back again to the project, formerly known as ssvm, which is now wasn't edge, runtime.

E

That is a mouthful.

B

Is it going to just end up being called wasmedge.

F

B

Mean I guess we can't really comment too much on the aesthetics of the name.

B

Any I mean we did talk about this last time. If I recall- and I think it was really that kind of separation of the project from its parent company that we were concerned about and the name being a very key.

B

You know element of that so chris, when you spoke to them about it, did that kind of highlight any.

I

They were open to the change they they get it. uh They understand the kind of concern of trademarks and overlap between company name project name and product, so they were cool. They were reasonable to work with.

I

D

Really have that.

I

Many dependencies too, so it's fairly fairly lightweight.

B

Any other comments or concerns before we start on a vote.

C

C

Yeah, I think it's.

B

B

Okay, moving on to chaos, blade.

G

G

I think we need a sick, alibaba.

B

Alibaba have been super active yeah.

F

Well, I was wondering if we needed a sick chaos since, because.

A

F

F

Sandbox projects.

B

We did say it was going to be a an area of.

B

C

B

Actually, reach.

C

Out to the kill, splay team at real- it's not my team. Of course. It's another team, but I think the interesting part is kalesplay is a new thing uh because it is trying to create a control plan for chaos tools. So if you look that at their presentation they have integration with the existing chaos mesh and uh lead miss project, so they are building a control plan. Follow these kills close blading uh kels tools, they have dashboards, so you can see what happens your cal system. I think that is a new idea.

C

Actually, of course they have. They also have their own chaos tool and they all there is overlapping. Actually, this approach, but I see there are doing some added value in the chaos engineering part. That's. Why also. I agree that we should have some kind of working group for that, maybe in the future, but this will be very interesting field. More project will come.

B

Yeah, I think it currently sort of sits under run time, but yeah sig testing there's an idea. We've got to start calling it tags from next week. They're tags.

H

Testing testing, maybe testing and conformance something that would hold the test frameworks and conformance frameworks under its umbrella.

B

Maybe um amy you could put on the um toc like agenda working document that we should talk about new sigs, because it seems like.

F

Tags, new tags.

A

New tags, I think that's fine, um simply because it's been one of those things that we have talked about as something that should be rising and hasn't, but focus. We only have a few minutes in here right all right, votes for chaos,.

B

B

The next one on the list is yet another reverse.

F

Proxy, I think this is the.

D

E

F

To submit to the cncf.

E

B

And I had to actually go and look up the fact that apparently.net is in a foundation and.

I

Yeah, it's a large community. Surprisingly.

B

I didn't just looking at it now: do we think it's a sufficiently diverse community not kind of entirely there's 48 contributors? I don't know if they're.

B

Where they will come from.

B

Any comments- or they don't actually have- I can't see in their am I looking in the wrong place, but I can't see in there.

B

Repo, I can't see like an owners or.

B

B

Anybody seen anything like.

B

F

um Just a kind of question: is it it's not.

I

F

Obvious from the description, I don't know what they've written in the dark, whether it's, whether we you would count it as cloud native.

F

I mean it's just says: it's a reverse proxy, which is.

H

They're just working on on on their kubernetes ingress controller, I believe, but then um I don't quite understand like what would be their differentiator uh from from other uh ingress controllers in the space.

H

Based on that, based on the youtube presentations.

G

The other one was in the readme itself, it says we expect er to ship as a library and a project template. So basically the idea is other people would include the in their applications. I guess.

H

Yeah that that is, that is what presents I wish there are youtube presentations, something that you include as a library directly to your c-sharp code code um and there you can see like how it's useful for c-sharp projects, but once it becomes like a kubernetes microservice.

H

I don't quite understand like what makes them um unique and different from from other ingress.

G

B

Okay, chris saying they'd also have to change the license from mit to apache. I imagine that's not a huge.

I

No there's a microsoft cla, so it'll be very fairly easy for them to do so.

B

Yeah supplied as a library.

I

How many people have net experience here.

B

I think this is the thing I mean if you look at that, why do they want to contribute for their point c? Is we think yarpshows.net is a capable platform for building cloud native, apps and infrastructure to run on both windows and linux?.

B

And they want to get more exposure to a wider ecosystem of customers and contributors beyond the.net ecosystem. I sort of feel like I I would. This is the first time I've really heard of like the net ecosystem wanting to yeah. It just feels like a very separate community to me. I don't.

B

But they do have a very large ecosystem, around.net.

C

I personally think if we should consider have something from the donate ecosystem. It's super large and it's growing very fast as more like alternative world. We barely have any connection with it.

C

If you have something from that community, it may actually maybe bring us with some even more projects in that field, and that is my personal consideration, but I'm not sure what is this project, I'm not sure the position of this project in the ecosystem, maybe just a small tool or it's already a you know- uh might be adopting technology, I'm not sure about that, because I'm not exporting this field. That will. But uh I personally think that donated community is something that we want to keep eye on and have some connection with them.

B

Yeah, I don't, I don't sort of feel anti.net per se. It just feels. I don't know whether whether this is a good fit with cncf or whether that should be a you know. Is it? Is it basically just a net library that should be part of some bigger ecosystem of dominate.

C

Libraries right right: we really need to look at what is this project or are you trying to solve.

F

I mean, I think that yeah I mean the reverse proxy's great, but I'm not sure I encounter it as far native. I think they have. They seem to have aspirations to have close their ties with kubernetes, but they're, currently all aspirational, rather than real. As my concern is that that it's I mean like I'm, not sure that, like nginx per se, is a cloud native project, even though it's used in quite a lot of environments for the same, you know as a reverse proxy. It's like.

F

It's not it's not by itself.

F

Got the characteristics like it's not not to say it's not useful in the context, but it's not a kind of it doesn't seem fundamentally that native with, without doing something else like envoy, has done and become it and had the kind of programming.

B

Model and everything.

F

B

Do we have any other projects that are libraries.

F

Yes, I'd say we I mean primarily libraries, I guess, like distribution is primarily a library, although you can run it as an application um and I'm sure there are others that are primarily libraries.

B

Although I feel like with distribution that you know it's, it's a giant piece of fundamental cloud native kind of you, gotta have a registry yeah.

F

Yeah yeah exactly but but it is kind of like it is architected to be used as a library.

B

All right do we want to vote on this one? Do we want to suggest that they go back to dot net foundation? Do we want to.

I

What what information would you need from them to make a decision? I guess is one way to think about it right.

C

Also, we need some information about what is this project, how this project allowed me the same safe reason, and that is a common question I have because, of course, I I think I don't have enough experience to look into the project deeply, so I think to have some input of feedback from their folks will be really helpful.

G

So should we send them to sig network yep.

I

Seems reasonable.

B

B

And what we said: we're we're asking sig network for their opinion on whether this is a specifically cloud native solution and whether it's offering something that our existing kind of proxy projects don't already offer above and beyond. Bing.Net.

J

Yeah I mean what I like to know is how it relates. You know: defers differentiates with envoy h, a proxy nginx. That's which is uh those are the similar projects, just just at least have some. Besides the fact that it it's written in c, sharp, yeah.

K

J

They I mean, given it's so modern, I'm sure they've incorporated some some some nice design decisions. I think it'll be good to know that.

B

Yeah all right moving on to cube invaders.

B

uh Question from amy presentation to sign network to review is this cloud native and houses different from existing cncf projects? Yes, I think that sounds good. Okay, perfect. We can move on cube invaders, which I think is basically written by one person.

K

uh It does look like one person I like how they they answered the question of: why do you want to come to the cncf? I think it's a great community and I'd like to be part of it.

C

That is actually a good reason, as long as I can add some value to the existing landscape. I think.

K

It's something fun.

B

It's fun is it going to. I is it serious, I kind of like it in a way. You know I love it as a kind of demo or something you know. That's the who doesn't like having space invaders in a in a project, but is it I guess two two things uh is this going to be a seriously useful contribution to the cloud native community, and the second thing is: is this potentially opening the floodgates to a absolute boatload of kind of demo fun projects that I.

F

I'd also kind of I mean we, we turned down a bunch of projects of this kind of maturity and number of contributors last time, because they were just too young.

F

It's it's not had a lot of contributions and I'd say that, from our experience with docker, we we had things that were found projects like the docker minecraft thing that um were great fun and great demos, but they didn't attract contributors in the long term, because it's not clear what you can do so much for a fun project like what the kind of scope of what it should be is um compared to a use, a more useful project.

F

um So I kind of availably concerned about how many contributors it would actually attract, given it doesn't have any kind of non trivial number of commit people contributing now.

H

All the commits are done like directly to muster not through the prs, so it's like yeah, just one person pushing code to muster, um and it would also be nice to to kind of like uh see how they differentiate themselves, uh from other case projects and in the sandbox like what? What extra do? They add.

H

um They add space invaders.

B

Yeah, do we need to vote on this one or are we gonna say thank you, but we I hate, voting, no on things.

I

It's like a policy of like no demo software, simple.

L

I

L

Looking for other ones like there's, one called wakapod as well, there's a couple of other fun projects like this.

B

Yeah yeah, I think fun though they are and much that we love them. Love to you know, I think, actually pulling them into the into the cncf is not the right, not the.

I

Right decision.

I

No kubernetes, I was actually dims, there's no place for like demo projects in kubernetes. I guess that was the incubator. There's a bunch of random stuff there for a while and that's been killed off right.

B

Okay and the next one.

G

I

B

All right should we move on to key plus.

B

I think I had a comment on this one as well.

B

I think this is just one developer as well.

B

If I recall it yeah, it's.

B

I mean it does have 295 stars, which is something, but it's a very, very small set of contributors.

F

Yeah- and I I mean given what it's trying to do, I think you'd kind of expect some contributors who were kind of uh software, vendors and things to at this point. In it I mean, I think at the point when we want to we'd be interested in it would be because there were people actually using it um to shape applications, which is not clear that there are at the moment agreed.

B

And it's been, I think it looks like it was started a couple of years ago. So it's had quite a lot of time to.

B

B

Yeah, I feel like maybe it just doesn't quite meet the uh community interest bar.

B

Anyone want to comment further, or should we move on.

B

All right so the next one is service, mesh performance and there's also mesherie, and I feel like. Maybe we should talk about both of those at the same time, um because they're, both kind of well service mesh performance, is figuring out how to test service, mesh performances, and it's coming out of the sig network service. Mesh working group meshery is a conformance testing tool. um I believe that the smi project has just I mean it's. It's all kind of tied up with lee calco and you know.

F

B

Roles that he's playing in these different groups- but I think mescheri- is now adopted by smi as the conformance testing tool. I am all for conformance testing tools that kind of establish whether or not something is con. You know like the way that kubernetes has kubernetes conformance test, and then we know that a distribution of kubernetes is really kubernetes.

B

This seems like a really good thing, but I just want when we were talking to um so dave and I had a sig network liaison meeting with lee earlier today. I feel like there's scope for getting alignment between the smi this smp and mesherie, so that we can say you know as the cncf. This is kind of how we go about understanding, whether something is a service mesh or what capabilities a service mesh has and what performance characteristics it has. um So I sort of feel like this.

B

We might want to accept or reject these projects separately, but I think there is also a kind of bigger consideration of whether we want to try and get these three projects to align a bit more.

G

A

G

Snp itself is a specification and the measuring is an implementation of the specification. uh The way I read it before.

B

Measure reason is a test framework, as I understand it to test whether or not a service mesh meets that smi specification.

G

And what it tests is basically, what is in the spec, which is smi.

F

It says here is both the canonical implementation of snp and a conformance tool for smi.

D

Which yeah yeah? I got the impression that measuring is kind of more than it's, like a combination of all of that. It almost wants to be like a control plane first for service meshes, so it's like a control plane for your control plane. It lets you deploy multiple ones and do these tests and measure their performance and all of that.

D

G

Is um we take measuring and not take the spec, like I mean, do they go together or do we need to do both? At the same time, I.

B

Think that's part of the question. Isn't it is like well? Are these things truly independent? I mean we do have uh precedent for things like no siri and tough. You know notary being the implementation of tough and they were treated separately, but this feels slightly different to me because it isn't the like. We were just saying it isn't the implementation of the interface. It's the way you conformance test the interface so.

F

It also is the canonical implementation for smp, which generally we've put a canonical implementation with the spec project. So tough has a canonical reference implementation and some of the other ones do. I think so. I would have thought that from that point of view, messaging sm should be with smp, but then, when it's also the conformance tool for smi, I think it should be with smi.

F

So I would tend to think that all three we should ask smi if they would want to adopt merceri and or smp first.

B

Yeah, I think I lean to the same conclusion.

B

I mean I don't have any doubt that they're, you know cloud native projects and they should somehow be inside the cncf, but I'm not sure they, I'm not sure it makes sense for them to be three separate things.

F

Yeah, I think it should be two or one things, probably I'll turn to lean towards one.

G

At the very least, I think we need sign off from the existing smi folks that they like this idea, and they want to like welcome this group uh to what they are doing together. Right, uh yeah.

B

Otherwise- and I think there's quite a lot of overlap between them and and you know that there is a blog post from lee on the smi blog post, talking about meshary, so yeah.

G

He's wearing multiple pads right, yeah.

B

Yeah I'd like to kind of see it all aligned a bit, you know and a bit clearer.

B

So I don't know whether that means we need to take a vote or whether we just say please work to try and turn these all into one thing, ideally under smi.

F

Well, I think we should yeah. We should ask if that makes sense or another alignment of them as as one these two as one project or something I mean, I think we should definitely get an opinion from smi and uh a kind of set of options that we could pick from our.

B

Yeah makes sense.

F

B

To be on another thing, um so if I drop off, ah there are several we'll.

A

Still have quorum and that's fine.

B

Yes, um okay, I will I'm gonna mute myself. I'm gonna drop in quickly what I think my votes would be for the other things. If that makes sense or actually is there anything? I was worried about.

B

We have actually.

A

No just go go ahead without me,.

B

uh All right, pixie's awesome before considering.

H

They genuinely want to donate.

B

It it's awesome.

B

Oh no, I do have a. There was some there's some issue about dependency in pixi. Yes, now I know for pixie.

B

It is not clear to me whether this is just the client parts and whether they're also so it's essentially um you know, observability using ebpf to collect metrics and observability data, which is awesome, but it's not clear to me whether this is just the client stuff and it's dependent on a server.

B

um You know component that may not be included that wasn't clear to me: does that make sense.

H

Did they make presentation to sigs um and if not sure we asked them to, would it help us to understand the project better.

K

Yeah seek observability.

I

Yeah, from my perspective, yeah.

B

That would be a good idea actually because it's pretty cool, but it would be interesting to understand how it how the components really fit all right. I really really need to drop off sorry take care everyone.

A

All right, I am marking pixie as the uh presentation requested, so we can move back to fluid.

A

Fluorescent fluid.

F

Yeah, this was the other big data one. So.

H

um It's an academia project as well and it seems to be pretty active a lot of.

F

Contributors uh yeah has a lot of a lot of contributors, yeah.

G

It's definitely a collaboration too alina. uh I remember talking to the person who submitted this proposal. I know him from ibm days and it looks like they are doing it because they are facing some issues and they came up with the solution to so.

L

There are other end users with similar requirements for this sort of.

G

G

The one thing that I couldn't figure out here was like, which sig would it fall under like we were talking before other than that? I'm a plus one for.

G

L

What would this idea of uh machine learning one cover? Also like data analytics um yeah, I think so I think would you want to because then then this would be a good thing.

G

If, if there was one question I had here, it would be like they're using a luxia which is not really cncf, so if they they, if they need alexio and can't be replaced, um you know how much of if alexa changes license. Then what happens to us right? That's a thing that I can think of right now, given all the agpl stuff going on.

I

The risk for a lot of projects- I wouldn't hold it against him,.

F

F

We've already got the issue with grafana.

F

I can see chris smiling, he enjoys this thing.

I

No, I spoke with them this morning. It was lovely.

G

Call for a vote on this then.

A

A

All right, our next one is submariner.

I

It's a rancher project or I'm confusing with something else.

H

It's a very big flashback. Yesterday, uh when I looked at the.

J

Words, it was, I think, now we have more red hat much more red hat.

E

J

E

Rancher um and then it kind of evolved.

J

Okay cool, so so the good thing is, we have a community, that's good.

F

Yeah, it seems pretty.

F

G

Yeah I like the fact that there is a good interaction with the kubernetes sig multi cluster. So I'm happy with that.

E

I agree, I think they've been very good about taking in kind of requirements and driving it from that perspective, to to support multi-clusters so and not just networking per.

E

G

Call for a word, then,.

A

All right, the vote is open.

A

And our very last one is anthria. I think I'm saying it right.

K

Did we close on pixie.

F

Liz wanted more details, so we should just say: we'll: do a presentation.

A

I

A

I

And what they're contributing yep.

G

Andrea is from my employer.

A

Thank you for the pronunciation correction appreciated.

E

How was it pronounced? I didn't hear the right one.

G

I don't know, come on.

E

Do you have all the answers.

G

I

It's an open v switch thing: okay, yep.

E

I think our next internal project should be a naming thing that just random generator of project.

G

Names, they have a slack channel on the kubernetes slack and they seem to be recruiting folks to do like uh microsoft, windows stuff as well.

G

And we are looking at like telco use cases and things.

G

F

But it's been around for a few years since 2019 seems to have got a lot of country, which is every time.

G

The thing I like about this project is the leads are like deep into this. They don't have. They are not open source background, like you know some of us, but they really want to do the right thing and, like it's really good, to see somebody who is putting in effort by like looking at other projects what they do and trying to adopt it and things like.

G

G

The only question I would have is like how many of cni projects would we end up with right, um but, given that some of the projects are uh not doing so well, that would be the thing to look at here.

I

Yeah, I see this as mostly supporting a lot of telcos, essentially that.

D

I

An open v switch as kind of primary use case, so helping them adopt cloud native technology.

J

Yeah, I think it's good. It's a you know it's kind of a little bit like the cube ovm project, which I think we also admitted into sandbox. I'd like definitely like to see more experimentation around. You know low-level networking for kubernetes there. There hasn't been a whole lot of projects in that area say even compare with storage. So that's that's kind of good to see.

F

We're actually going to have a whole bunch more kubernetes networking project projects than a while ago, fairly soon.

G

We need a separate foundation for networking.

G

Projects: okay, call.

G

A

All right that wraps us up for today, um unless anybody else has things they want to be able to put on the uh public recording for.

A

A