Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Solving Multicloud with Seamless Connectivity and AI - with Rob Croteau
1.4k Plays1 year ago
In this episode of KubernetesBytes, Bhavin and Ryan interview Rob Croteau of Avesha. Avesha is behind the open source project KubeSlice which aims to enable admins to seamlessly connect multiple Kubernetes clusters no matter where they are located with a few command or clicks. Learn about the challenges teams face today with networking and how Kubeslice and avesha are trying to make connectivity for clusters simple.
Join the Kubernetes Bytes slack using: https://bit.ly/k8sbytes
Try Nom Nom today, go to https://trynom.com/kubernetesbytes and get 50% off your first order plus free shipping.
Ready to shop better hydration, use code "kubernetesbytes" at this special link https://zen.ai/apaSnaIFOuee5jScqZ28a03tKKvQiqkyz8mtm9wipoE to save 20% off anything you order.
Time Stamps
- News 00:05:36
- Interview 00:14:05
- Takeaways 00:53:50
Show Links
- - https://kubeslice.io/
- - https://avesha.io/
- - https://github.com/kubeslice
Show Notes:
- - https://cloud.google.com/blog/products/ai-machine-learning/duet-ai-in-google-cloud-preview
- - https://cloud.google.com/blog/products/containers-kubernetes/whats-new-with-gke-at-google-cloud-next
- - Kubecost Cloud now GA - install agent using Helm chart https://siliconangle.com/2023/08/21/kubecost-debuts-kubecost-cloud-help-enterprises-rein-kubernetes-spending/
- - KEDA graduates - https://www.prnewswire.com/news-releases/cloud-native-computing-foundation-announces-graduation-of-kubernetes-autoscaler-keda-301907019.html
- - Netapp files in google https://www.techtarget.com/searchstorage/news/366550341/NetApp-cloud-storage-evolving-for-stateful-K8s-AI
- - GKE Enterprise https://techcrunch.com/2023/08/29/google-introduces-gke-enterprise-to-help-companies-manage-complex-kubernetes-environments/
- - API GW article https://thenewstack.io/the-api-gateway-and-the-future-of-cloud-native-applications/
- - Ngrok static domains https://ngrok.com/blog-post/free-static-domains-ngrok-users
Recommended
Transcript
Finding the Right Dog Food
00:00:00
Speaker
If you've ever had a puppy and raised it to become a big dog, you know that changing food and finding the right food is hard to get right. Ultimately, you want them to feel good and act happy and be okay with what they're eating. They're part of your family, after all. I have an eight-year-old golden retriever named Roscoe, and he's always had a sensitive stomach, so finding the right food is kind of a pain. That's where Nom Nom comes in.
00:00:24
Speaker
Nom Nom's food is full of fresh protein that your dog loves, and the vitamins and nutrients they need to thrive. You can actually see proteins and vegetables like beef, chicken, pork, peas, carrots, kale, and more in the ingredients.
00:00:39
Speaker
So here's how it works. You tell them about your puppy, the age, breed, weight, allergies, protein preferences, chicken, pork, beef, and they'll tailor a specific amount of individually packaged Nom Nom meals and send them straight to you. If you're ready to make the switch to fresh, order Nom Nom today and go to https forward slash forward slash trinom.com slash Kubernetes Bites.
00:01:03
Speaker
and get your 50% off of your first order, plus free shipping.
Introducing Kubernetes Bites Podcast
00:01:09
Speaker
Plus, Nom Nom comes with a money back guarantee. If your dog's tail isn't wagging within 30 days, Nom Nom will refund your first order. No fillers, no nonsense, just Nom Nom.
00:01:25
Speaker
You are listening to Kubernetes Bites, a podcast bringing you the latest from the world of cloud native data management. My name is Ryan Walner and I'm joined by Bob and Shaw coming to you from Boston, Massachusetts. We'll be sharing our thoughts on recent cloud native news and talking to industry experts about their experiences and challenges managing the wealth of data in today's cloud native ecosystem.
00:01:52
Speaker
Good morning, good afternoon, and good evening wherever you are. We're coming to you from Boston, Massachusetts. Today is August 31st, 2023. I hope everyone is doing well and staying safe. Let's dive into it. I know it's still August. It's the last day of August.
00:02:09
Speaker
And it feels kind of like fall outside. I know the temperature has definitely gone down, which this is perfect temperature for me. Like I don't want it in the 90s and hundreds. I don't want it in the 40s. So like, as long as it's between that 50 and 70 range, perfect for me.
00:02:24
Speaker
yeah if it could be like this all year round i don't might be heavy pumped i mean i know it's still like 115 in phoenix right now so like you know i feel for those people but uh in texas but hey you know we chose three feet of snow in the winter you chose that
00:02:41
Speaker
You know, we all have our grapes, but what have you what have you been up to this past weekend? I don't know. I've been to a conference in Vegas. VMware explored last week. So I was out there. Yeah. How was that? It was good. I didn't go in with a lot of expectations. Like, personally, I feel that that audience still is a lot of like.
00:03:00
Speaker
legacy virtualization admins that want to talk about running VM. But I was surprised. I did have more conversations than I expected at that show about Kubernetes and about data on Kubernetes and things like that. Overall, a good show. Where was that? Is it Vegas? Yeah, it was. Another Vegas trip. Two into one, I think. I'm not going to be at reInvent this year, which I am not terribly mad about.
00:03:28
Speaker
I think I'm somewhere there too. Maybe I'm not on the list, but music is still far out. I mean, people love Vegas and power to them, but it's not my jam. I know. It's not my jam. If I just venture outside of Vegas and you get me into Grand Canyon and that kind of stuff, that's my jam. Not in summer though, right? Like 150 degrees. I've been there in summer. I've been there when it's 150 degrees. Hiking in the Grand Canyon, actually. Not fun.
00:03:54
Speaker
You know, as long as you make sure to drink water and now soda, you won't have to get helicopter now. Um, speaking, speaking about the friends experience. Um, anyway, so I know I texted you over the weekend cause I was in Chinatown. I know. Come on, dude. Like give me some heads up.
00:04:10
Speaker
Well, I was in there, and I was like, oh, I should let Baba know I'm here, because a buddy of mine, we were going to some, like, whiskeys of the world. Oh, nice. That's a continental. And, which was great. And I'm like, I like bourbon and stuff like that, but it's not like my thing. But I did find one I really enjoyed. But we went in the day because it was beautiful. Yeah. And I haven't been to Chinatown in a while, so I was like, I got to go to a few of these places. Get up some, get some ramen. And he hadn't been there either. Yeah, yeah. Oh, we didn't go ramen. We went dumplings. Oh, OK.
00:04:39
Speaker
This place is like Windsor Dumpling in Boston. You sit at shared tables and you just put check boxes with the stuff you want with dangerous, by the way, because you just keep it coming. And then we were walking towards the north end and I was like, I wonder if Bob is around. But yeah, that was nice. It was a beautiful day out.
00:04:58
Speaker
I couldn't complain, which is a pretty good weekend. And we have a long weekend coming up, so perfect. We do, we do. Do you have off tomorrow too, or just the Monday? Yes, the Monday. My father has off like tomorrow and Monday and Tuesday. I was like, I need to get your manager. I know, that's awesome. No gripe to my manager if you're listening out there, by the way.
Cube Slice and Networking in Kubernetes
00:05:20
Speaker
Or maybe this is the push he needs, right? All right. Well, so we have a really cool topic today. We're going to be talking about Cube Slice with Rob. We'll give you a little more about him. But before we dive into that networking world, let's do a little bit of news. Bob and I want you to kick us off. Yeah, sure. So Google Cloud Next is this week, right? And they did, following their Google I-O announcement, the first keynote was all AI, AI, AI, AI. I think they also have a session in the catalog with Justin.
00:05:50
Speaker
But they introduced something called do it AI in Google Cloud. And it definitely has a component where it works with Google workspaces and helps you do Google Sheets and Google Docs better. But then from a Google Cloud perspective and
00:06:03
Speaker
the app modernization space that we work with usually, it can help you modernize your applications, right? So you can just tell it like, okay, this is my C++ app or this is my C++ function. Please rewrite it in Go and make sure that instead of using a local DB, I use Cloud SQL. Again, this is trying to move applications to the other cloud, but it just does all of that and rewrites your code, gives you a new Go app or Go function.
00:06:28
Speaker
Uh, and you can start using cloud databases. So I think that was pretty cool. Like the demo that I saw during the keynote. So, uh, I just wanted to share that scary though. Is that replacing someone's job? It is, but it's borderline replacing someone's job. Like you still need people who can understand go to verify that what AI has written is actually correct and what you want to do. So that still needs to be barriers. And I don't know.
00:06:52
Speaker
of tests written and made sure that the functionality is the same. But at least it removes some of the heavy lifting. I feel like this is one of those turtles all the way down. This is when we're going to have AI verifying the AI output. And then it keeps turtles. That's it. Some AI has to keep the other AI honest.
00:07:11
Speaker
Awesome. Well, that's cool. That's good stuff. And it can also help developers who don't want to spend time or learn to write complex SQL queries. They keep more done. Yep, that's another way to look at it. Yeah. They can just ask in natural language. Just generate a query that gives me x, y, and z from this table and join that table with it. And then it just gives everything in a proper query format. So that was cool, too. That's fair. Yeah.
00:07:35
Speaker
So those were a couple of announcements I know. And then moving on, I think I saw a cube cost cloud. And we have spoken to the cube cost guys. We have spoken about how cube cost cloud was in beta. Now it's GA, finally. So you can now install it using a simple Helm chart on your Kubernetes clusters, EKS, GKE, EKS, I think, and get everything integrated into one unified portal and manage your costs from that single pane of glass. So that's a cool thing to check out.
00:08:04
Speaker
They do have a free trial. And then KEDA, when we did the whole serverless approach, I spent some time researching more about KEDA and understanding how the event-driven autoscaler works. That project inside CNCF has now moved to a graduated state. So it's now a graduated CNCF project for people who are keeping track at home, I guess, of CNCF projects. But yeah, that's a quick news section for me.
00:08:29
Speaker
Awesome. I just have a few here. The first one is around what NetApp's up to in Google Cloud. So they have offered basically a first party managed file store service built into Google Cloud now. So I think this was announced at Google Cloud Next, I believe, to GA. So I think it's been around. But this is Google Cloud NetApp volumes.
00:08:58
Speaker
for some buzzword bingo there. But this is all about the network file system support across those platforms. And they have some other stuff in that announcement around using the right types of storage for a lot of these AI workloads.
00:09:15
Speaker
Reality is most announcements will try to mention the word AI at some point, but it's centered around that. Funny thing is every conference will try to get the NVIDIA CEO Jensen on the keynote stage. VMware did it last week. Google Cloud did it this week. Google is a popular guy. We need some AI, some NVIDIA in our keynote to boost our stock price up.
00:09:36
Speaker
Okay, one fun fact. Well, yeah, all this stuff is still powered by their on tap operating system. So all the things you'd normally get like snapshots and replication and stuff should be all familiar, but now it is effectively GA and available for...
00:09:52
Speaker
You to use a fun fact about that. Once again, I used to work at NetApp, right? So the NetApp CEO. Yeah, I did. The NetApp CEO and the Google Cloud CEO are actually twin brothers, like identical twin brothers. Identical twin brothers. That is funny. I didn't think I knew that. So just a fun fact. In your technology bingo.
00:10:15
Speaker
wherever you play at your next local brewery. You'll have that one in your pocket. Yeah. It should be a night. Successful family there. Yep. Okay. So the next one I had was around GKE Enterprise. So Google obviously launched Kubernetes way back in the day. I shouldn't say way back in the day, about 10 years ago or less.
00:10:37
Speaker
And they've had obviously their GKE offering for quite some time, but this newest announcement is really around GKE Enterprise, which builds on top of everything they've done with things like Anthos and GKE and everything. And it really is, I think, aimed at managing more complexity of multiple clusters, whether that's
00:10:57
Speaker
GKE clusters and Anthos clusters and things like that. So it ties in security and governance and all these things. So it's kind of a level up to Charizard, if you're into that kind of thing, leveling up their service. It was weird when the first line in that blog post was, oh, GKE enterprise is the premium edition of GKE. And my mind went to like, why isn't GKE the premium edition of GKE? Why isn't it premium? Well, we had to give the new thing a new name.
00:11:27
Speaker
So it is multi-cluster like they have a concept called fleets and you can manage multiple clusters easily and things like that. And we're going to talk about this hopefully today a little bit around and we have been talking about with previous guests and now I think the whole industry as entirety is moving towards those more complex how do you do things multi-cloud multi-cluster those kind of things so
00:11:48
Speaker
Not too surprising that they're centering in on this and along with other companies. But yeah, it's called GKE Enterprise. I've always been a fan of GKE stuff. So I'm sure it's some good stuff. We'll have all these links in the show notes if you want to go dig into it some more. I have a couple others that I haven't mentioned. I won't go into them. But one's an article about API gateways. If you're new to API gateways and what they're about and how they work, there's cool in there. And then one on ngrok static domains.
00:12:15
Speaker
You're a fan of in Grok like I am about publishing some stuff you're working on locally. You can have your own static domain now for free. So you can name it. I don't know what you can name it, but maybe something identifiable, which is nice. You can always go back to the same place and have randomly generated stuff. So go check those two articles out as well.
00:12:35
Speaker
All right, so today we have Rob Curteau from Visha. He kind of manages directive growth over there. And he's here to talk to us about Cube Slice, which is an open source project. They have surrounding, what do you know, multicloud, multicluster networking and connectivity. So without further ado, let's get Rob on the show.
00:12:59
Speaker
As long time listeners of the Kubernetes Bites podcast know, I like to visit different national parks and go on day hikes. As part of these hikes, it's always necessary to hydrate during and after its turn.
00:13:13
Speaker
This is where our next sponsor comes in, Liquid IV. I've been using Liquid IV since last year on all of my National Park trips because it's really easy to carry and I don't have to worry about buying and carrying Gatorade bottles with me. A single stick of Liquid IV in 16 ounces of water hydrates two times faster than water and has more electrolytes than ever.
00:13:37
Speaker
The best part is I can choose my own flavor. Personally, I like passion fruit, but they have 12 different options available. If you want to change the way you hydrate when you're outside, you can get 20% off when you go to liquidiv.com and use code KubernetesBytes at checkout. That's 20% off anything you order when you shop better hydration today using promo code KubernetesBytes at liquidiv.com.
00:14:04
Speaker
All right. Welcome to Kubernetes Bites, Rob. It's great to have you here.
CubeSlice: Features and Benefits
00:14:08
Speaker
We're going to talk all things Kubernetes and networking, hopefully. But before we jump into all those questions we have for you, please introduce yourself to everyone and what you do. Great. Thanks for having me. So my name is Rob Crutteau. I'm the director of enablement at Avisha. And I basically help our customers understand and maximize the potential of using our products for their Kubernetes ecosystems.
00:14:31
Speaker
Very, very cool. So you work for Avisha, and today we're going to talk a lot about sort of what CubeSlices is. So give us a brief introduction about what that project's all about and really what challenges are out there today around it.
00:14:46
Speaker
Yeah, so Kube Slice is one of our flagship products, and it simplifies cross-cluster communications, either in a hybrid environment across multiple clouds, or if you're in a single cloud across regions. So we're able to do that seamlessly. And the real nut of the product is the communication between those different clusters. It's done seamlessly. We can get that up and running in 10, 15 minutes, and you have that cluster connectivity.
00:15:13
Speaker
Very cool. Yeah. I know multi-cloud, multi-cluster are hot topics this year. I think there'll be hot topics going into 2024 as well. I think more people as they run Kubernetes are like, how do I do this more than once and in different places? More so than that. Like I've come across with a lot of the customers where I just talked to one yesterday and he's like, Hey, we were trying to figure out how to do this in house.
00:15:40
Speaker
The table is starting to turn because back when we all started containers, we're like, hey, I'm going to containerize my application because then I can run it anywhere.
00:15:49
Speaker
And then we're like, well, I can't really run it anywhere. But in pockets. The need now for those corporations in the enterprise that really want to have those disparate clouds and not be vendor locked in and really deploy their applications anywhere. Let's just take like a DR scenario. They want to have something running in Azure and something running in AWS. And it's very difficult to do that today.
00:16:15
Speaker
networking-wise and Kubernetes-wise. So our whole goal was to simplify that and that's exactly what Cube Slice does. So you're able to connect those two clusters and then deploy the workloads of the microservices where it makes sense across those cloud providers. Or if it's a DR scenario, just keep it as a warm standby. If something happens in AWS, just flip the switch and you're over in Azure and your customers are on the wiser.
00:16:40
Speaker
Absolutely. And we'll get in some use cases, but Cube Slice is open source. Has it always been open source? When we first started, it was not. Okay. So what was that? Why was the choice to open source it?
00:16:51
Speaker
Yeah, so we really wanted to harness the power of the community, not only give back. So when we were developing Cube Slice, we were following close with the CNCS, SIG Multicluster Group, just to make sure that we're in line with what the community is looking for. So we took pieces of Cube Slice, not the whole thing, right? We took the connectivity piece of it and then open sourced that out to the community.
00:17:14
Speaker
So we really want to, you know, obviously enable the developers, any DevOps teams out there, the community, you know, the Kubernetes community as a whole, because, you know, we all take from it anyways. We want to be good, good stewards. You can dribble it back. Right. That's awesome. You know, and it gives like those younger developers that are coming into the space, a new project to work on. Right.
00:17:37
Speaker
No, that's a good reason, like building community and just making sure people have the tools and technology to set up. And it's great. Like we don't want to develop in a, in a vacuum, right? Oh, I think the customers want this. So by giving it out to the developers, they'd be out their own little unique use cases. And you know, those, those commits and PRs will help us just grow the platform.
00:17:58
Speaker
for sure. So like I want to focus on the hybrid cloud multi cloud story, right? Like speaking from my past terrible experiences, I've configured like a Cisco ASR router on prem to talk to like a Cisco CSR instance in the cloud and establish a VPN.
00:18:16
Speaker
And that's really difficult and really tedious. So I'm sure that's one of the challenges that CubeSlice is solving for with the whole cross-cloud thing. But can you talk a bit more about just the general challenges that you see when you're talking to customers that want to adopt such an architecture?
00:18:31
Speaker
Yeah, sure. And I can just give you some of my own previous experience outside of our customers today. So prior to joining Avisha, I was at a consulting company called Taos, and they were bought by IBM. And one of the things I did there was I led the Google practice. So part of my consulting pieces on the Google side was really to get those Google customers onto Anthos.
00:18:55
Speaker
And so I started the Anthos practice there and just exactly what you said there, just me getting it set up in my lab.
00:19:03
Speaker
of just getting Anthos up, right? Just getting the developer workstations in place, doing that connectivity. I could only work with the F5s. If I was deploying it on VMware, I had to use the seesaw load balancer. And they're like, well, if you've got a PANS or anything else, you're on your own, you've got to kind of figure that out. So once we did all that, part of the practice, the big part of the professional services is going into the customer site and doing IP mapping and management
00:19:31
Speaker
from day one. That is really the bulk of it. So depending on, like if you're a huge enterprise, you could have thousands of, depending on your account strategy, let's take AWS, right? You could have thousands of accounts. Every developer has their own AWS account, how those systems talk to each other. Now you've got STS gateways, you've got a Ravi nightmare, you've got it all. And then if you're trying to connect, say from developer ones,
00:19:55
Speaker
Developer ones account developers too and they want some services to talk. Well, you can't really do that Right if you have the IP overlap
00:20:05
Speaker
So maybe in that case, a lot of customers just stick with the Tenda and they just give everybody the default Tenda and that's it. And so when you get into those sticky situations, it's a nightmare and it just takes months, months to just untangle that and plan for it before you even do any type of connectivity between the clusters or workload.
00:20:26
Speaker
And you're talking about how organizations are distributing these CIDR ranges. But even inside Kubernetes, inside a specific cluster, I know initially I used to have a specific parameter, CIDR range, and I used QBADM. But I no longer do that. It's just whatever Kubernetes gives me by default. And obviously, that will create the overlap in IP spaces. And then it's not fun when you have same ranges across two different clusters trying to talk to each other.
00:20:57
Speaker
So I'm glad that we have a solution like Cube Slice that's open source that can help us solve that. Do you want to talk about how Cube Slice actually does this? Yeah, I certainly can. So what Cube Slice does, let me just explain the slice just for the audience so they understand the contract. It is not a pizza. It is not a pizza.
00:21:18
Speaker
So let's say we have three clusters, one in each major cloud provider, right? We've got one in AWS, Azure, and Google. And so the concept of the slice is being able to take resources from all those clusters and create a virtual cluster, if you will, across all three physical clusters in the different cloud environments.
00:21:38
Speaker
And that encapsulates a slice. You onboard your application onto that slice, underneath it doesn't know what's going on. And the mechanics behind it is we label some of the nodes. You can create your own node group if you want for the gateways, but you just have to label one of the nodes in each of the cluster. And that's when we do the Helm install where the bits get installed.
00:21:59
Speaker
So when we install the bits, we create a point-to-point VPN tunnel automatically between all the clusters that are engaged. And so every slice has its own router, its own DNS, it's isolated. The application is isolated. So the connectivity is per slice.
00:22:18
Speaker
And we handle that. We handle the certificates. We handle the service imports and exports across the cluster so they're aware. We handle the namespace sameness across all those different clusters. So from the app level, if you're looking at it from all platform teams' point of view, they can put the resources wherever they want and just give that slice to the developer, let them deploy their application, and that's all they need to do.
00:22:42
Speaker
And all that traffic, I know there's another question maybe around this, but the traffic itself is layer three, east to west across all those clusters. There's no need unless the application requires it, right? There's no need for a separate ingress for those services to talk now. Okay. That makes it really easy because like ingress and points are great, but yeah, it's an additional resource that you have to develop and manage on top of your clusters. Yes.
00:23:07
Speaker
Yeah. Correct. It's funny. This brings me back when I first got into just this industry as a whole, like coming out of school. I was working on OpenVSwitch, and I was working in the OpenFlow lab where we were testing switches and all these things. But I remember specifically a little later on when Docker first got announced, some of the, I forget what start of it was, who got bought by Docker back in the day. They developed this basically overlay technology where you could just
00:23:34
Speaker
create an overlay network within Docker and Swarm, and it would just kind of, you could design whatever set or address you wanted, and it would just be like, poof, you could use the same one over and over again. And I was like, this is magic, right? And that's the, I think a lot of the power is kind of,
00:23:49
Speaker
You have your own little space. Everything can communicate onto it. And now, being able to do this kind of across clouds is really powerful. And I guess the question that comes up when you were talking about that is, do the end users need to be aware of any requirements between those? So are there latency constraints or requirements that people need to be aware of? What does that look like across regions in cloud?
00:24:18
Speaker
Yeah, they don't really. It doesn't even come into play, right? Because it's not going all. If this was the typical cloud networking where we, let's say AWS and we were getting the STS gateways, we had to figure out where we're going to route it through. There's another management network. We don't need to because it's point to point. So the latency between the two clusters is very minimal.
00:24:39
Speaker
Okay. That's awesome. Like, hopefully this helps users that, I don't know, by mistake, keep their cube API server exposed to the public. Making sure that this doesn't require any public connection definitely helps. Yes. And the, you know, the other piece I just want to tack onto that, because you mentioned it too, Ryan, is that cider address, the overlapping cider address. So in the slice construct, because it's isolated, let's say you were constrained with your IP space itself. If you only had a few cider blocks that you could
00:25:09
Speaker
Um, that you could carve out. So we could take the same RFC 19 and address, whether it's one, nine, two, 10, what have you a slash 16. And you could reuse that same cider on multiple slices. Oh, nice. Okay. You could have 10, 10 slices across those same three clusters using the same cider. And they never will conflict.
00:25:34
Speaker
Yeah, that's very powerful. I know. That's awesome. Well, OK. So before we go into more of those nitty-gritty details, which I do want to get into, I think I want to go back to a little bit of the use cases again, right? So communication across region, cloud, those kind of cluster, that makes sense. But there's also one that we've talked about, Bob, and with previous guests is how teams do migration, both from we've done this from sort of a storage capacity, but also just from like
00:26:00
Speaker
I'm a team developing things on-prem. I'm using Kubernetes, but now I want to start using Cloud, or maybe they're moving off Cloud. How does this come into play when you're crossing that chasm of on-prem to Cloud, and what does that migration look like in this world?
00:26:15
Speaker
Yeah, so that's a great question. And it's when I get asked all the time, it's actually the use case I lead into with the customers. Now, having mentioned that I was at the Taos practice, this was one of the things I had to do all the time, right? Cloud migration is from on-prem into the cloud.
00:26:30
Speaker
And so I had a large online, not a retailer, but a vacation travel kind of site. And they had a bunch of Kubernetes clusters on prem. We were building their landing zone in Google, creating some GKE clusters. And they're like, hey, guys, when can we get these clusters? Because we've got a
00:26:51
Speaker
fix our pipelines, we've got to, you know, do some massaging in the back and like, well, guys, you know, we got to build a platform, right? So we have to do all the routing. We have to work with your network guys to get the PGP tunnels set up between Megaport and your on-premise and all that. And that's weeks, right? It's weeks worth of time. So had I had this software,
00:27:13
Speaker
we could have just connected to still while we were doing the platform stuff, right? Because it's a point of connectivity, at least for them to provide them a test bid so they can set up all the ancillary things that they need to do in the meantime. So that's one half of that use case. The second half, let's just say that the foundation was already built.
00:27:32
Speaker
Now it's just me connecting those two clusters together or three or five or wherever the case may be and redeploying my app onto the slice and splitting the microservices wherever I wanted them. So if I had 15 microservices, I had three clusters and it made sense to split them into five. So whatever the case may be, I could put five on AWS, five on Azure and five on Google and it would make no difference to the end customer. And that's how I demo it too.
00:28:02
Speaker
demo to customers. I use the on Google's online boutique. It's got 12 micro services. I sprinkle them all around. We go through the front interface. I make, you know, adjustments in the carts and I buy the fictitious stuff and there you go. Yeah. So, so in terms of like, you know, doing like blue green and things like that, you can kind of like,
00:28:20
Speaker
role services or applications out in either cloud and start hitting more and more of the one you're deploying in a different cluster. And because it's on the same network, you'll still kind of, you can set it up so it round robins. Absolutely. Yes. So everything that we've done, it's all Kubernetes native. So all the Kubernetes constructs apply rolling updates, rep cassettes, the whole, the whole spiel.
00:28:41
Speaker
And I think this is one of the solutions that actually will help me with the cloud bursting scenario, where I don't want to use up all the capacity or elasticity that I have in the cloud, but I want to have that option open. And if it's a slice that's pre-configured, come Black Friday, I just have to ramp up my instances and then just spin up more pods.
00:29:02
Speaker
Exactly. That's awesome. And so that's a use case we're doing with one of the large financial institutions. That exact use case. Awesome. Speaking of demoing, I feel like now, given the slice, and I talked about pizza before, there's got to be a demo in your future that has to be selling pizza slices or something. Oh, I don't. Or is there a slice app? Isn't that a real application that you can connect to? I've ordered pizza through it. I think so. I think there's a thing in the future there. Yeah. I've used it plenty.
00:29:32
Speaker
Okay. So Rob, you mentioned that everything with Cube Slice is Kubernetes native. So how do we install it? Are there operators available? What does the installation workflow look like? Yep. So you install it via Helm. We have a Cube Slice CLI wrapper. So we create a topology file. You fill out the context of the clusters that you're trying to configure.
00:29:56
Speaker
You can use that to install and it'll install all the operators on all the clusters that are in that topology file. Oh, perfect. And it creates all the connections between the different clusters that are part of the slice. You don't need to do a thing.
00:30:09
Speaker
Once it's done, it's all configured, then you log into the UI. And from inside the UI, if you want it to, or you don't have to, but you can use the UI to create the slice, do the isolation, the resource quotas, all that fun stuff. If not, you can go to the YAML section of our documentation and create the slice config YAML if you want and tailor it as you need to and go through the normal construction.
00:30:30
Speaker
That's awesome. So like, is this when you are seeing users and customers adopting this, right? Do you see this, like developers doing this on self service basis, or they're still relying on your network teams, quote unquote, to like, Oh, please set up a slice for me. And then as a second half, I know it's a long question, but as a second part, right? Can I integrate the slice construct in my CI CD pipelines or my GitHub GitOps pipelines that might have to deploy applications for specific slides?
00:30:58
Speaker
The answer is yes and yes. The first half of it is it's really more of
00:31:06
Speaker
we're seeing the platform teams. And the larger enterprises, they want the control to create the slices, and they'll use whatever their normal service now or Jira tickets, whatever, to get the request to provision something. And some are also saying, yeah, we as a platform team, we can totally do that, but now we need to educate our internal developers on our own.
00:31:31
Speaker
the freedom to deploy wherever they want now. So when they're making changes to their CACD pipelines and they're doing the redeploy in the slice, where clusters do they put those microservices? Yeah, I think that's one thing as we keep interviewing various people from around the industry, it's one thing we're seeing is it really depends on how much that organization has really adopted sort of that DevOps and sort of platform team because
00:31:58
Speaker
The reality is it does vary, right? The unicorn would be one team knows all these things and can manage these things. But as we're asking one team to also know storage, also know networking, also know all these other things, I think that's a natural to see a yes and yes scenario. So it's good to know.
CubeSlice: Advanced Use Cases and Integrations
00:32:22
Speaker
Speaking of those types of requirements, provisioning the network or a slice is one thing.
00:32:28
Speaker
But what about beyond that? How is that network secure? Does it also provide things like firewalling or network policies? You mentioned it was just L3, right? But does it help with other layers as well? It does, yeah. So the L3 connectivity between the clusters, it's obviously encrypted. The slice construct itself is the isolation barrier.
00:32:55
Speaker
we layer on network policies onto the slice at the slice level. That encompasses those namespaces. So if you already got network policies, great, it's just an additive policy already. So when you go into the UI, you click on, let's say you want to onboard a couple more namespaces for that application to communicate with, you create the policy, you apply it, and the UI does the rest.
00:33:20
Speaker
Okay, interesting. So I think, like, since you brought up UI a couple of times, right, I wanted to ask, like, is the UI component of CubeSlice also open source? Or that's something that you have to pay for? Because you see other vendors in the ecosystem do that, right? Like, oh, the CLI thing is free, but then you need a UI, it's paid. So I wanted to, like, start there, and then I have a follow-up question as well, Rob. Sure, yeah, it's the latter.
00:33:43
Speaker
So the UI comes with the enterprise version. All the things that we're talking about, you can do with the open source piece. So the open source piece is the connectivity piece. So if you want to do any of the other bells and whistles, the resource quotas on the slice, you know, graphically, you can do it obviously in Kubernetes constructs, right? But the automation in the UI is where that's at.
00:34:05
Speaker
OK, gotcha. And then I saw on your website, like when Rand and I were doing our research, Cube Slides also helps me with IP address management or IPAM. Is that a feature that's available in the open source one? How do I manage IP addresses across different clusters?
00:34:21
Speaker
Yeah, so the IP management piece is around that RFC 1918 address. That's what it's referring to. So if you've got that one, you don't have to do the bulk of the planning across all your different cloud providers and set up. Obviously, you have to do some to make sure that that one that you choose does not conflict in your other sites. And then you could repurpose it across all the slices. Or if you have a handful, you can just use a subset of them.
00:34:51
Speaker
Okay. And how does it work with like CNI plugins? Do I need the same CNI plugin empty? Where different CNI plugins will work? So the way it actually works is under the covers, anything that's going on traffic wise, and it's communicating in that local cluster, it uses the local CNI.
00:35:10
Speaker
If it needed to talk to a service in another cluster, it's going to use the NSM interface which is going over the tunnel to talk to the other cluster to get the information back and then it routes it back through the CNI.
00:35:29
Speaker
CubeSlice work with service meshes. I'm just thinking about the different scenarios, right? CNI transparent that works, but if I, as an organization, I'm already using Istio or something else from a service mesh perspective and enforcing or doing traffic routing, how does that work with CubeSlice? Yeah, we say, God bless you.
00:35:48
Speaker
That's a fun answer. Yeah, because we don't care. It has nothing to do. It does not change the slice behavior. If you're already entrenched in Istio, you can certainly configure the slice to use Istio ingress and egress for your application. Nothing really changes there. We do have some customers that ask us, if I'm going down this path, do I need a service map?
00:36:16
Speaker
And we say, well, you know, most of the time you're doing the service mesh for the intercluster traffic management, right? And we're more looking at the intercluster traffic management. So if you're using Istio in the mesh in your existing cluster,
00:36:35
Speaker
Keep it if it's serving the purpose for you. It doesn't interfere with what we're trying to accomplish in the connectivity piece. Got it. So you effectively have pods that are sort of dual-honed to the network that is a slice and also to the rest of the Kubernetes cluster. I'm just picturing this weird Tron tunnel on my head for some reason.
00:37:03
Speaker
But yeah, no, I think that's really powerful. You know, this show started off, we talked a lot about container storage because of Bob Nye's background. And I did see a couple things around, specifically around databases, say running like Postgres or Mongo or CockroachDB, which even, you know, Cockroach being sort of a multi
00:37:24
Speaker
multi-cloud, multi-region sort of database as well. What can Cube Slice, I think, help with when running databases specifically? Because things get naturally a little more complicated when you start adding storage to the database.
00:37:39
Speaker
Yeah, so we have, if you go to our YouTube channel, well, you can do this at the end, but we have a video here we're doing a demo with a Pacman application, right? And we're using MongoDB with three replica sets. And those replicas are across three different cloud providers. And so in that entrance,
00:37:57
Speaker
instance, we're showing the resiliency of having the application on the slice across the cloud providers, and we'll generally kill the pods that are running the DB service on the master and wherever it's running. And, you know, Mongo does its thing, as it's alleging, brings it back up, and the slice is none the where. So when you go back to play the application, in the front end, you'll see that the cloud provider changed.
00:38:21
Speaker
But the data that has all the historical high, like your high scores, is thrown intact.
00:38:28
Speaker
So it helps by, you know, allowing you to distribute that, your databases across the cloud providers and not having to do anything necessarily special that they already do. Okay. So you can, you can have essentially replicas of each of those types of databases in different regions or clouds and let them do their typical application level replication. Correct. Over the network, but doesn't stop you from using something like storage or volumes or PPCs locally to keep things safe as well. Okay.
00:38:58
Speaker
Correct. Interesting. And okay, that's a good use case. My point was like, are people actually doing that? That's like a really complicated setup, right? When you have. I hope so. Yeah.
00:39:10
Speaker
We actually created this demo because one of the large financial institutions in Toronto, a big one over there, this was the stem from their use case. We've got these Mongo replicas all over the place and we need to protect the data. Even if it's on-prem, they wanted to be able to safeguard their data and if something goes bump in the night, that there's no interruptions to the transactions.
00:39:37
Speaker
That makes sense. I mean, I always think back to, I don't even know when this use case, but the Netflix Cassandra use case, where they had those things kind of everywhere on those rings. I know we didn't talk about Cassandra, but people have been doing these types of architectures. And now I think the more people are deploying them on Kubernetes, these things are just necessities that come as part of, hey, I want to do this architecture for availability or something.
00:40:07
Speaker
And in that same data vein, we're seeing customers that are coming out of the woodwork, especially around their Kafka clusters. So having distributed, they're having a hard time doing distributive Kafka across cloud providers. So they can have multiple Kafka clusters wherever they are in AWS, but getting them to talk to each other in either cross region, what have you, is difficult. So they're trying to, let's,
00:40:34
Speaker
One of the next demos we're trying to put out there is doing multiple Kafka clusters across different cloud providers on the slice and having it stream as data through using the resources from both clusters. So is that like a broker would be replicating? Correct. I forget the actual terminology in Cascaland. We didn't broke it.
00:40:59
Speaker
It's awesome. And Ryan, thank you for bringing up Cassandra, right? I just remembered when you were talking like Kate Sandra, the operator actually now supports deploying Cassandra instances across different geographies. And then, but in their documentation, they're like, you have to take care of networking. If all the classes can talk to each other, we can deploy it. And this just fits perfectly. So it's awesome.
00:41:20
Speaker
At the very beginning, I think we did the same thing with Postgres. Before we had all the UI and all the fancy stuff, that was one of the first demos, just doing a Postgres replication across different cloud providers. It was great. Gotcha. And I think I want to talk a bit about autoscaling. I know we spoke about cloud bursting. Can CubeSlice help me with specifying any autoscaling rules? If I'm moving things around to a different cluster inside the slice,
00:41:49
Speaker
Can I scale up the worker nodes, things like that? Can AI, the new buzzword, play any role in that? So Cube Slice, no, we have a second product called Smart Scaler for that. Smart Scaler does exactly what you say. We have an RL-based AI engine that learns the application behavior.
00:42:11
Speaker
and it will set the configuration for your HPA Autoscaler now, constantly over time. So you won't have to go into every single one of your clusters for all those particular workloads and set the minimum and maximum thresholds and waste time and money, right? So the goal there with Smart Scaler is really about the savings.
00:42:28
Speaker
So we've seen in some of the training models that we have anywhere up to almost 70% savings on the clock. And so we're going to be modest. Let's just say it's 35. Even 35% savings is great. And the other thing that we kind of tout around smart scalers, now you can actually offer your users SLOs.
00:42:51
Speaker
against that, where before you can't. And then of course it's around, really around the cost save, right? If you're just really nearly spinning up pods, like to your case, like Black Friday, I'm just gonna, I'm gonna burst into the cloud, but I'm gonna have my auto scaler set to however many thousands of pods, cause I don't know.
00:43:06
Speaker
Right. It's back in the AWS days when the black Friday came around, we had to talk to the customers and say, Hey, uh, we noticed the data from last year. We need to warm all your EC2 ends. You get them ready. Right. So we do all that fun stuff. It's in the same concept, but now, you know, it's a particular AI that's doing it. I like, are we, are we in that part of, uh,
00:43:28
Speaker
the advancement of these technologies where we're allowed to say back in the AWS days. I feel like, I mean, it's the reality is I think we think that way, given the word technology is going and kind of where we want it to go. But I just had to key in on that. Even now they've changed that, even at AWS, right? There's no more pre-warming. They kind of figured all that out. So we say, oh, gee, AWS days, maybe.
00:43:57
Speaker
So, you know, we've seen other applications and platforms do similar things in terms of using our models to kind of help different aspects of running Kubernetes platforms.
00:44:12
Speaker
whether that be cost or kind of monitoring or even troubleshooting, does the auto scaler piece or the smart scaler piece tap into the kube slice networks to know what's going on at all? Or where does it get metrics or data from? So it's pulling its metrics from Prometheus. OK. So it's taking all the Prometheus metrics, and that's what they're using to train the model. Or if you're using Datadog, it'll pull your Datadog metrics from all the clusters and collate it and do its algorithms based on that.
00:44:40
Speaker
or the traffic prediction. Yeah. And Prometheus being so widely, I think used that's, you know, so you have Prometheus, Datadog, any others that I think are supported out there or? Right now off the top of my head, it's I think it's those two. Okay. Yeah. I mean, if you Prometheus, you got probably 80% of the market, right? I would imagine I have that there's no basis in my percentages right there. But that's a lot of having a podcast, right? You can make it.
00:45:09
Speaker
And then the feature for that too that they're developing based on some feedback from a customer was they want to have the ability to do like scheduling via the calendar. So let's say you know Black Friday is coming and it's based on calendar events. Or the use case is really around like an online betting retailer. They're like, we cannot lose
00:45:32
Speaker
any data, there can be no disruption whatsoever. If a fight's coming out, as simple as that, there's going to be a fight coming out. Let's say it was the Musk and Zuckerberg fight, and you're going to put bets on it. They don't want to lose not a dollar from that. They want to use that feature to make sure they're scheduling it based on what the events are.
CubeSlice in Retail and Edge Computing
00:45:56
Speaker
And that, do we, do we know that's not happening or is there? I think that's not happening. Cause I would have paid that pay per view. I would pay for that. Super bowl. Like the gambling company doesn't want to lose any bets during the super bowl. Let's say, there you go. Well, speaking of use cases, do you, um, do you have, uh, some really good use cases that you could talk about? You don't have to, you know, mention names or anything like that. I know you've mentioned a few already, but any favorites that come to mind that would be kind of.
00:46:25
Speaker
Yeah, the only one I think I haven't mentioned was we had a large retailer that
00:46:32
Speaker
they're in a precarious situation for pop-up stores, especially during the summer, where getting the connectivity from whatever that location may be and connecting it to the backend, it would just take too long for them to do it. So we came up with, I call it the edge stack, edge stack retail accelerator. So we basically took some Novo,
00:47:01
Speaker
nooks. And then we installed Kubernetes on the nooks. Now we connected those nooks to their backend Kubernetes, OMS systems and all that fun stuff. So now there's a mesh. I don't, I don't want to use the word mesh. Now there's connectivity between pop up store and, and their backend seamless.
00:47:22
Speaker
And so they also had problems. And we partnered with Cox Edge around this because we needed that last mile of connectivity. And so their big problem was they do some pop-up stores, let's say, here in New York City. And some of the buildings are so old that they can't get reliable internet. So they end up having to rent trucks, put the antenna on the damn roof, and then run all the children down so they can operate the pop-up store.
00:47:48
Speaker
And the other use case for that same retailer was around space in the retail stores. So if you go into any high-end retailer, any retailer really, you go in the back and there's a server rack in there with a switch and some other stuff. And so in pop-up stores, having that space, it's a loss, right? They need to have merchandise in there so they can sell. So having just this underneath the counter,
00:48:12
Speaker
at the end where they're doing the POS point of view or the POS checkouts in those days so they could get more square footage to do what they need to do.
00:48:20
Speaker
got it. So that basically, that point of sale software is running sort of at the edge closest to where these pop ups happen and kind of provide that front end connectivity. If we want it like the full solution was it was three parts. It was huge slice Cox edge for the for the edge location. And then we also partnered with Spectra cloud. So they did that. Okay. Cool manage track. Yeah, we were able to stack all the the corporation's applications into the
00:48:48
Speaker
into the blueprint or the pancake, which you call it. And so we deployed it. It did all its configurations on your own, which kind of alludes back to your CI CD question. So yes, since it's, you know, native Kubernetes commands, you just do whatever you need to do. Very cool. I mean, I'll just talk about slices and pancakes. I'm getting hungry.
00:49:06
Speaker
I got some pizza in the fridge, actually. Why is there always a connection between technology and food? I'm not disliking it. Well, I think this has been a super powerful conversation. I think there's so much, I think, in the networking space, especially I think if you're new to Kubernetes, a lot of things can be complicated, but the networking side of things
00:49:28
Speaker
might be one of the tougher ones to really wrap your head around, right? Just because there's at a base level without adding more overlays on top of, of networks. I think coming out with new solutions that drive simplicity has to be sort of at the, at the front, right. Yeah. And so mentioning on that note, um, Avisha themselves are now in the Gartner hype cycle.
00:49:49
Speaker
Okay.
The Fun Side of Tech Talks
00:49:50
Speaker
They put us into the new one is for Kubernetes networking for, you know, is there a trust networking in the hype cycle for container technology? And then the other one was for the programmable platform. So same thing for infrastructure strategy and, you know, data set and infrastructure technology. So it started to pick up more and more and hopefully that utopia of deploy your application anywhere, we're able to help those customers do that. Got it. That's one done is in between, I guess. Yeah.
00:50:16
Speaker
Okay, I guess Rob, one more question, right? I think you brought up a lot of demos. Are there communities that our listeners can join, places where they can find these demos, getting started guides, anything that you can, you want to link or give a call out to?
00:50:34
Speaker
So if you go to Avisha.io and then go to the documentation section, that'll bring you to the enterprise documentation. There we have a link for a playground. So if you just want a playground, we have a bunch of kind clusters set up and you can go through and it proves around. There's also a quick start guide also.
00:50:52
Speaker
If you wanted to try it in your own environment, we give a 30-day free trial license. Go ahead, download it, install, play with it, break it. If you needed support, you can go to support.wcio, put in a ticket. If you're interested in the open source version, that's cubeslice.io. Then we're also in the Kubernetes Slack ecosystem. If you're already part of the Kubernetes,
00:51:16
Speaker
uh, Slack workspace, you can just go to, uh, the hashtag cube slice channel and talk to the rest of the community users there. Great. Great. And will you be at KubeCon Chicago yourself?
00:51:27
Speaker
I personally, I don't know if they're going to send me. I'm pretty sure there'll be a cohort of some kind, but I may or may not be there. I hope to be. Well, whether it's you or not, Bob and I usually do some live shows on the floor. So if there's someone you'd be interested about talking to us again, we usually do sort of smaller 15 minute sort of conversations, but then we jam them all together for some kind of episodes. We'd like to get an update for kind of what you guys are up to there as well.
00:51:54
Speaker
All right. So time for chat GPT question segment. So I mentioned earlier that the question would be, if networking and Kubernetes were a superhero, what would it be? And you can give your best answer, or I can let you know what chat GPT says, which is quite funny. Well, if it was smart, they would say cube slice, for sure.
00:52:14
Speaker
It gave the superpower sorry it gave the superhero the name of connectron or netweaver I don't give to I don't know why I gave to You know, it's funny didn't
00:52:27
Speaker
I think HashiCorp used Netweaver under the cover for its regulator. Maybe they got it from here or maybe it got it from here. The superpower it says here would be seamless and dynamic networking. And I won't give you all the, I'll give you the brief intro to its super abilities.
00:52:46
Speaker
would be a teleconnect net weaving firewall manipulation traffic shaping subspace message messaging i'm not even sure um yeah what what to do with that one and a resilience aura uh against things like ddos attacks and it's arch nemesis the latency lurker nice it's just weird that it's making new character i would have just gone with like spiderman i don't know helping
00:53:15
Speaker
I would say there needs to be some t-shirts made out of that one. We can make some t-shirts. I didn't even mention it. I didn't mention this Connectron superhero has a battle cry. It goes like this. Forge the digital web. Unite the code. I don't know why it chose that phrase, but there's this battle cry. It apparently has very little to do with networking, I think.
00:53:36
Speaker
Depends on who you ask. Oh, man. Well, Rob, it's been a pleasure having you on the show. And hopefully we'll talk to you again soon. Otherwise, take care. You too. Thanks, fellas. Thanks again.
00:53:50
Speaker
Thanks, Rob. All right. Well, that finishes up our conversation with Rob. I think the entire networking space around Kubernetes, like I was saying in our conversation, is a really complex one. I started, like I said, my roots in sort of this networking space. And just if you were to take Kubernetes, vanilla Kubernetes,
00:54:08
Speaker
and try to dive into the networking side of things and the CNIs and how the cluster IPs work versus external IPs versus load balancing and all these things. It's quite a complex space. So I do like the simplicity of the multi-cluster kind of cube slice technology that it seems to be kind of targeting because it's already
00:54:33
Speaker
a huge space. So, I don't know. What did you take out of this conversation about it? Again, I really liked the conversation because I don't know of any other solution right now, right? Obviously, I can do some research and find that out, but this is an interesting problem to solve and an actual problem to solve instead of just making shit out. I think this was a real thing, like a real challenge. As I said on the episode, right, I have configured like for VM-based workloads, I've configured like the Cisco routers on-prem and in the cloud and
00:55:02
Speaker
created these VPN tunnels, and it took a lot of time. And if we want developers to have the agility and set up multi-cluster, multi-cloud things, we can't wait a couple of weeks to configure the tunnels between two clusters. Having something like QSlice create that Slice construct definitely helps, makes multi-cloud easy, makes hybrid cloud easy. So I really like that part. I also like the fact that most of what we spoke about today was open source. I know the UI thing, I was glad that we clarified that piece.
00:55:29
Speaker
The UI for it is an enterprise thing. But if you just want to deploy Cube Slides on your own using YAML files on the Helm chart, it's all open source. CubeSlides.io. So those were a couple of my key takeaways. Yeah, I mean, it's a good point around the complexity of creating these networks. And we didn't really talk about underlay networks at all in this episode. But the reality of it is none of that really goes away. There's still a lot of base connectivity, and there's still a lot that goes on in that space.
00:55:58
Speaker
When you get to the point of this, again, abstractions are key when you're working at the level of these types of Kubernetes namespaces and slices, being able to click a button and connect to clusters is super powerful. Given that you already have the rest of that connectivity in place for those clusters to talk to each other, now you can just make your applications seamlessly.
00:56:23
Speaker
be able to discover each other and talk to each other. I think it's super powerful. As we mentioned earlier in the intro, multi-cluster is definitely something companies are thinking about and releasing new projects and products around. So I think we're going to see that more and more as time goes on. So for me, I think an interesting point is that migration is still a leading kind of use case that Rob still uses when he talks to customers.
00:56:47
Speaker
I think this definitely shows that as we talk to these guests on the show, we often talk about leading edge things, but the reality of it is tons of people are still moving to this type of technology, moving to Kubernetes and still really thinking about that migration use case, about how do I get the workloads that I have?
00:57:08
Speaker
here that we've been testing and, you know, stabbling with Kubernetes into these different spaces or from clouds now that we want to start using them. So I think that was just an interesting point that was not so surprising, but really validating, I guess, to hear from me and just kind of shows you that a technology like networking doesn't stop at just, you know, connectivity and discovering of services as other things that you can obviously use it for in test, dev, staging, those kind of things.
00:57:39
Speaker
Cool. Well, that was a fun show. And I think that brings us to the end of today's episode, Bobbin. So I'm Ryan. I'm Bobbin. And thanks for joining another episode of Kubernetes Bites. Thank you for listening to the Kubernetes Bites podcast.