 So, once more, hello everyone, good morning, afternoon, evening, whatever there is in your time zone, so thank you for getting up early or staying up late to join the Edge Computing Group discussions. Today is day one at the PTG but it will be day zero for our Edge conversations. So, before we dive into the sessions, do we want to do a round table introduction since it's not a lot of people or do you all just want to deep dive into the sessions because everybody's so excited. Well, I'd like, I'll go if you recall we discussed about sort of laying out the organization of the three days I think that might be worthwhile, because not everybody on this call is, is familiar with the, with what we're doing so that the laying out of the, you know, day one day to day three, or day zero day one, day two, I should say. Yeah, that's, that will, that will be the next, if we want a round, quick round table intro in terms of who's who on the call. Do we want to do that, or should we. I'll be happy to, you know, particularly since I knew there's a couple new people. So, I'll start Beth Cohen from Verizon been involved with the Edge working group for since its founding was one of the founders and primarily interested in at developing its use cases. I do a lot of what I've done with this group that take it back and developing products within Verizon. I will introduce myself to since I've been talking a lot. So my name is Eliko Vancha I work as senior manager of community and ecosystem at the open info foundation. And as part of my job, I'm community manager for the starting X project that we mentioned a bit earlier today it's an integrated cloud platform for edge and IOT use cases. And I'm also one of the co leaders of the edge computing group. And working a lot with adjacent communities such as etiquette state of the edge and some other ones in the edge and telecom space. Who wants to go next. Yeah, maybe. Yes, I can. Hi, my name is going to get your Terry I'm working in the open source program office of Nokia. And I'm also involved in the edge computing group, maybe not since it's foundation but from very early times, and I'm of course bringing in the, the telecom aspect of edge into the discussions. Get some new people about Nate. Sure. My name is Nate Johnston. I work at Red Hat. I'm a for neutron reviewer. I've been in the community for five years or so is on the TC for a little bit. I also now manage a team who that works focused on Octavia and designate. And I've not really, I've been aware of the edge discussions and watching them from a distance but what really attracted me to this was dimension in the notes of DNS. And I don't think the edge story for designate is very well formed at this point. And so I'm interested to listen and learn and see what I can help contribute back to the designate team from the edge perspective. That'd be great. Because we've been, we've been definitely struggling. A lot of people are not super knowledgeable about the whole, you know, network services side of the house. And so, yeah, it's definitely an issue that needs to be brought back into other components. Okay, maybe I can, I can go then. So I'm batis-jonglet. So I work at INRIGA in France as a research engineer. And I work in the team of Fabriann Lab. And basically, I just arrived in the group. I think I've been to a few meetings, but there's no big meetings basically. And so I, for now I'm getting to, I'm trying to get an overview of the edge activity in the group to better understand what's happening there and what we can contribute as a research institute. And I have a bag one in networking and in, so networking mostly on Linux and on research infrastructure and research platforms to do all kind of network and system experiments. And that's it. Sounds great. Thanks. Adrian, maybe you want to go next. So, hi everybody. My name is Adrian Labre, professor at IMT Atlantic, head of the Stack Research Group. So the group that is just mentioned. Mainly we are doing research and development activity on topics such as for edge computing, computing continuum from the cloud to the IoT. My goal is actually to try to design the next resource management system to operate this computing continuum. We are mainly eight permanent and between 10 and 15 non-permanent staff all working on those questions for the IoT edge computing. That's all. Thanks. Thank you. And go next. Everybody I'm Rob Hirschfeld with Racken. And, boy, Beth, I certainly don't have your tenure on this in this group, but I have been in and out of it for many years. And I'll serve some time on the LF edge, the LF Edge organization. I've been doing edge and edge related work for years and years. And I've been in the OpenStack community for a long time before that, too. My interests here are very much around operations and distributed operations for Edge. And so a lot of the things that I enjoy bringing into the group and discussing in the group are related to the automation and operations pieces. That touches on everything else too, but those are usually my focus areas. Oh, and bare metal. I do a ton of bare metal stuff. I'll take a smile. All right. No, I just think that that bare metal will become really interesting for Edge. It's one of the things that we have to get right. Yeah, we really do. Okay. Pankaj. Pankaj, I'm based out of Colorado. Sorry, I don't have my camera on. Oh, that's good. Disconnect one of the monitors and switch on my laptop. Too lazy for this early morning. I met Microsoft. I met the team that was sold from AT&T to Microsoft last year. I'm basically engaged with creating the network cloud or the recently re-labeled Azure for operators, distributors, services. I have been with this group for a while, but my main focus was from Aniket. Thank you. Oh, and I saw Karine joined. We are doing quick round table intro. Yes. Hello, everyone. Sorry for my late arrival at another meeting just before. So I'm working with Orange and my main focus is also my main attention is with the Aniket project at this moment. I'm interested on the OpenStack and Azure topics. So I don't know. I'm just joining, so I don't know if you want to be more introduction or not, if it's efficient. It's fine. It's good. I just would like to encourage everyone that I know that at least from from previous PTGs and other sessions. People are mostly joining with the intention to listen and learn. I think that we really arrived to the time when, even if you don't have a hands on experience with something, you still might have a question or a comment to add and just would like to encourage everyone to please speak up and grab the mic. If you have any time or if you don't want to interrupt the discussions then zoom has a function to raise hands and show in all kinds of ways that you would like to chime in as well. I will do my best to pay attention to those as we are going through the meeting. I also just shared the etherpad that we prepared for the event. So, you can have taking notes or you can just open it so that you're able to follow what's happening a bit more closely. And when it comes to one more logistics item. We will have a bit over an hour and a half left for today. We will see if we get to the point of where everyone needs a five minute break, or if we can, or if we have really interesting discussions and we will just kind of fly through the remaining amount of time. If more of you feel like that you need a short break, please again raise hands or just chime in at some point so we can maybe coordinate on that one as well. And with that, Beth, if you don't mind, I think it would be great if you could start kicking off the discussions with talking a little bit about the concept and the scope. I will ask others to join in who have been working in an area where this day zero day one and day two slash and concept is a thing. Right. So, Nate, this is a directly in in your interest. So the, what we did this, this time is we focus the discussions around operations and service delivery and this comes out of my experience in actually building a product, an edge product which is virtual network services. And we have the app horizon and we shipped it, what, five years ago now. But you know it has been attorney. Let me say that. And the concept is day zero. The first is is sort of the templates and the automation and sort of the standard features that that can be kind of predefined. And there's a whole lot of things that go into it, particularly around edge, you know, in the data center, you know, basically you just have those ISOs and you have, you know, you have the additional complexity of getting it out to the edge, which means it's on a possibly unstable or unreliable network connection. And so you need to, to, you know, bootstrapping is a whole lot harder. And so day one, sorry, today, we're going to focus on day zero, which is, you know, how to how to build those templates what needs to be in those templates. And, you know, who's responsible for creating them, maintaining them there's there's a whole lot of stuff that goes into sort of the day zero stuff and I know Rob, your, you know, neck deep in this stuff. And then the tomorrow we're going to be talking about day one and day one is really around this within Verizon we call it service delivery. Day zero and Verizon's terminology is product development. Day one is service delivery, i.e. customers ordered something, we need to actually deliver it to the customer's edge or deliver the service, whatever. And what that means is you take the templates that come from day zero and you inject specific information, let's say an IP address, or, you know, something about the customer's network connection or, you know, specific to that to that user that end node, you inject that and then you get it out to the site. There's a ton of additional information that needs to go into it, you know, particularly when you get involved in licensing, which means you need to have, you know, some way, you know, typically licenses are managed centrally, but you know obviously each site has to have a unique license you know how does it check things to make sure that the license is the right license is being delivered, etc. So, and then you know how do you validate that the hardware at the other end can support what has been ordered. This is actually a big problem. Again, because at the edge you don't necessarily have additional resources that you can just, we'll just add a couple, a couple more. You know, you can see pews and, you know, and, you know, use up more resources. Well there is no more resources to use up. So anyhow, so service deployment and delivery is kind of the focus of tomorrow. And then the third day is sliding down. So this is once once you've delivered the service and you've turned it on you've turned it over to operations. There's a whole lot of things that need to be taken care of, once you're in operating mode. And, you know, just, just I use a simple example that we've run into, you know, we've, we've turned up the service customers happy we're running it and then there says oh here's the new update of the service and we need to push it out to the edge. Well, how do you do that, you know, because you're in operations and, you know, you can't just say to the customer hey, we're taking the service down for a couple hours while we do an upgrade. That doesn't fly. How do you access it if the network connections fail. How do you maintain your licenses right a lot of the vendors have a, you know, a check in mechanism you know has to check in once a month or once a week or once a day or whatever. To make sure that you know it's still a valid license and if you lose that connection. It's sometimes it reverts down to a lower level of service, which with nasty nasty surprises for customers. So anyhow that's the focus on the third day and and the thread through all three of these is driving automation and and manage and manageability and lifecycle management, you know, over, you know, through the entire lifecycle of all of these applications that are at the edge. So, hopefully it didn't go on too long. Any questions. That makes sense and I've been involved in exit delivery and administration at previous jobs, Comcast and America online so this, this all feels very, very normal and straightforward to make sense. If you've done, if you've done Comcast, yeah. Verizon does the same thing. Any other comments or additional thoughts to the concept before we dive into day zero. Zero let's go. Let's go. Okay. So as Beth mentioned, templates play or play an important role. In case of day zero preparation steps. So we prepared a couple of questions. As we were planning for the PTG that you can see on the etherpad and beyond templates we were also touching on things like how to do automation from early on, like, even before things starting to happen. So if you have the right level of automation to support day zero, then touched on licensing issues that could be a topic to discuss, and also platforms that you can choose to deliver your use case and applications with any high availability is appearing on the etherpad as that's also a pretty important question or a topic to decide before you start to roll things out because high availability setups are different. They're tricky and complicated. Yeah, that's my experience with them. Okay. So, especially those of you with experience. Do you have a favorite question to dive in should we start with templates. I'd like to start with. Yeah templates is good. Okay, let me actually let me share some of my personal experience with developing the product and and sort of some of the road, the bumps in the road. Oh yeah, trying to get the cut of the it started. So, we, we started with the idea of creating a product that could be delivered on the edge, as well as on a hosted core platform. So we in fact use open stack. And the idea was to create, we use templates in fact, and with workers with an orchestration platform to to deliver those templates. And we quickly ran into issues with how standard should those templates be and how many, you know, the idea being that, you know, do you do you have like, six, six standard templates and then you kind of add a bunch of variables when you do the delivery, or do you have a couple hundred standard templates. And over time we ended up with, we now have a couple hundred standard templates and the reason is is because the images are actually there's not a whole lot of them. But the templates go into details about the network connectivity. So it turns out that the way we architected it and maybe that wasn't the right way but it is the way we're stuck with. So we have like if, if let's say we have a MPLS connection we have a wireless connect you know LTE or 5G connection actually LTE and 5G separate templates. So, you know, the network templates drive the templates or the network component and the combination that's in that drives the templates which actually causes us major pain because, you know, somebody will come in and say, I want to do this architecture. And then we have to run around and create a new template. Yeah, each new template takes, you know, a couple hours to do. It's not a big deal, but as I said, we now have hundreds of them. So, other experience. Is this the right way to do it. Are you're the templates that you're using. Are they messing with my sound sorry. Do I sound okay. Yeah, you sound fine. All right, I was getting too much too much monitor back on my side. So do you version control all those like are they set up in a bundle and then site site specific. They're not version controlled actually it's just a library. It's just a catalog. So, you know if one goes out of date we just throw it away. And then back up. I'm trying I'm trying to think I have questions for you I also have what what sort of what we've been helping people do and our approach to it. I'm happy to share that if that's helpful because this is, it's hugely painful to go site to site. And then also know what. Just to back back in the old days we used to have a complete automated system and bundle everything. But what we found was that if we change the automation on on site to be to fix something. It was really hard to know if to bring back the change and then roll it out to other people. And maybe this is different this is automation it's different than templates. Although for us that they've started to cross into the same thing. Yeah, the automation actually it doesn't change as much it's the templates that I mean it's the same automation just using an orchestrator. Our experience has been that the that is if the templates are the configuration that you're using as input. The changes to those templates also have to be respected by the automation or the automation has to be able to pick up the changes to those templates. Yeah, but we're not doing it that way we're just saying, during the order process, we take. We, you know, based on what the order says it gets converted into Oh, you want this template. And it just the template gets injected into the automation. How do you how do you keep. So, what we've seen is that as the automation changes the information the templates can can need to be updated and so there's a vert there's a synchronization issue where especially past templates when you're talking about a lot of sites or a lot of deployments that you know you're doing all the things all the time and so you have a case where the information you're providing to one site or a collection of sites has to be changed in relation you know it has to be changed in relationship to the automation that's behind it or the systems that are behind it. And for this we call it a content pack and you can version individual content packs and then you can create dependencies between content packs and then a content pack and have templates and other things but all of the pieces that you need to make something go because that can include because this can include binaries as a as a requirement. Right now we're talking day zero so this is like hey I'm building a site I need to know all the stuff that goes into that site that could include a set of, you know, configuration templates it could include the automation for it it could include the DHCP configuration for the system it could include binaries that you need to have loaded license files I know we're going to talk about that. Which is which is fine site site to site but then what we see is it once that's, you know, you start replicating that pattern and it's going to. It's the next site's going to change and the next site's going to change the next site's going to change. How do you isolate the differences between sites like do you have a way to break out some dependency some inheritance graph or some reuse components. Well, remember the site. So we're, it's day zero. So this is not, it's not. There's nothing unique about these templates. Don't you have like. Each site has some configuration about what you know what its name is what it's IP ranges are. That's injected in day one. That those are all variables that are injected as we, as we do the service deployment. And do your, the, how do the templates accommodate so you're starting day zero with these templates they expect to have injection of this is good, good. Yeah, yeah, yeah, yeah. That's I think I kind of mentioned that in the line where she's how standard are they and how much they can rely on variables to support day one deployment activities. I was talking about that. I see. So they're vanilla. The idea is that when the, when the solutions architects are architecting a customer's deployment, they're going down and saying, Oh, this, this is, this is site type acts that has these, these types of network connections and the and these services. Okay, so. I don't know what a blueprint. Okay, it's exactly what we call them actually we call them blueprints. Okay. So can that go all the way back to some of the reference architectures that are getting built like would have, would have blueprint here reference. I don't know if it's a Prano or some other standard, you know, some other blueprint system what does that look like for y'all. Well we built it ourselves but that was because we built it five years ago so. Yeah, so we use T templates but there's nothing that says, I don't know what's a great a Prano is based on itself or it's got its own YAML. Leave. I don't, I don't know enough about a crane. Yeah, I'm not, I mean the challenge is anything that you do this so when you build a template like this is it looking to be an actionable like this gets fed into the automation I'm assuming. Yes, yes, totally. That's why you're saying he, right. That's why we use he doesn't have to be he just that's happens to be our system. Right. It needs to be an automation driven as to go back into the automation. I think a quick interruption, a little bit of a technical issue with zoom. So it will most probably disconnect in less than a minute. So just join back to the room and after that it should work. We will see may not draw but I think it will. So if zoom drops just please connect back to the same room and hopefully it will work. So I think about that. It's just, you know, how Monday mornings are going. It is what it is. Okay. I know.