Become a Creator today!Start creating today - Share your story with the world!
Start for free
00:00:00
00:00:01
Kubernetes Alternatives - when NOT to use Kubernetes!
698 Plays2 years ago
In this episode, Ryan and Bhavin interview Alexander Mattoni - Co-founder and Head of Engineering at Cycle.io about When to use and When to not use Kubernetes. The discussion focuses on the challenges associated with Kubernetes adoption - On Day 0 and Day 2, and what are other alternatives available to organizations that are just looking to run their applications easily. We talk about how Cycle.io can help organizations build a simplified infrastructure stack to run their applications. Have a listen and let us know what you think about Kubernetes.
Also, send us your 3-4 mins clips about your experience with Kubernetes - to be shared on future episodes
Show Notes:
Alexander Mattoni - https://twitter.com/alexmattoni
Cycle.io - https://cycle.io/
News:
- AWS Controllers for Kubernetes - ACK for Amazon EC2 https://aws.amazon.com/about-aws/whats-new/2022/11/aws-controllers-kubernetes-ack-elastic-compute-cloud-ec2-generally-available/
- Removal of GlusterFS in 1.26 - https://kubernetes.io/blog/2022/11/18/upcoming-changes-in-kubernetes-1-26/
- Two possible data inconsistency issues in etcd v3.4.[20-21] and v3.5 - https://groups.google.com/a/kubernetes.io/g/dev/c/sEVopPxKPDo?pli=1
- Kubecon NA 2022 recordings - https://youtube.com/playlist?list=PLj6h78yzYM2O5aNpRM71NQyx3WUe1xpTn
- Kubernetes Bytes season 1 on youtube - https://youtube.com/playlist?list=PLCOmEAve4xr2lbCd6sPXMRf6XcZeWuaJ5
- Kubernetes Bytes at Data On Kubernetes Day - Kubecon NA - https://youtu.be/q_K8Ma9LxWA
- Cloud Native Security Con NA - Feb1-2 https://events.linuxfoundation.org/cloudnativesecuritycon-north-america/
- TiKV is an open-source, distributed, and transactional key-value database - evolutions of TiKV https://community.cncf.io/events/details/cncf-cncf-online-programs-presents-cncf-on-demand-webinar-the-evolution-of-tikv
- Backup and Restore using alpha k8s checkpointing feature - https://martinheinz.dev/blog/85 | https://kubernetes.io/docs/reference/node/kubelet-checkpoint-api/
Recommended
Transcript
Introduction to Kubernetes Bites
00:00:03
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.
Thanksgiving Anticipations
00:00:14
Speaker
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.
00:00:29
Speaker
Good morning, good afternoon, and good evening wherever you are. We're coming to you from Boston, Massachusetts. Today is November 23rd, 2022.
00:00:37
Speaker
I hope everyone is doing well and staying safe. Bhavan, it is the day before Thanksgiving. I hope you are planning some sort of fun activity for the rest of the week. You're off, right? Yeah, it's Thanksgiving week. It's like the holiday season is about to start. I actually need to like wrap up a couple of meetings today, but that's it. Then Thursday, Friday, like our holidays. And I think I'm doing like a friendsgiving thing on Saturday.
00:01:06
Speaker
Yeah. Yeah. So quiet weekend. How about you? Yeah.
Ryan's First Thanksgiving
00:01:11
Speaker
So I normally always travel to New York to see a lot of my family and my friends do a Friendsgiving. They do like two turkeys, but a bunch of friends. Those friends all had babies this year, so that's not happening.
00:01:24
Speaker
And so we decided to basically host Thanksgiving the first time up here since we moved into this house. So we've only been here three years. So we're having my sister-in-law and her six kids, God bless her, and four dogs. So we're gonna have a lot of dogs and a lot of kids in the house. But I'm excited because it gets me to be able to cook. I like cooking. So I went and found a turkey farm. So to get fresh turkey,
00:01:50
Speaker
This year, instead of frozen, nothing wrong with that technically, but a lot of people talk about the fresh turkey doesn't have those crystals that form when it's frozen and you get a juicier bird, blah, blah, blah. I bought into it and I found this place in Lancaster, which turns out everyone in Massachusetts and everyone in Boston apparently goes there because I signed up for this.
00:02:18
Speaker
which was a pickup date of Tuesday. It was like, okay, come between eight and six. And I was like, oh, that's pretty lenient. You know, that's fine. Like, sounds like it's kind of an easy thing to do. Well, I show up to this road with this Turkey farm, Bob's Turkey farm. It's a great Turkey farm. I'm not, I'm not dissing it at all, but, um, I show up and there's lines of cars on both sides of the road, right? The, no way you're getting close to this parking lot first of all.
00:02:45
Speaker
So I was like, what is happening? People, there's so many people here. I got there at 9am. Anyway, so on a Tuesday, right? I'm like, I guess everyone's not working. Still, I walk, you know, I parked down the road, I walk like, I don't know, a couple football fields, like I had a decent walk.
00:03:04
Speaker
And I get to the back of the store and there's this Disney park style line through the back of this yard, right? It's not even in the parking lot. It's just in the grass. Just people just snaking around. And there had to have been, I don't know, a couple hundred people in line. So I was like, all right, here we go. I'm glad it wasn't freezing, right? It was actually pretty warm. I waited an hour.
00:03:29
Speaker
Pick up my prepaid turkey right wow and this was my name what about like what happened what would have happened if you resell like 4 p.m. What I'm saying I don't know so I had a buddy also go in there I warned him I sent a video of the whole line better be prepared cuz there's nothing to do here either if you're bringing kids and stuff like that Needless to say I'm happy with what I got and I took today to Brian do my own
00:03:55
Speaker
Brining solution. I got but 7am to make my Brian. That's how nice it I am. You are dedicated
00:04:02
Speaker
Anyway, so, you know, I'm going to be cooking for the family. I enjoy that. Um, I'm not hosting like 30 people. Maybe I wouldn't enjoy it if I was doing that, but I'm excited to have everyone out there coming up tonight. Um, it's going to be chaos, but that's part of the Thanksgiving.
Kubernetes Complexities and Alternatives
00:04:17
Speaker
Yeah, I know. I think we need an update from your next, uh, in the next part, like how was it? Okay. Was it really worth it? Or did you not see as much difference when you, when you actually do it way too much and I never want to do it again. Yeah.
00:04:34
Speaker
Anyway, we have a cool show today. Bhavan, I know you and I have been talking about this idea of Kubernetes bringing a lot of complexity, right? When to Kubernetes it or when to Kate sit is what I've been calling it, you know, for better or worse. And we're going to be talking with Alexander Matoni. He's the co founder and head of engineering at cycle.io. But
00:04:56
Speaker
But before we get into that, we have some news to get through. So why don't you kick us off? Yeah, it is not as heavy a newsweek as I think we had the last time when we were discussing KubeCon news, but still a few things that people can note down. I think AWS controllers for Kubernetes. So the short form for that is ACK for Kubernetes now supports
00:05:20
Speaker
managing and provisioning EC2 networking resources as well. So using Kubernetes APIs, you can provision AWS VPCs, security groups, internet gateways, et cetera. So that's something, if you are in the AWS ecosystem using EKS, this is something that might be helpful for you. I don't love that acronym. Yeah. Unintended acronym. Coming from a networking background. I know. Acknowledged, right? Every time in my brain, so I can't get moving on. Thank you.
00:05:48
Speaker
Then, I think we are getting ready for 1.26. There is a new blog that highlights the different changes that are coming in Kubernetes version 1.26. The one thing that I wanted to highlight was the removal of GlusterFS. I know in 1.25, when we did a similar discussion, we said that GlusterFS, the entry volume provisioner, is going to be deprecated.
00:06:09
Speaker
In 1.26, it's going to be removed completely. Again, GlusterFS, I know people don't usually run the latest and greatest as soon as it comes out, but start planning things because usually Kubernetes ecosystem takes a couple of releases to completely remove things that were deprecated, but this seems a lot of sudden like 1.25 and 1.26, it's gone. Something to keep in mind.
00:06:31
Speaker
Next, I think two possible data inconsistency issues with HCD. This is something that showed up in one of the things that I subscribe to. If you're running HCD version 3.4.20 and 21 or 3.5, there are a couple of issues. For example, in case an HCD cluster crashes, during the defragmentation process, you might have a data inconsistency issue.
00:06:56
Speaker
or if you're running 3.4, if you had authentication enabled and a new member joined the cluster, there might be some data inconsistency issues. So both of these issues have been fixed in version 3.5.6. So for people who really care about XCD, make sure that you update to the latest version and are not affected by this data issue.
00:07:16
Speaker
Yeah, and if you're running Kubernetes or on a platform team, you do care about it. The amount of data and metadata that's stored there, take these things seriously. Definitely go check them out. It's highly recommended. And then I think the next two things were just things you should do over the holidays. If you're looking for more content to watch, Kubernetes North America basically uploaded all of its sessions.
00:07:41
Speaker
all of the Day Zero events that happened at KubeCon, all the actual sessions that happened at KubeCon, North America 2022, are now available on YouTube. It's on the CNCF channel, but we'll still have a link for it on the show notes. We have to put our data on Kubernetes recording. We didn't put that in the last episode.
00:08:02
Speaker
Oh, yeah. Yep. Yeah, the dok day one. It's like a nine hour video. Well, we'll link to the panel at least that's for you know, this, the show.
00:08:13
Speaker
Yeah. And then I think more content, and I think this is feedback that we, Ryan, you and I got at KubeCon, right? Whenever people were learning about our podcast, they were looking up YouTube first, which was weird for us. But again, that's something that we thought, okay, this is something we should change. And since this is an audio only podcast, we basically went back and
00:08:33
Speaker
made some visual spectrum things for all the episodes that we had in season one. So all of our episodes from last year are now on our Community Spites YouTube channel. While you watch that, we'll be working on processing season two. So this season or all the episodes that we did in 2022, and we'll have those available on YouTube as well.
00:08:51
Speaker
Yeah, that is definitely some feedback we got. And I appreciate anyone who did give us that feedback of wishing it was on YouTube. So yeah, we will, of course, be adding video, hopefully in the future. We don't know exactly when, but we'll announce it. It's something that we've talked about. Now that we have the YouTube channel, there's a place for it to go, right?
00:09:14
Speaker
And if you prefer audio only versus video, actually give us the
Enhancing Kubernetes Bites Content
00:09:18
Speaker
feedback. Let us know how you consume it. It always helps. We're just trying to reach more listeners, that's for sure. So for news that I had, it was more some articles that I pinpointed
00:09:30
Speaker
uh, and wanted to bring up here. Well, the first one is there's a cloud native security con North America, uh, February, uh, first through second in 2023. I know security has been a huge topic. Bob and I have been starting to dive into it more in the show and with our guests, um, probably good place. If you're
Upcoming Events and Discussions
00:09:49
Speaker
interested or work in this sort of space, it's in Seattle, Washington. Um, and we'll link to it. And so that's only nine weeks away, right? If you're interested, I know Seattle's
00:10:00
Speaker
an awesome city. Personally, I really like it. So definitely go check it out if you're interested in going physically to that conference. The next one is a I know I mentioned
00:10:13
Speaker
Uh, the T I K V, which is the value store in our previous episode, I believe. Um, and personally I hadn't didn't know much about it. So, um, I did find an article that was more about sort of the evolution of, of it and sort of how it's, you know, it's a lifetime as an open source, distributed.
00:10:33
Speaker
transactional key value database is, I think it's a pretty cool technology. Also part of the CNCF community. It's a presentation webinar. Go check it out if you're interested and or did like I did and wanted to find more information out about it.
Kubernetes Complexity vs. Alternatives
00:10:48
Speaker
The next one is, I believe it was 125 came out with a check pointing API. Yes, Kubernetes 1.25 came out with
00:11:00
Speaker
uh, an alpha checkpoint API. And there's an article about how to use it for backup and restore, um, and in the alpha state. And I wanted to bring this up because it's a different concept of back and restore than I think we are used to covering on this show with stateful containers, right? When we talk about back and restore, we're really talking about application data that's on disc most of the time, um, and persisted, right? We back up that as well as the application metadata. Well, this, this checkpoint API,
00:11:29
Speaker
goes into more how to capture sort of the memory pages so more like a conceptually more similar to the sort of VM you know pause and restore that kind of thing but be able to kind of
00:11:45
Speaker
snapshot that thing and move it around everything in memory. I think it's an interesting API. I could see it be combined in the future with other backup and restore mechanisms that also capture things like application metadata, data on disk, and maybe even what's in memory. It could be a third point that we're interested to hear in.
00:12:08
Speaker
definitely go check this out. There's a blog post data on it as well as the Kubernetes Docs reference. Yeah, I didn't know about this. So this is definitely something that even I need to like go and read more about. So thank you for finding this link.
00:12:22
Speaker
Yeah, absolutely. So that is it for news that we had today. Without further talking, let's get Alex on the show. Alex, it's great to have you here on Kubernetes Bites. Thanks for spending your time with us, you know, the day before Thanksgiving. I know a lot of people are busy, but if you want to just jump in and give yourself an intro for everyone listening and a brief sort of title and background, that'd be great.
00:12:49
Speaker
Yeah, thanks, Ryan. Yeah, like you said, my name's Alex Matoney. I am the head of engineering and co-founder at Cycle.io. Right, great. So, you know, what we're here to talk about today is really about the idea of, Bob and I have been kind of throwing back and forth this idea of when do Kubernetes at, or Kate's at, I guess is the short answer we've been talking about. Really, this boils down to Kubernetes alternatives.
00:13:14
Speaker
So, you know, we'll get into a little bit about what Cycle does a little bit more later in the show, but I want to start off with asking you sort of what do you see in terms of challenges today with organizations that are really starting off or getting into using or wanting to use Kubernetes?
00:13:30
Speaker
Well, I'm sure this comes up all the time, but in short, complexity, right? There's so many options, so many things to choose from, and so many different ways to put together Kubernetes. Kubernetes is really just an engine, right? And there's so many things that go into that that you need to put together. And for people that are new to it, or even people who have probably done it a couple of times, getting started is such a panic-inducing event, at least for me.
00:13:56
Speaker
So apart from getting started, our organizations have already adopted Kubernetes, also looking for alternatives, given that they've already invested in Kubernetes a bit. How does that feel? Getting started, obviously, as you said, is really difficult. What about day two? Well, day two, I would say, is almost even harder. As you continue to go with Kubernetes, you can get the first stuff set up maybe within a day.
00:14:20
Speaker
But as you continue to scale, the costs and the complexities also continue to go up because your needs increase, right? What happens when you need to deploy even more instances of that app that's not scaling as well as you hoped? Or what happens when you need to increase your storage or you need to deploy more infrastructure? There's a lot of things that are continuous. What happens when you need to update? I mean, oh, God, what happens if I have to update my production cluster?
00:14:46
Speaker
Yeah, I think generally we can agree even of us that are using Kubernetes and successful with it and folks we've talked to that no doubt it adds a layer of complexity. And I think this most recent KubeCon, we did see a trend towards that user experience, right, of being sort of a challenge to create a better experience, meaning, you know, all these sort of
00:15:12
Speaker
built in cloud-based SaaS approaches to things making your life easier on managing control planes. That's we've seen for some time, but I think we're even trending further. It's really no surprise. I think the point and what I'm trying to get across is that even those who adopted and are successful with it take on that complexity and we can all agree on that.
00:15:36
Speaker
Definitely. I think my next question is, OK, we can cover the challenges in a little bit more detail later on. But if not Kubernetes, what are the other options available? We know that in 2016, there were multiple orchestration platforms. And then I know all of us have been using slides, at least people who are in the Kubernetes vendor ecosystem, that Kubernetes has become the de facto standard. And it has won that race. But if not Kubernetes, what else? I think, to summarize the question,
00:16:04
Speaker
Yeah, I mean I'm definitely biased on this conversation, but you know One of the alternatives is cycle and we've been around 2015 kind of back when everything was a little bit crazy in the ecosystem I mean when we first started half of the problem was we had to explain what a container was in the first place Luckily, it's not so much a problem anymore But like you said Kubernetes did come out on top And I think a lot of that is actually we were solving a problem that people didn't know was going to exist yet at that time So we're almost a little
00:16:34
Speaker
too early with our kind of all in one solution that we come up with. So I don't know how far into it you want me to dive into with this question. Yeah, sure. Sure. It's a good it's a good interesting point that you know, in 2015, your company kind of existed. And during that time, we saw sort of the mesosphere sort of peak and, and, you know, I use that actually at a company,
00:16:57
Speaker
because Kubernetes wasn't the de facto standard yet. People weren't as keen to use it. And I think Mesos did a good job of reducing complexity. A lot of it was UI driven, all that kind of jazz. So I guess what makes something like Cycle different and how you see it succeeding where these others have kind of fallen off?
00:17:19
Speaker
Great question. I mean, we're still here today. So obviously we're doing something right. But I would say kind of our biggest realization when we first started, we were really just upload your containers and we run them for you kind of a platform. And we didn't we ran into a lot of issues with that where larger companies are like, well, I'm not going to trust this little startup out of Reno, Nevada to run all of my enterprise operations. And we kind of agreed with that. We're like, yeah, we really don't want to do that.
00:17:48
Speaker
And so one of the things we changed early on, I want to say like 2017 is we made it so that users can directly put in their API keys for multiple clouds, right? Google, AWS, what have you, and made it so that we will actually distribute our operating system and compute layer onto those different
00:18:07
Speaker
compute boxes that the client owns, and we will orchestrate and manage them for them. So it's kind of like a federated approach. We have the centralized platform, and then we push out updates. We tell their stuff to start when we need it to start and do all the other things. But at the end of the day, they own that infrastructure. And God forbid, we use this all the time early on when we were first getting started, but God forbid our whole platform went down, their stuff would still be running.
00:18:31
Speaker
Obviously, they couldn't start to stop it, but their stuff would still be running on their infrastructure, which gave a lot of peace of mind to people in the earlier days.
00:18:40
Speaker
Gotcha. So they still own everything. They own the infrastructure where your applications are on. So if they had to like, comply to any regulations, local regulations, all of those conditions were still met. So like, cycle is one of those options, right? I think the other option that has been around alongside Kubernetes has been hashing out Nomad. Like, have you come across customers that are using Nomad today? And like, can we talk about
00:19:06
Speaker
how Nomad offers a different experience? Yeah, I mean, we've come across a few people, I would say, that are using Nomad, but not very many, honestly. So, I actually haven't spent too much time using Nomad directly, so I'm probably not the expert to ask on that, but I'm welcome to discuss anything that specific. Gotcha.
00:19:27
Speaker
I think this question came in because I was watching a new stack video today for this podcast. SeedGeek actually uses Nomad and they use that orchestration layer to do both containerized apps and non-containerized apps deployment across different clouds. I think they are primarily on AWS. I just wanted to know if you have come up with customers who are maybe even looking away from Nomad or did you find people that
00:19:55
Speaker
agreed that we need to move away from Kubernetes and find an alternative solution. So that was the background behind that question. Yeah, I would say that we've probably found 50-50 people who are moving away from Kubernetes and people who are just getting started into DevOps, maybe containers, maybe not containers, and are looking for an easier solution than getting started with Kubernetes because that can be overwhelming if you don't know what you're doing.
00:20:20
Speaker
So I would say we have a pretty even mix of people coming in who are migrating off of Kubernetes because of the cost and the complexity and people who are just getting started and don't want to go down that route path or have started slightly down that path and decided this is not for me. So Alex, I know we had met in KubeCon. I know one of the things we talked about was sort of the concept of
00:20:44
Speaker
deploying applications in a low operations platform type
Cycle.io's Solutions for Container Management
00:20:51
Speaker
of thing. I know I brought up Heroku, which is one I really enjoyed back in the day. I ran a couple of sites off of it. I also had experience with Meteor.js. I don't know if you've ever used that. It was a very similar command line deploy the thing, but what I really liked about those was the simplicity of it. I sent this
00:21:10
Speaker
I sent this thing somewhere else and they took care of all the complexity of actually running the thing. And that was nice. But there are sort of a lot of things that Kubernetes has also gotten right, right when it comes to having the flexibility for organizations to really use either
00:21:27
Speaker
Add a small or scale i want a small but small or a very large scale things like rolling upgrades the csi standards the ecosystem around it so why would why would you see or guess what users are you seeing move away from.
00:21:44
Speaker
Yeah, I mean, you're right. Kubernetes has gotten quite a bit right. They obviously are able to solve massive scale issues. But, you know, they require massive scale teams. I mean, we got to remember that Kubernetes came out of Google to solve Google scale problems and not many companies, I would say, arguably have Google scale problems. There's definitely out there, but I wouldn't say that it's the majority by any stretch.
00:22:08
Speaker
So what we're seeing, and it's interesting you brought up Heroku because Heroku recently, I believe they've removed their free plan, right? And so what we've noticed are there's a lot of people out there who like the simplicity of Heroku, kind of like what you'd said, but the ceiling on Heroku is really, really low, right? You can't get a lot done. You can't build massive platforms like Heroku on Heroku.
00:22:32
Speaker
Right. So there's there's that layer that that it just can't go below. And cycles kind of a step down from that. It's it's a lot of the simplicity with the, you know, a lot of the the simplicity of what Heroku offers, but has a much, much higher ceiling because at the end of the day, it's just
00:22:48
Speaker
container orchestration and containers. So we see a lot of people moving away from Kubernetes who the complexity has gotten out of hand or they're moving from Heroku to something like what they want to be able to do with Kubernetes without actually having to use Kubernetes.
00:23:05
Speaker
Okay, interesting. So I think we spoke about how it's easier, right? But when we started with Kubernetes and that journey, everybody was focused on stateless application and finding that Kubernetes was the best place to run stateless. And then now we have crossed the chasm and people are running stateful applications as well. Is there a similar distinction or a different type of application that works best with some of these other platforms? What do you think about that?
00:23:34
Speaker
Yeah, I think that that's, you're right. We started out really hoping that things could just be done statelessly. And that's just never the case. A lot of the times you need databases, you need all of these persistent storage things to build any kind of normal application. And so it was natural that people wouldn't want to have two different ways of managing two different types of applications. I don't want to have my SQL database sitting on this server as a standalone application, everything else being orchestrated via containers. That's just, you know,
00:24:03
Speaker
It's nightmares trying to maintain all of that. The way that we saw that at Cycles specifically is we actually came up with two types of containers. You have stateless and stateful, and when you're going to deploy them, you make that decision. The platform treats the stateful containers as their own conceptualized, we're not going to auto-scale this, we're not going to mess with the storage on this. You're telling me that this is something that has very precious data that you want to maintain, and we're going to let you do that. We'll still run the containers and set up storage and do all that stuff.
00:24:32
Speaker
We're treating that as, for lack of a better term, the pet rather than the cattle. It's a very special container. It's got its storage and it knows what it's doing with that.
00:24:44
Speaker
But it still uses the Docker format for building my container image and maybe container D for running the container, or is there a different standard involved? Yeah, no, it's the same. We are OCI compliant, so we follow the open standard. It's not specific to Docker necessarily, but that's the tool that I'd say 90% of people are probably using. But yeah, it's just built the same way. You just specify what storage you're going to be attaching to that.
00:25:12
Speaker
Got it, got it. And now does this use host-based storage? When you say stateful, I'm just trying to get a little more detail there. I knew you were going to... We have to here. We have to talk about state. I got the four works guys here, I got it.
00:25:29
Speaker
Yeah, so it does use host-based storage, and there's a very good reason for that. I know we talked a little bit about this at KubeCon. There's definitely different paradigms, and with Cycle, since it's opinionated, and that's the difference between Cycle and Kubernetes set up,
00:25:44
Speaker
there's pretty much one way to do things and that's what causes a lot of the simplicity. Things just work together and they work that way. We use host-based storage and the reason being is that, let's say we had network storage or something else. We thought really long and hard about this. What happens when
00:26:05
Speaker
that network storage drops mid-right. I know there's solutions that can handle a lot of this stuff, but for us, it's a reliability and a consistency thing. If you've got a specific application like a database, that database understands that data. If we're just blindly zipping things up and moving it around for backups or splitting it across multiple containers, we don't necessarily know what that container is trying to do with that data.
00:26:29
Speaker
The database layer, the application layer storage management, sharding and all of that kind of stuff is much, much better and much, much smarter at it when it comes to managing how that storage is going to be split up. So for us, it was a consistency and reliability choice to use host storage. There's nothing stopping somebody from like cooking up a sand or something else like that. But at a platform level, we're going to support the features that are the most stable and the most reliable and things that are going to benefit
00:26:58
Speaker
80% plus of our customers.
00:27:01
Speaker
Got it. I know later in the show, I have a question here to talk more about the infrastructure, but in terms of storage, you say, how much control does someone have over that host layer? Like you said, cook up a sand, and that's what I'm keying it on, is could they bring an environment like that to the infrastructure that is being used, or is there a limited amount of administration that, since we're trying to get away from a lot of ops?
00:27:29
Speaker
No, I mean, absolutely, they can, they can, it's their infrastructure, remember? So we're technically deploying, like, let's say you deploy to AWS, we're technically just setting up a VPC in AWS that cycles VPC. If you want to connect it to something else, you can peer another VPC, you can connect your SAN. Obviously, those are more advanced use cases, and fall a little bit outside of what the platform is going to be super, super, you know, ease of use for.
00:27:53
Speaker
There's nothing stopping somebody from going all the way down. I mean, even our compute layer is literally cycle OS and our compute layer giving you full access. You can you can run containers in full privilege mode on your own infrastructure. It's your infrastructure, right? So in terms of support for complexity, you can go as deep as you want, but we're going to make it as easy as possible to do that.
00:28:14
Speaker
Interesting. So I think talking about your infrastructure stack, right? So you said once the user provides you with the access credentials for your cloud accounts, you will deploy an infrastructure stack for them inside a dedicated VPC sounds like it. And then they can, they have access to a UI or an API maybe to submit their container images that they want to run. Is that true? And then I'll move to the second question. Sorry, second part of our question.
00:28:41
Speaker
Yes, yes, that's all true. And they get to choose what infrastructure it is they're deploying through those providers. Okay, gotcha. And then so my second question is like, okay, if I and who manages that infrastructure for them? I know you said it is a federated model. But if it's running on EC2, am I responsible for managing the lifecycle of cycle OS running on EC2? Or how does that work?
00:29:02
Speaker
No, the platform will push all updates that are necessary to those to those boxes for you. It's literally just and that's kind of where the paradigm shift happens. Cycle ends up sitting as a layer on top of all of this federated infrastructure across multiple clouds, across multiple providers and manages all of that for you.
00:29:21
Speaker
through our platform. And so you don't have to worry about the updates. You don't have to worry about making sure that things are configured correctly. We deploy what you want and set it up the way that you want it set up. But we continuously keep that up to date for you. Okay, perfect. So then, like, I think now is my original question. Like, what does developer experience look like, right? If I have code in my repo, how do I actually deploy it on a cycle cluster or whatever we call that construct?
00:29:51
Speaker
Yeah, great. So there's a couple different ways. We have our user interface, which is kind of like the introductory the way that most people get started on our platform. That UI consumes our own API that we make public to all users. So you can do everything programmatically.
00:30:07
Speaker
Then we have different ways of building images and other things like that. Once you've deployed your infrastructure, Cycle can hook into your Git repos, and we have this concept of stacks, which is basically Docker Compose, where you can orchestrate not just the containers, but everything around those containers. What are their privileges? All of the really detailed nitty-gritty stuff the platform provides can also be done via this automated method.
00:30:34
Speaker
sort of like a CI CD, more the CD portion of CI CD, where say, I push a commit to main, you know, we can have GitHub actions or whatever, hit a hit a webhook with a with a secret token, that cycle will then go through an entire you can automate your entire, you know, anything, whatever happens, build multiple in multiple images
Exploring 'Low Ops' Platforms
00:30:54
Speaker
from Docker Hub from my repo, re image this entire environment with a whole new set of containers, etc.
00:31:00
Speaker
So everything can be automated and anything you can do in the UI can be done via programmatically as well. So that's awesome, right? Like if let's say the customers who maybe adopted Kubernetes but are struggling with the complexity and because they were only working with open source projects, they can kind of seamlessly move because you integrate with the same set of tools that they might be using for their build pipelines or their CI CD workflows. Exactly, yeah.
00:31:25
Speaker
Yeah, to me, I feel like this is, you know, just just from not actually trying it yet. I definitely want to. But to me, I feel like this is a is the huge part of the value for me and sort of the way I look at it, because, you know, dealing with Kubernetes a lot, we often think about this huge layer cake of things you have to think about. Right. And
00:31:45
Speaker
and certain providers, certain, you know, distributors of Kubernetes can give you some of those. But I feel like, you know, kind of combining a bunch of these in the control plane, and then sort of you on the infrastructure, but it comes with a lot of the CICD kind of stuff baked in where you normally have to sort of, I don't know, choose and kind of bring your own, which it does, does beg the question, right, where a lot of organizations, I imagine, have
00:32:09
Speaker
specific tools they like using right for these types of developer workflows can they bring their own or do they sort of have to tap into cycles.
00:32:22
Speaker
I mean, what do you mean by their own tool? Like for what specifically I would say? You said like GitHub actions, right? Can they can they use those directly? Or do you have something like that that's built into your platform? I mean, they can definitely use GitHub actions. The platform itself, the we call them pipelines on cycle side, they allow you to kind of orchestrate stuff on the cycle side, right? So you hit it via webhook. So any you don't have to use GitHub actions, anything that can hit a webhook basically can kick off that entire pipeline.
00:32:48
Speaker
Yeah, so for the most part, they can bring their own little notebooks. Okay, makes sense. And do you find a lot of, you know, I'm sort of thinking about the type of user kind of, I say leaving Kubernetes, it's probably the wrong way to think about it, but coming to use this type of platform, this low ops thing, are they still highly sort of integrated with APIs? Or do you find more of the users kind of tailoring towards user interface?
00:33:17
Speaker
A lot of them love the CLI, but a lot of people don't, too, I would say. I'm a DevOps guy, obviously, but I don't like memorizing CLI commands and stuff. 99% of my day is not spent just messing around in the terminal. It's using different various UIs and other things for ease of use.
00:33:35
Speaker
So there's definitely the crowd that loves being low level and having that control, which is great. And then there's the people who love doing the UI portion of things as well. So I would say that a lot of people will generally start in the UI for us and then move to the API. We're actually working on a CLI right now. We're even exploring the ability to basically just clone your Kubernetes cluster and move it over to Cycle and match exactly what they've done there.
00:34:03
Speaker
and move it over. So a lot of cool stuff we're still working on as well. We've gotten more feedback and more people are starting to ask for a CLI-like functionality for the platform, but it's kind of a recent thing, actually.
00:34:16
Speaker
Gotcha. Okay, so I think my next question is like we are trying to avoid the complexity that Kubernetes has, right? But what do you think about things like Azure Container Apps or some of those Knative kind of services where you don't have to worry about Kubernetes or something brand new like an Acon Labs
00:34:38
Speaker
If I'm just using that, I don't really care about what's running on Kubernetes. How do you feel about those kinds of solutions or alternatives compared to cycle? I know we had like, again, as we know, this is not a cycle podcast.
00:34:53
Speaker
about trying to figure out what else is around that can help organizations. So I was like, have you looked at that or what do you feel about those alternatives? Yeah, I mean, a lot of the time the, let's say the APIs, or I'm sorry, let's say the things that are built on top of Kubernetes are implicitly saying we are going to only do what, mostly what Kubernetes at the foundational level is capable of doing. They're all kind of doing the same thing at the same time.
00:35:20
Speaker
Right. So it's hard to build something easier to use that's built inherently on something extremely complex, in my opinion. And so you have a lot of the companies. And this is why I think a lot of the previous companies that wrapped around Kubernetes, there's not too many of them anymore because they all come out with the same things at the same time. Right. The features are all tied to whatever Kubernetes is doing. Sure, there's some sugar on top. There's some, you know, a couple different things that differentiate them when it comes to the big providers.
00:35:50
Speaker
Is there a huge difference between using Azure's Kubernetes service versus AWS Kubernetes service? And the only difference is what does AWS and what does Azure provide at that point? You're locking yourselves into those ecosystems. You're locked into what you can't really do in a multi-cloud easily without setting up distributed clusters and doing all of that kind of manually. So you're stuck with
00:36:13
Speaker
Okay, I'm buying into this ecosystem with this provider. What happens when their prices get jacked up way too high? Or what happens if, you know, let's say all of AWS is East Coast goes down or whatever. It's happened, right? During the holiday season is when US East One tries decides to go on a one day vacation. Exactly. And so I'd say that there's a lot of
00:36:38
Speaker
Yes, they make it easier. And it's great. But you're still getting hooked into that ecosystem. You're still hooked into Kubernetes in, you know, whatever, whatever Kubernetes is doing, that's what you're tied to doing. And whatever Amazon decides to support or Azure decides to support versions of Kubernetes. That's what you're stuck with, right? I think they only recently updated to the version of Kubernetes that came out in almost a March. Don't worry on that. But yeah, you're
00:37:04
Speaker
you're stuck to what they're doing, you're stuck to how they're updating and what versions of Kubernetes they're running, right? So I would say that it helps, but I wouldn't say that it's a complete solution to solving the complexities and the complications that come with a platform like that.
00:37:21
Speaker
Gotcha. And I think all three of us were at KubeCon North America last month. So while you were there, or even in your regular discussions, what kind of a split do you see between organizations that want to adopt communities and make it work somehow, even if it entails investing a lot of money and resources, to the percentage of
00:37:43
Speaker
organizations that are like, okay, I need to solve a business problem. I need to increase my developer agility. How do I do that? What's the best solution for me? For like just talking to people at KubeCon, again, this is a very community specific show. It might be an 80-20 split for me, but is there a similar split that you observe in the market or is there a different number in your mind?
00:38:04
Speaker
And I know I came into the Hornets nest here with the Kubernetes based podcast. I would actually I would say it's the inverse 80 20 from what you're saying. I would say 80% of people don't need Kubernetes and 20% do.
00:38:21
Speaker
When we were at KubeCon, I talked with a lot of folks who they're not necessarily tech companies. They're financial, they're healthcare, they're several different verticals, but none of their core competencies are tech-based necessarily. And so I talked with several who are spending hundreds of thousands of dollars managing and operating Kubernetes who are at KubeCon trying to figure out, how do I do this better? I've got to be doing something wrong, but they're not alone. There's so many people who
00:38:51
Speaker
They don't want to manage Kubernetes, they want to deploy their apps. They want their apps online and they want to run their apps and they don't want any downtime. The infrastructure and everything else that comes with Kubernetes is like a secondary thought to them, and it's a cost thought. It's a financial-based calculation. How much is this costing me and how much is it going to cost to not use this? And so they're at KubeCon looking for other alternatives. If not,
00:39:17
Speaker
you know if not something that's different from kubernetes how can i make kubernetes do what i needed to do without spending you know hundreds of thousands or even millions of dollars let's be honest for some of the larger companies out there millions of dollars per year
00:39:31
Speaker
I think I clearly saw bigger organizations having issues hiring the right folks with the right skill set. So like, not everybody has been able to hire communities experts on their teams. And if they are, they are spending a lot of money to get those resources. So I think people listening to this podcast, you're in the right place, right time. Yeah, definitely. Definitely. But yeah, I would say that the 80 20 split is more 80% of companies do not need this 20% would.
00:40:01
Speaker
Yeah, I think I wouldn't necessarily disagree in most cases, right? In the sense that I definitely got the sense that there was a push towards
00:40:12
Speaker
still managing a lot of the complexity. And a lot of the ecosystem is very strong in that we're layering things on top of Kubernetes in this ecosystem that do aim at making that 80% more amendable towards using something so complex. And I think the 80% where I see it is the people interacting with
00:40:36
Speaker
The 80% of people in our active careers don't need the complexity. I mean, they're still at 80% that are sitting in like, you know, certain platform engineering that, you know, do benefit from that, having like the fine grain scale and things like that. But yeah, I definitely, I would, I would kind of agree with that in some cases. So, um,
00:40:55
Speaker
I want to switch back to this LilOps term. I know it's something you've used on your website, but I think maybe our listeners could benefit a little more to understand what that actually means when we're talking about
00:41:12
Speaker
managing things like the underlying infrastructure so maybe you could walk us through like a day in the life of setting up something with a low ops platform like cycle versus maybe doing you know doing some kind of touch point comparisons to the Kubernetes world and sort of.
00:41:29
Speaker
Give us a little insight on what underlying effort do you use? What are some of the challenges along the way? Those kind of things, if that makes sense. Yeah, yeah. Well, I would say the concept behind the low ops platform is, think of how easy it is to get something set up and running locally on your machine, right? The dream. You got one machine to worry about, you've got
00:41:50
Speaker
you know you run a couple commands and boom things are running things are stable and everything's working that's awesome what if we could transfer that to the cloud what if we could take that experience of i know what i want to do i know what i want to have happen.
00:42:04
Speaker
and just make it so right into the cloud. And so the low ops platform, we use that terminology because it's, you know, obviously you need to know a little bit about running and managing applications. You know, there's especially if you're a larger company, you have a lot of stuff to do, you need, you know, probably a few people who are knowledgeable on that, but you don't need an entire dedicated team.
00:42:24
Speaker
of very high skilled engineers just to keep your applications running. The platform is going to make, you know, a low off platform will make all of those underlying decisions for you much, much easier and make it a lot easier for you to manipulate and manage and do what you would normally do with a more complex solution or something you'd have running on your own machine, but make it easy and just work across the cloud.
00:42:51
Speaker
Got it. Makes sense. And so, you know, we're there as you've been around since 2015, what were sort of maybe some of the biggest challenges you hit between then and now and maybe what did you learn from them?
00:43:02
Speaker
Yeah, great question. I would say, one, businesses don't want to trust startups with their really critical infrastructure. Totally understandable. I would say we definitely learned that there's a fine balance between simplicity and power and capability. And it's really hard to get that right. It's hard to have something where you have a lot of control in a really easy way.
00:43:30
Speaker
And so that's kind of the underlying challenge that low ops platforms that are competing with Kubernetes are really trying to solve is Kubernetes does a lot of things right because it does a lot of things, right? There's a lot of pieces and a lot of parts that you can put together to get kind of what you want. How do you take all of that and make it so much easier to do where you can think more about the angle, more like a declarative
00:43:52
Speaker
Thing for the programmers out there, right? I want a declarative way to set all this stuff up in a way where I know what it's going to be. And I don't want to have to think super hard about how it's all going to interact together all the way down to like the operating system, right? Because running Kubernetes on, you know, Linux works great, but there's other people who are trying to run containers on like Windows and other things like that. It gets really complicated and you have to make sure all the pieces work together. Perfect. Yeah.
00:44:17
Speaker
Yeah, I definitely see sort of the future and I don't know, you know, we talking with a number of folks, there's I don't think there's one thing that's going to win out and that makes total sense. We're going to live in this world where there's going to be certain workloads that run certain ways on Windows and VMs or whatever. And then also these other things. So having that flexibility is certainly
00:44:38
Speaker
I think something a lot of people
Future of Hybrid Cloud and Low Ops
00:44:40
Speaker
want. We see that today with Kubernetes doing lots of things. We're already seeing people saying, well, I like this, but can you put my VMs on? I understand that complexity piece while trying to give, it's almost this fine line. I feel like Kubernetes walks this fine line where it doesn't want to be the old open stack of the days of doing way too many things and then people are just like, my head hurts, I'm leaving.
00:45:06
Speaker
But then I feel like this low ops piece has this also has a fine line of, you know, leaning too far towards the Heroku side of like, don't take too much away, because we still want to do some fun stuff, right? Or more interesting flexibility stuff. So I, I am intrigued by it, to say the least. So a couple more questions before we wrap things up here is, you know, what, what's not solved? And where is it? Where are you going with low ops, right? So what, where, where are the challenges in the future?
00:45:35
Speaker
that you're sort of targeting sort of personally or as a company, I'm interested to hear what that really looks like. Yeah, you know, we've seen a huge movement and this is also probably a little hot take. We've seen a lot of people moving off the cloud in back on the on-prem. We've seen a lot of people moving into colo and hybrid cloud.
00:46:01
Speaker
That's something that we've very recently started supporting because we've seen this huge migration away from trusting other companies with your precious stuff that you're running, or larger companies who already have a lot of this stuff.
00:46:17
Speaker
already in their data centers or in their, I don't know, their closet, right? Who are now moving into the container orchestration space and setting that up and managing that on local stuff is hard as well, right? And so, you know, we've started solving this problem because this is where we see the future going is much more hybrid and on-prem
00:46:39
Speaker
And so we actually, at Cycle, we just released, we're calling it the IAL, the Infrastructure Extraction Layer, where you can basically prop up an API with a certain pre-registered list of endpoints that our platform will call out to in order to provision IPs, provision your infrastructure for you.
00:46:57
Speaker
so you can create a bridge into your colo or on-prem solution that the platform can interact with, just like it interacts with any of the other providers that we support, so you can get this true hybrid cloud experience, which is really cool. We've seen a lot of stuff moving in that direction recently.
00:47:13
Speaker
Yeah, I mean the the move I think to bring compete where data is to talk about edge, right? There's still you know, there's definitely a lot going on there I think you know, we could spend a whole entire podcast talking about that but in light of time I do want to give sort of our listeners who may be intrigued by the concept of low ops and what cycles up to you know Where can people get connected with you first of all? Where could they learn more about cycle itself and what you're trying to do or even to get started?
00:47:42
Speaker
Yeah, our website is cycle.io. And you can definitely get signed up and kind of start on your own. But I'd highly recommend for somebody who wants to know how this is going to solve their problems to reach out and schedule a demo with us. My email, Alex at cycle.io, reach out directly and we'll help tailor the solution to exactly what problems you're trying to solve and see if the platform is going to be able to work for you.
00:48:08
Speaker
So I highly encourage people to go check it out and reach out for anybody who wants to figure out how to get it working with what they want. Awesome. Well, this has been awesome. Thanks for coming on the show again. I think this concept is something I've been wanting to discuss a little bit more and probably not the last show. We'll talk about it just because I think it's definitely a trend and interesting point of conversation in the Kubernetes community around the complexity and managing that complexity.
00:48:35
Speaker
Whatever that mean or game to talk about it. So thanks for coming on and addressing some of this as the first time. So appreciate it, Alex. Yeah. Thank you, Ryan. Thank you, Bob. And appreciate it.
00:48:47
Speaker
All right, that was great having Alex on the show. I know what they're doing at Cycle.io, and I didn't know they were around originally from 2015, given that they weathered a lot of this turbulence of de facto standards and other orchestrators. I think it's really interesting to hear his perspective on the complexity and where users are going.
00:49:09
Speaker
I don't know. Give me your takeaways that you had from this conversation, Baba. No, I think I definitely agree. Alex had an awesome perspective of why you should only do things that actually add value. If something is going to cost you a million dollars, but only you end up producing half a million in terms of revenue, that's not worth doing. I think my first key takeaway was trust.
00:49:33
Speaker
do what's best for your organization, do what's best for your applications. Everything now is being built as microservices, built on Docker containers or just containers in general. Then the way you deploy it, it will depend on what platform your platform engineering team has for you. If they are willing to take on that Kubernetes burden, yes, go for it. If you are a really small shop, really small organization that is still doing some in-house development, but you don't want
00:50:02
Speaker
all of the overhead that comes with Kubernetes, choose for something else, right? There are different options that are available to you, cycle being one of them. I know we discussed a few during this episode, things like HashiCorp, Nomad, smaller communities, similar kind of features. And obviously, HashiCorp has a lot of influence in that community. Amazon ECS, I know of bigger enterprises that don't want to move to EKS. They are like, I'm confined to AWS, I would rather use ECS because I've been doing that for years. So there are different options out there. So choose wisely, I guess that's my point.
00:50:32
Speaker
Yeah, and to your point around do its best for the application, look at your application first, right? You know, something like Heroku, which I mentioned, is very 12 factor, right? Data databases and things like that, we normally talk about in the show, usually sit outside so you can, you know, but it gives a lot of value to easily deploying the thing and leaving a lot of that management away. But what it doesn't do, right, is give you a lot of flexibility to manage that underlying infrastructure.
00:50:58
Speaker
And then sort of the other end of that spectrum, Kubernetes from scratch, so to speak. And so I think the reality is, what is the current state of what your application is? Does it need that level of complexity and scale?
00:51:17
Speaker
Yes, there is some thought of where will you go with this application? Meaning, do you plan to scale and scale quickly? Is there a place you can start? Do you have a small team? I think that's a big part of what I'm hearing from low ops is maybe there's a benefit to bringing on this type of control level type of solution where you can hopefully get away from doing a lot
00:51:45
Speaker
Of that sort of management of the online infrastructure and be able to kind of deploy these things by giving access and trust right of course there's a big trust component there. Yeah think about what the application actually needs where will go what is it look like how what's your scaling look like right.
00:52:03
Speaker
maybe in the next few years. Again, being an advocate for the Kubernetes ecosystem in general, I think if you are just planning on using the open source components and trying to put together all the different LEGO pieces on your own,
00:52:17
Speaker
That might be too much complexity, but there are vendors in the ecosystem, right? I know at KubeCon, you had hundreds and hundreds of sponsors in the Solution Expo trying to talk about how they can make a specific part inside the Kubernetes ecosystem easier. So again, we're not saying that you should completely move away from Kubernetes because it's complex, but
00:52:37
Speaker
see how you can make your experience easier. And there are alternatives out there if that's something that you want to check out or look for options. We're definitely not saying that at all. More or less, there's a lot of instances where I think technology being as popular as it is and being a sort of standard of container orchestration
00:52:58
Speaker
It's easy to just say, well, let's just use Kubernetes. But often, that will actually introduce a lot more headache in smaller scale applications where it may be too much for exactly what you need right now, which is fine. That's a fine conclusion. And there may be a time when you give yourself a little more time to dig into everything that Kubernetes gets you. That's really the point.
00:53:27
Speaker
I guess a takeaway that I'll bring out of this is the idea of a common control plane. Having the ability to be able to manage infrastructure, deploy applications,
00:53:39
Speaker
to various clouds and to things like on-prem and certain locations. I think that part of it has always intrigued me in sort of all my time and career in this space of being able to maybe do this through something like analytics and a common control plane where you kind of tell
00:53:59
Speaker
this layer of you know hey here's my application here's here's what it needs go figure out where it needs to put it in this world of compute and networking stores that we have let me let me not think about nor do i want to think about it right just in a way and i'll pay for that.
00:54:15
Speaker
You know, that ability to keep things online wherever it is, right? I don't actually need to know. So that whole part of it intrigues me. I think that's something that we've also seen in Kubernetes communities, right? Being able to kind of abstract and federate where your applications are running in Kubernetes. But that whole concept, I think, is something I think Alex talked about a bit as well, and something I see a lot of value on in the future, both
00:54:42
Speaker
both in Kubernetes ecosystems and in other application deployment systems. For sure. Cool. Well, I know it is November 23rd, so everyone, if you are celebrating Thanksgiving, please enjoy the rest of the weekend and thank you for listening. And one of the things that we wanted to reach out to our listeners this week was we have an idea
00:55:09
Speaker
around putting together sort of a community-based episode, meaning that it'd be great if you had something that maybe you want to talk about that you've been waiting for us to address or maybe you haven't sent in a comment or haven't reached out to us, but you want to talk about it is we would invite anyone listening to send a clip, maybe max three, max four minutes, something like that around a topic.
00:55:35
Speaker
that you want to discuss or maybe a tech tip or whatever you want to call it. We want to put your ideas into an episode. It might be cool. We don't know when this will happen. We have to get enough submissions, but we want to start advertising it now. We will probably in the next foreseeing episodes until we get some more submissions and we'll see what it brings us. But we'd like to bring your voice
00:56:00
Speaker
to this show, and let us know what you think about that as well. Bobby, do you want to edit anything to that? Yeah, and you can find us everywhere, right? So on Twitter, if you want to send us a DM, it is at Kubernetes Bites. LinkedIn, you'll have to reach out to me and Ryan, because I don't think we have a LinkedIn account with Kubernetes Bites, but our email is open. So please don't spam us, but it is kubernetesbites.gmail.com.
00:56:22
Speaker
So share your recordings with us. We'll see if we get enough. We'll put them together in one single episode or if like we, if it takes some time to gather, we might introduce clips here and there, but we just wanted to make this more interactive, right? Like let's hear from people that are our listeners and want to share their thoughts with the rest of the community. Absolutely. All right, Bobbin, this brings us to the end of today's episode. I'm Ryan. I'm Bobbin. Thanks for joining another episode of Kubernetes Bites.
00:56:53
Speaker
Thank you for listening to the Kubernetes Bites Podcast.