GC - #36 Kubernetes at the Edge

00:00:00: Giant Conversations, episode thirty-six.

00:00:04: Today we are blessed with the presence of Antonia and Javier who have just come back on a tour talking edge to cloud traveling many different cities And giving their perspective

00:00:21: On

00:00:22: what they've been upto.

00:00:44: I would like thank Jim and i for that song And my fifteen minutes of due diligence on asking for less saxophone.

00:00:53: So hi Javier, how are you today?

00:00:56: Hi

00:00:56: Tommy.

00:00:57: I'm good Thank You.

00:00:58: thank you for having us in the podcast.

00:01:00: yeah we appreciate being here Antonia you

00:01:03: Yeah doing great

00:01:04: excellent all right well We're gonna start off with a couple questions but i'm mostly interested at The very beginning Here to understand.

00:01:13: Kind Of an introduction

00:01:15: To

00:01:16: Giant Swarm and edge computing.

00:01:19: smart factories distributed Kubernetes, you know the whole nine

00:01:23: lay it out.

00:01:24: Yeah So we went to different cities to give talks around.

00:01:30: You know Edge Computing in smart fact smart factories because that's what Jen swarm has been doing in the last couple years.

00:01:37: We've seen more and more interest coming inside Around the area of smart smart factories and smart manufacturing about making plants more efficient and using communities like frameworks to achieve that with distributed architecture.

00:01:57: And this is what we wanted to talk about.

00:02:00: That's great, so on your tour you went onto a lot of different cities.

00:02:06: I remember when you went to Edinburgh or Ahus up in Denmark Berlin Atlanta kind of my neck at the woods, interested in what your favorite cities were and why.

00:02:25: So that was Hamburg I think?

00:02:30: Well actually haven't been to Berlin although i wish we did but yeah...I think my favourite city so far has been Edinburgh because it just loved this city.

00:02:42: It's beautiful and very historic.

00:02:46: Antonio, what about you?

00:02:48: It's kind of hard to say.

00:02:49: There were many nice places I... Of course i was excited to go to Atlanta because I really liked the Netflix series called Atlanta.

00:02:57: so I expected everything there just like The Netflix Series and it was a bit different but I did enjoy the lunch.

00:03:05: So that was great.

00:03:06: When y'all we're giving your different talks What is one thing that surprised people most when they first heard about what Cube Edge can do?

00:03:17: So when we give a talk, there's always like... A section at the end that is saved for questions from the audience.

00:03:26: And we had quite a few questions around reliability of the product because- Of the projects I was talking about QBH here Because they're all a few extra components compared to core Kubernetes and how it works in connectivity.

00:03:42: so people had concerns and legitimate questions around how it would work, if something breaks.

00:03:49: How does it break over?

00:03:51: We had quite a lot of interest around how data synchronizes between the cluster in the nodes when it goes offline.

00:04:01: And yeah there was also I think part.

00:04:06: The draw to these kind of technologies is the memory footprint or resources footprint rather, which allows you to run Kubernetes on smaller nodes like Raspberry Pi.

00:04:18: So we often also had enquiries around how does that work and how do you save resources?

00:04:32: With... That's interesting!

00:04:34: with the resource footprint.

00:04:36: Was that a surprise to you when you started putting this together or was it kind of a known goal?

00:04:43: To lower your, your footprint?

00:04:46: Yeah I think generally with Edge.

00:04:48: uh The Goal is uh That among other things Uh that-that You do have A smaller footprint.

00:04:55: um And for example you don't use A full cupelet But like a simplified version Of a cupelet.

00:05:03: And I think also some of us have experienced with K-three S, which is often used in that field.

00:05:08: So i guess it wasn't surprising but It was still cool to see basically how space efficient it Is.

00:05:20: The overall results now what?

00:05:21: Is it a comparison too?

00:05:23: when you say here for example the resource footprint Was um seventy megabytes What would be That if You didn't go this route.

00:05:35: So in the case of cube edge or K-threads, you have one binary that runs all your Kubernetes components.

00:05:42: That schedules your workload and everything.

00:05:45: so that's pretty tiny.

00:05:47: if you take a normal regular nodes we are talking about proper kubelet And this thing could take I don't know maybe a gigabyte something According to how heavily used the node is, there's a significant difference.

00:06:08: Theoretically you could probably run a Qubelet on these small devices but it would be huge waste of resources and then you wouldn't have lots left for your actual workloads that create business value.

00:06:20: Wow okay!

00:06:21: That is a significant different.

00:06:26: I had only heard about Cube Edge once i started watching.

00:06:29: I watched a lot of your talk in Warsaw, and I did see some of the great questions afterward.

00:06:34: that's kind of by... Once I got lost I fast-forwarded it to see.

00:06:39: um ...the interest but lots of people there.

00:06:41: they really enjoyed your

00:06:42: talk.

00:06:43: And remember in Warsaw someone asked about running at an chicken farm?

00:06:46: Does does that really stayed on my mind?

00:06:50: yeah she had asked that!

00:06:52: She was like.

00:06:52: this is real use case here.

00:06:54: People kind of in simple terms for a guy like me.

00:07:02: What is QBedge and if you can give me real-world example what problem it would solve?

00:07:10: So, QBedge is an open source project which used to basically connect Kubernetes cluster with Edge devices.

00:07:25: If you have edge nodes like Edge Machines, or these can join your Kubernetes cluster as nodes basically.

00:07:32: So it's just going to look at your Kubernetes Cluster and all the nodes will look normal.

00:07:37: but some of this are going to be Edge machines basically And then what Kube-Edge allows us is run workloads on these.

00:07:47: basically That was one part of Cube Edge is that you can also manage basically devices, like some IoT devices inside your cluster as custom resources.

00:08:04: So you can use Kubernetes to manage your IoT devices.

00:08:09: essentially

00:08:10: Can we dig a little bit deeper on the kinds of companies who come to us and need these services and how we work with them in what we do for them.

00:08:25: Actually, lots of different types of companies that use edge computing technologies.

00:08:29: but the case of JN Swarm like I mentioned the intro it's mostly smart manufacturing.

00:08:38: so often they will have a factory flow like legacy processes, let's call them that with workers and factory floor workers who will do actions manually.

00:08:52: And bringing in something like edge computing you are bringing all the benefits from running Kubernetes in a traditional environment at the edge, so you can schedule your workloads everywhere in the plants.

00:09:07: You will have like a small cluster or something and these clusters can then collect data from their cameras or sensors and all this things um... And that way they can get more accurate readings faster iteration, they get more consistent because sometimes people can forget to do things or meet repos and the robot doesn't do that.

00:09:32: They just do things

00:09:33: consistently.".

00:09:34: That

00:09:34: kind of answers my next question but I would like to know once it's up and running how does it actually change for the people on the ground?

00:09:45: Like people who are working in a factory How would they notice a difference once this is installed and running?

00:09:55: Well, I think it depends on what actually working.

00:09:59: Ideally you will receive benefits from having cloud-native technology in your factory so that could be related to the way you analyze data or processes are automated.

00:10:15: but then really big pro You would still have all these benefits even when you're not online.

00:10:24: So like a technology, like Cube Edge basically allows you to kind of run these things separately.

00:10:32: Even if your network isn't available it can still basically Have all this benefit And I think that is probably the biggest difference That you would see If you are working inside a factory or doing something else where edge computing's used

00:10:48: When you say if you don't have access to the network or is that because it's on-premise?

00:10:55: Yeah, so I think this a really big one also.

00:10:58: That companies actually ask us to help them out with... ...that they either intermittently lose connectivity.

00:11:07: Or in some cases like more extreme cases we had companies asking about air gap where they do not want connectivity to the internet or other connections depending on their use cases,

00:11:22: right?

00:11:24: Is that based on privacy?

00:11:25: I'm guessing.

00:11:27: Yeah there's lots of like...I think the biggest consumer of air gap environment is around a defense sector where it's security related.

00:11:38: so they are going to run down stuff in machines ships planes whatever and these things need to be completely autonomous.

00:11:49: For an organization, manufacturing healthcare defense who are convinced that this is the direction they needed to go?

00:11:56: Why does it matter who helps them get there?

00:11:58: and okay... This is the marketing side.

00:12:00: I don't know if anybody's not interested anymore to hear the spiel.

00:12:06: Yeah like To be honest i think all the companies go through the same journey when before getting to the stage where they took to Gen Swarm.

00:12:16: So, they have a problem statement and then in options each option has different costs so lower cost often appears to be DIYs.

00:12:27: you do it yourself.

00:12:29: any company could figure all of these things out by themselves.

00:12:33: It's an open source, available online, free.

00:12:37: mostly You could do it all yourself, but it takes quite a bit of time.

00:12:42: It takes money and people.

00:12:44: so you need to hire teams that's wages in everything And take some time on board and gain experience and build the products.

00:12:54: Once you've built your product, run it, update or maintain it.

00:12:59: That is where you don't allocate resources for business value.

00:13:02: So that makes sense.

00:13:07: Just hire a company.

00:13:08: They get up and running in couple weeks, they take care of everything And then like that's kind of the hack A shortcut to production

00:13:20: Back-to being on stage.

00:13:22: any anything else?

00:13:24: On the docket you have other stages to grace.

00:13:29: I think we've done with this talk.

00:13:31: We gave it quite few times.

00:13:36: I think maybe people now want to see something else.

00:13:40: Yeah, they get it Well cool well i guess we've.

00:13:44: um We've squeezed as much water out of this rock as we can.

00:13:49: So Xavier and Antonia thank you for joining us today To talk about edge-to-cloud.