Can we build Open Source ecosystems (K8s+) [Cloud 2030 March 18 2021]

00:24 - I saw something yesterday that I haven’t been keeping up. k three s.

00:33 - Oh, yeah. Kubernetes light. I use that. Yeah.

00:39 - Okay. That’s what it is. I was wondering, that’s the first time I saw a mention of it.

00:44 - There’s at least two, maybe three different Kubernetes lights that are popular right now.

00:49 - Right. There’s three. Yeah, I’ve been tracking. Because the AWS distro that they released is a light is a is functionally a light distro. Also, it’s opinionated. It’s not exactly light. It’s the difference. Yeah, there’s k three s, which was rancher and now Susa. And it’s been quieter since since they came out. We I used it I have some some automation on it. That’s actually pretty slick. For the Raspberry Pi lab that I do, and then I test it on the clouds.

It’s it’s very usable. And then there’s micro Kate’s I think they call it Yeah, micro. Yes. Yeah.

01:39 - And that one’s mirantis is behind that. But somebody there is not just mirantis. Like brantas is promoting it, which is their mo but there I think it’s actually by a different a different group. Originally, I’m not familiar, I could easily go do a lazy web search, but I don’t know off the top of my head.

02:04 - And then the Amazon, I was surprised because I saw the Amazon thing come through and that distro is, is being presented as a micro Kate’s alternative. And that doesn’t include like mini cube, which is the developer version that runs all all in one all in one on a desktop.

02:24 - There’s like eight of them now. Wow. Just think it’s Google’s single node Kubernetes cluster.

02:34 - So are we saying it’s not bad enough that you can’t share between standard Kubernetes, the different distros. And that now you can’t share between Kubernetes and Kubernetes light or Kubernetes? light to Kubernetes? light? I actually think the distros are relatively compatible, right? Because there’s there’s a API test that makes them work pretty pretty much the same way that but I’ve heard I’ve heard that from a from a get up to speed for developers.

And I’ve only heard this anecdotally from a few people, but that there’s a pretty steep curve. Like if you’re using you familiar with openshift. And you decided to go use somebody else’s that’s a pretty steep learning curve to get up to speed.

03:24 - Oh, like the differences? Yeah. I think openshift is that’s more true for openshift than other products.

03:34 - So maybe it’s specific to open shift versus alternatives with micro K and s other than starting and stopping the cluster. I mean, there’s really functionally no difference.

03:50 - Yeah. can kick. Yeah, I can’t speak to micro cades. But k three s is basically Kubernetes code base, but they made some they hardwired in a couple of pieces. So it’s it’s not really a fork, but it’s sort of a build decision. There’s a couple of things that they are cleaved in but yeah, it’s basically the same code job for micro. Kate’s it’s the same thing, right? That’s, yeah, basically Kubernetes just packaged for smaller footprint, right? Yeah, it was basically done for developer desktops, right.

So you don’t want to farm an entire cluster, they get these things done. And then you have mini cube which was the original one was by far the most painful to install. It was not a full It was not a full Kubernetes implementation. So there’s a whole bunch of things you could not do in May cube. So with micro K and S or K three s right, it’s pretty much a full on cluster. You’re just running a control plane on the same node.

04:56 - And you will in my case, three is eliminated at CD which I think was a good call. So they, they switched the backend. So you didn’t have to worry about that part of the bring up. And then they put all the tools into one binary. So you don’t have to install six binaries, the server and the controller, and the client, even even the cube cuddle are all in the same binary.

05:25 - That made it microcampus. But I would say that the microsuede s, make the challenge of microcat ABS is when you start looking it up, you put promethease on, you put all the other services on, you know, your desktop slows down. Let’s put it that way. Yeah.

05:44 - Yeah, but I mean, is, you know, I’ve also heard that concern, and I’m, I’m trying to be fair here as someone who is openly and regularly, not hateful, but not appreciative of Kubernetes in the enterprise.

06:05 - I am, you know, still, you know, sensitive to the fact that there are many areas that I’m not that familiar with, and the folks that that I’ve been talking to, and again, it’s a relatively limited group, I wish I had a regular, wider audience of potential adopters in the enterprise. But most of the people that I know, that are using Kubernetes, on a regular basis are using it at scale, or they’re using it as an Amazon distribution on Amazon, or, you know, Google.

06:49 - But it just, it seems like the one of the biggest concerns is the inability to manage anything inside Kubernetes without adding significant difference, additional tooling or solution sets, like Splunk, or promethease, or something like that. And is and I guess the reason I’m asking whether that’s fair or not, is when of vSphere was two or three years old? How much monitoring Did it have on VMs? I frankly, I just don’t remember, I remember we had a lot of gaps when we were first building VMs, based on VMware, at gillean.

We had a lot of gaps, like you put a VM in a machine and you couldn’t even find the VM, you know, things like that. So it was and the ability to balance memory and CPU utilization were horrific and and determining how much a particular VM was using versus others, or whether or not you were efficiently utilizing the box as a whole. Where real problems bases in the first two years do we see Kubernetes ever doing any of that itself? Or do we see that Kubernetes is always going to have to have three or four bolt ons in order for it to be a comprehensive package for actually not just delivering and distributing applications, but also effectively managing the resources and the risks that are supporting those applications.

08:31 - I’m happy to take a shot at answering the question and see where other people and the design that the community around Kubernetes has wanted is that Kubernetes is a is a kernel.

08:46 - Right? And so that is that drives exactly that what you’re describing, so that there will be by design is a growing body of add on extensions, capabilities and components? Sort of answering those questions rather than it coming out of, you know, come out of the cncf perhaps, but it won’t, it won’t, it won’t get added into Kubernetes as a as a product. Which I actually think is one of one of my concerns about the the product, the Kubernetes systems, right? I mean, I guess if you this is where Red Hat openshift is not Kubernetes.

It’s an ecosystem curated by Red Hat around, you know, that makes Kubernetes into a product.

09:36 - Right? Well possible. So why Red Hat is so vociferous in their refusal to support anything that is in Red Hat top to bottom. If If you want their support, right, this is what makes it very hard for a company like RackN to be in the Kubernetes space is that the people who are doing Kubernetes distros are typically lining up around Those types of statements, right? And so they need they want to own the account top to bottom.

10:05 - Whether they can or not as a different question, but they that’s that’s it’s funny.

10:12 - Right? Sorry. It’s no different than Linux.

10:19 - I think you’re right. I’m interested in your expanding on the comment.

10:25 - Yeah, actually Me too, john. I mean, I have a couple of thoughts on Linux as well. And I’d love to hear your take.

10:34 - Well, when you got Linux, he didn’t get all the management utilities around it, right, you still have to deploy monitor yourself, well, the other pieces to it. And in the case of Linux, you know, Red Hat did the exact same thing, they became the person to wrap the tools around at the provided supported distribution around it.

10:52 - Right. So I think I look at Kubernetes and say it least it comes with something unlike Linux, right? It’s a much more complete orchestration system than just a simple Linux server where you have to go add the orchestration to it. But yeah, I think that that’s not where they’re headed, they are very much trying to be an orchestration platform, not a comprehensive solution. They’re happy, let’s see ncf trying drive out in the best of breed in the ecosystem.

11:23 - That’s, I think one of the things that we need to see happen on that is, in the Kubernetes, space, I’d actually love to get back to the Linux space. And you know, we’re talking about open source effectively, is you’re getting distro, vendors who are pulling together bits and pieces of, you know, open source projects.

11:43 - Right now, a lot of those projects are single companies like a company that does one thing in it, when what customers actually need is a spectrum of components to make all this stuff work. And so I think it’s it’s, it’s works as an open source community, maybe to have somebody be like, I am the, you know, open security platform around Kubernetes, and open source and get a whole bunch of adoption. But it’s ultimately a product that they’re selling cross planes like this.

We’ve talked about cross plane, some where there’s a company, clearly behind cross plane, doing a lot of that work and putting in the open, and hopefully it gets pulled into Kubernetes. But that makes it hard then for other companies to embed that, that product, that component into their platform. This is this is what I want to try. And I’m trying to say this concisely. We’re in this balkanized world with open source projects were around Kubernetes, which is predominantly major vendor backed the, there’s a whole bunch of small people solving small problems as individual companies that become essential in the Kubernetes build that then, if you were going to build a workable stack of Kubernetes stuff, you’re going to either be crossing a whole bunch of vendors to build your stack, or you’re going to have a vendor who has to then own stacks of things managed by other companies.

Even though they’re open source, they’re still really managed by other companies. And it strikes me as a very difficult thing for companies to navigate. users to navigate and vendors to navigate.

13:24 - Sorry, Rob, when you say companies managed by other companies, are you talking about this? vendor sources? I’m trying to gather projects or vendor projects. Okay. So the or whatever the organization is that takes responsive editorial responsibility for for this small tool or the component that’s being integrated? Right. Okay. There, there been a couple of places where I’ve, I’ve seen like, in the observability space, you had two startups who managed to collaborate pretty well, around the spec, and then that got adopted.

And it seems like the observability spec turned into a pretty good collaboration around this. So I’m not to me, it’s it’s not a always a problem. But there are times when are you referring to in in the observe abilities? I know honey come I’m trying to remember the other company.

14:22 - Should honeycomb honeycomb as an example, I think they’ve done they’ve done. They’ve done an amazing job on the on branding and marketing around observability. But, but they pulled something back into Kubernetes. That became an observability standard run around Kubernetes as at the monitoring level, and then they’re providing a service. Don’t you don’t you foresee someone doing much the same with Kubernetes is based on some, you know, one of the long poles or maybe multiple long poles in the tent for the use of Kubernetes basically doing, uh, doing a good packaging job, the way honeycombed.

15:06 - But here’s the here’s my thing with with what honeycomb does is, I think they did it right, there is an API that will ship logs to a observability. Service. Right? So they open source the client, if you will, not the service and the back end.

15:27 - Yeah, it’s well, and that’s smart for a number of reasons, mostly because their customer base that really needs this has already made a choice regarding how they want to consume login information, and, and do the analytics on it. So trying to trying to come in with an opinion as to which of those you want to jump on, you know, basically restricted their, their field of operation, let’s put it that way. But I mean, to me, eventually, and probably sooner rather than later, the observability question is just part of an application, performance monitoring.

component. Right. And so the idea that I have, that I’m going to piece together, my platform out of all these different vendors, doesn’t strike me as the as, as sustainable, especially for most people use well using.

16:33 - Yep, I think you’re right, you’re absolutely right. And, you know, I, I keep thinking of it, and I use the image of, you know, an end user being, you know, kind of committing to something with Kubernetes. And, and being surrounded by a swarm of mosquitoes, or these kind of little, little pesky.

16:56 - And, and bits of operation that they have to have to deal with in order to make use of it.

17:02 - And the the integration or the the unification of it is up to the end user. And that’s a bad, that’s a bad choice.

17:14 - So sorry, rich, go ahead and finish. Yeah, so I guess my question is, who’s going to do something? Well, if we’re looking at some things other than observability, if you were looking at for someone who was going to do the, the confluent of, of Kubernetes, you know, the way confluent, you know, took on Kafka, Bhaskar? I think that would be the that would be the, that would be the thing that I’d be looking for from a commercialization point of view.

And the question there is, does anybody have a leg up on that kind of an offer? But if you have equity, open source, right, so yeah, so we’ve been working with and so one of the things we started with the notion of saying someone needs security that out, and there needs to be some ability to customize it. And I can tell you that it’s a pain in the butt. But when you start trying to put your your mosquitoes in, organize them, and everything’s moving underneath you, keeping a coherent, tested stack is no trivial, no trivial Feat.

But when you do get there, you now face the question of saying, Okay, well, can I actually open source? That’s right. And and the answer is, it’s really not from an investable. No, that didn’t, or they’re not going to take Apache two anymore, you have to go into a restrictive licensing agreement.

18:50 - And that’s what we’re kind of saying, because we tried to curate the entire development stack through all the deployment and operation side.

18:57 - And I think we’ve done a decent job of doing that. But it’s a significant engineering feat to keep that going. You know, in our model was, you know, originally saying, let’s just make a new LAMP stack where everything’s curated a ecosystem. If people could just adopt it, you get community to maintain it, which I think is the right thing to do. But I think from a fungibility perspective that no one wants to see Apache do I think elastic kind of scared everyone off? Scared the vendors off? Sure.

scared that scared, but the people writing checks off? Which I don’t think it should have.

19:33 - I mean, they’re still doing, you know, they’re making 80 plus percent of their revenue on SAS, right? They still found a good way to make money. There’s still like a billion plus valuation company into it. But from a defensibility perspective, right kind of stuff.

19:51 - If you don’t keep core your technology protected somehow. It’s just not popular right now.

19:58 - I mean, this is I think that’s true. What’s happening? It’s been happening with investors slowly for a while. I mean that they could see if they’re SAS, they have no reason to actually release their code. Right? I mean that. They don’t they don’t have to, they could just stop making commits to the Apache project and maintain an internal fork. what’s what’s their advantage to actually continuing to publish their SAS code publicly? And maybe why they suspect they don’t do all of the like their operational components.

I don’t know if they publish them or not. But why bother? Well, for me, and I can’t talk for anyone else.

20:42 - But why bother is good to build a team, when you’re done putting into think of just all the orchestration pieces you need to put together to create a coherent managed system? Right. So we talked VMware, we’re talking a pretty sizable engineering team, to maintain what is mostly a proprietary system, right, good enough to deal with making changes to it. Right? So the advantage of open source is when you get a community, you can actually start supporting things, you can leverage the community, right, without having to build up an entire engineering team, wherever.

I would say that was, you know, the thought process for me it was to read it, but stupid reality.

21:22 - I mean, to do you, I mean, this is we just giving our history if you don’t know it, right, we had started our work is open source, open source, basically, the core. And we found that where people wanted to contribute, were the pieces that we had proprietary in the places where they never touched with places we had open, and we flipped our license. From that perspective, but the consequences, you can’t use our product without a license, it’s it’s a licensed product, no matter how much of the code is open at this point.

And it didn’t, doesn’t impact I mean, it impacts the people who want to come by and not pay for it and pay for the product and company IP, but it doesn’t.

22:10 - Isn’t that kind of the point? Aren’t you in exactly the point that you want to be where you, you’ve got a situation where you, you are a commercial? You know, you are not a philanthropic organization? Rob? And so, you know, there is a this is this is a commercial, this is a commercial offer, and what that what is the value that the customer gets for, you know, paying, paying for the license.

22:39 - The thing, the challenge that we see is that there are the value prop in open source is supposed to be this community effort. And, and from our perspective, people, because and we saw this happening, there were people who were using our software for 1000s, of running production data centers in the 1000s of units. And because we couldn’t write software that was fragile, and required consulting, it, there was no, there was no monetization drive for those people, and they didn’t want to participate in help either.

And maybe we should just get over that. And so it provided the public good. But that wasn’t that that that was that didn’t work. It didn’t work in the places we’ve seen in, in industry across the board, right, I was gonna say where where have you seen it? So what Rob’s model is, is the old Unix model, where it where you provide the the core aspects of it, and then there’s a community that puts in the stuff that they want to contribute.

But what Rob was say, so you’ve got the user groups and everything else along those lines. And it’s, again, the Unix model as opposed to the Linux model. got robbed thinking.

24:03 - I was in some ways the Linux Yeah, there are so many distros of Linux, I think there’s two Linux models is the reason why part of what I’m thinking Rocky, because I think the Red Hat canonical Susa Linux model is different than the BSD. And, you know, every other, you know, Alpine and all these other models where there’s almost to Linux is from that perspective. Is that Do you agree with that? Yeah, there it’s, it’s the way in lots of ways it’s the Linux Foundation model as opposed to the open source Linux model.

25:00 - because no one’s willing to pay at&t licensing fees.

25:05 - And so another paying Red Hat licenses instead.

25:08 - Oh, and that’s so I don’t know if this is a worthy of conversation or not just throw it out there. But the the idea for open source beyond what Rob already pointed out, is effectively free consumption. Right? I mean, that’s what it boils down to, is that people assume they can download something and start using it like they did with with the base of Rob’s platform and what some people try to do with Kubernetes. And, and what many people do with Linux, but is his open source of success, if in fact, the only time that good luck, quote unquote, and I’m using the quotes very seriously here, because I’m trying to deny the idea that they’re actually open source, that the product is only really usable? When it’s configured and sold by by a vendor, like red hat or Microsoft or Amazon or Google? Is it still open source? I mean, so what that it’s that its base is open source, if the only way a good enterprise can effectively consume it, is by consuming it through one of those vendors, otherwise, the overhead and difficulties are too high, to surmount, then does the term open source even a flop apply? Well, why don’t we take a different? Excellent question.

26:33 - Why don’t we take nginx? Right, that’s an open source product. It’s had massive wide adoption without licensing. Right? It is massively used.

26:45 - And then you go to the end there, they’re contrived, you know, pay for our paid version, you get these additional features, additional features, you get our preset. Right. But it is a valid open source model. And what they’re really driving in that is not licensing for the product. They’re, they’re charging for support for the product. Right, right. So you’ve got three open source models you charge for support.

27:11 - And I argue nginx has done a very good job of doing that. you operate it as a staff, so they don’t have to use it, and then you charge for consumption fees into it. And there’s a third escaping my mind right now.

27:27 - Eric, dirt calls and it’s an area, you know, it’s it’s the exact same model, obviously, it doesn’t have the same kind of widespread adoption in the market yet. And it’s much newer, but it’s the same idea. The product is super easy to use for 80 90% of the usage models need Cynthia’s benefit on top of that, his support and complex implementations? Yeah, it’s certainly I think it was open core was the third model.

27:57 - Yeah, open open core would be the product works, but you’re going to need to, you know, buy something critical to use it in production.

28:06 - Maybe a little bit when I go back, they think think about Hadoop is another example. Right? Now, it’s extremely widely used. And and, you know, there’s a number of people that have, you know, put it into SAS services to charge for support components. But you can’t argue open source wasn’t effective, you would not have the amount of data processing you have today had Hadoop been a licensed product. Right, if I had to go write a check for $100,000, you would not have widespread data analytics.

28:50 - I’m thinking about that. Because Hadoop. Hadoop is interesting, because you definitely had to vendors, right, who were duking it out early.

28:59 - And to make a dupe work, you actually had to write pretty big checks just to get the infrastructure in place. It wasn’t right to do it, it wasn’t just a install some software in the cloud and and get data out you had to make an investment.

29:16 - So right. But that’s one of the one that’s the older model of that’s Richard Stallman his model of free software and that you pay the consulting fees to get the work done. But the product is open to the the work product is open to anyone and everyone. So large companies with money, pay for the work and everyone benefits once that work is out there. And that model work for a lot of the Apache and whatnot. Cygnus, the company But he actually was very successful for quite some time as they built compilers and operating systems and stuff for telco where telco paid them to do it.

And then it became open. And then all the telcos got to use it. And I think at some point, the telcos probably decided who was going to pay up this round.

30:26 - Oh, but I mean, I along those lines, I’m thinking like react and go, Lang. And there’s there’s some very successful open source.

30:37 - Yeah, who paid for exactly like what you described came out of? Yeah. Okay. Google, React came out of Facebook.

30:46 - MySQL came out of sun. Right? Yep. Hadoop came out of what’s it Yahoo.

30:56 - Yeah, that’s right. Yeah, I think so. That’s exactly it’s, yeah. So that that’s a very traditional model that we’re not seeing as much of. And redhead is kind of the anti pattern for that. Especially now that IBM owns them.

31:19 - I have a curiosity. Why do you say that? I mean, I think they’re kind of a poster child for their charter support.

31:25 - Business, they say, I say that, because they’re, they’re actually grabbing other people’s ideas, and then locking everyone into their model of the idea of watching them in OpenStack. It was kind of Yeah, well, OpenStack, it was an interesting thing where they didn’t have a strategy. And then their strategy was let everything mature to a certain point, and then buy up all the talent that created the mature product. And now that IBM has that, it’s, it’s the, like Rob was saying the opinionated.

If it’s not in the redhead stack, we won’t play.

32:14 - Yeah, I mean, that, john, and I know that all too painfully through our work with Erickson. I mean, I can’t tell you how many meetings Erickson had with, with Chris, the CTO at Red Hat at the time. And they kept asking the same thing. Okay, we’d like to do you know, this part with Red Hat, but we want to be able to put our stack in the middle and Red Hat kept saying no.

32:37 - And they kept going back and doing it over and over. And every single time they got the same rejection, and they they must have done that five times. And then they finally asked me to do it. And I didn’t get any better results. You know, they just flat out said no, which is why I actually suggested Erickson should by Red Hat.

32:54 - That was six months before IBM did it. Ah, I think we would have been better off a very tentative but Red Hat.

33:03 - Maybe Yeah, I just, it seemed logical at the time as a way to, to pick a platform that was extremely common among operators anyway, at some level, whether it was at just the Linux level, or whether it was openshift, or OpenStack, or whatever. So there was almost every operator was using something from Red Hat. And then on top of that Red Hat had the access to the enterprise that Erickson had been longing for. So I just seemed to make sense to me.

33:31 - Yep. And that’s exactly IBM’s play. It is, but I question whether it’s a good long term play.

33:40 - Yeah, it’s hard to say. And at 33, I suggested the price would have been 32 billion. But IBM paid 34 billion. But nonetheless, that’s a that’s a giant bet. But, you know, it’s the kind of bet that a company like Ericsson has to make, if I think if they ever want to break out of the, the typical model of supporting, you know, 18 month to 36 month product projects with the operators. And they, um, you know, they allowed at&t to pull them around like a, like, they had a nose ring in the years attempting, yeah, attempting to build at&t magical cloud on OpenStack.

And then they just threw out a bunch of code and, and scripts to the public and called it, airship, or whatever they did, and said, Look, look, we’re all done, all you have to do is make it work. And then they just dumped it all and went to Microsoft, you know, I mean, it’s like, it’s just they I told him to dump the whole thing. I said, just ignore at&t. But, you know, it’s easy for me to say I’m not, I’m not the CEO trying to figure out how to replace you know, 10% of our revenue if at&t decided to bail on us.

34:53 - Well, you know, I look at it. I look at that and kind of go if you look at what’s happening in the telco world, and You think about, you know, what comes after 5g, and how much of this is already virtualized out and how much this is going to be, you know, the core value in the core business is going to get eroded. Right, I look at it’s a more fundamental question, what Who am I going to be in 20 years, right? Because the business where I make money today is not going to be working to make money more.

Yep. So it’s not a question of if I’m going to make a change. It’s a question if I have to make a change, or or get diabetes had the example of the frog in the boiling water, just turn it up slowly? Well, that’s that’s what a lot of these companies are looking at, you know, so, yeah, but they have to do something. I don’t think they I’m not gonna comment anymore. I’ll leave it at that, you know, that I think is at the core of the Red Hat model.

And my question is, what Red Hat is playing to, is all the enterprise companies that are moved slow, right, they want stability, and they want a very safe path on which they’re going to trip. Right? I don’t think that’s a winning business strategy anymore. Well, I also think back to the back to whoever said anti pattern first. It’s, um, you know, you if you’re going 160 miles an hour around the oval, and the oval is made up of a bunch of companies that play in the open source world, the first corner that’s gonna throw you off the track and crash your car into the crowd is the one that says red hat on it, because they are no more an open source provider than then I am a free gold distributor.

Yep. Well, so back to open source, Red Hat, etc. The Linux found? Well, it’s not a foundation a more than the Linux mega marketing company. Yeah, just just watching it. Literally, it’s like, oh, you want to, you want to put open source on your company, especially if you’re a small company, give us money. And we’ll put your little logo up with the rest in the right Little Box Of cn CF, or open networking, or this or that or the other.

And now, there’s so many different boxes up there that it’s it’s absolute noise, because they’re all tiny little companies, with a couple of giants. So who are you going to pick the logo that you can sort of make out that’s one that’s familiar to you, because everything’s in microprint. So it’s, it’s no longer it’s not useful to have 1000 different companies all doing virtually the same thing and open sourcing it and saying, oh, we’re open.

Everything’s free. Well, no, everything’s just noise or garbage until something comes out that actually plays more. And is expanded. It’s expandable and extensible to larger problems. Yeah, it’s all noise at this point. So signal I’m thinking through what that means. Because otherwise you’re you’re we’re building giant monoliths are giant tech stacks that are really hard to build.

38:41 - Well, the key The thing is, is Linux has gotten to the point where they’re saying, we’re not playing favorites. So right, we’ll just accept money for from everybody and no favorites. And let let the market decide. But it’s gotten to the point where there’s too much noise for anybody to decide until Red Hat buys. Or, or Google.

39:12 - Or one of the things God we’ve created a market where it’s the the, you’re building, you don’t care if you’re giving away your software and IP because your your goal is not to sell the software, it’s to be acquired by one of the the larger companies that are verticalized in the software and have a month zation strategy.

39:29 - Yes, yes. Rob? Yeah, do some future history here. If you know given the conversation just had what would be a a believable scenario around Kubernetes.

39:56 - Like what what what would happen going forward in the future Yeah.

40:02 - Yeah, literally lots of tiny little companies doing Kubernetes and Google pulling the strings. I’ll give you I’ll give you my scenario.

40:12 - Okay. Well, before you do that, Rob, rich, can you provide maybe a little context to that as to? Well, what’s your thinking? Why? Why that? Because that? I mean, I’m having a hard time following along.

40:28 - Okay, wait. During the first, the early part of the conversation, we were discussing the fact that there were so many small companies, putting them together and organizing them as an end user was a problem. That curation seemed to be one of the few ways in which a commercial commercialization of Kubernetes could could proceed. And that Now, given what we’ve just discussed about limits, how Rob discussed the approach that a company like honeycomb has adopted for observability.

41:27 - Let’s, I what I was asking Rob to do is point to the future and say, Alright, maybe name a name, but at least run out a scenario for the, for the successful commercial delivery, and therefore the adoption of Kubernetes by a larger group of consuming organizations set tight enough, Kim said it, does that give you the context? or do any drill down further now? Go? Go ahead. I probably should have just let you. Let Rob Go ahead.

You know, and having joined late, I’m probably missing. Missing more to it.

42:14 - Why don’t let me just catch up. Go ahead, Robin.

42:17 - And let me catch up on it. I it’s, I think your question is reasonable. And it helps it helps me understand what riches what rich was asking. So the when I think about the future of Kubernetes. My I look at software generations, and what it takes to get a platform, right. And, to me, Kubernetes is not the final generation of this platform. And so what what I think is happening at the moment, and is reinforced by what I see is that we’re actually looking at a platform with some core functionality, and then a whole bunch of gaps, and other components that actually should have should be baked into the, into the platform.

And including serverless. I actually, I think that the distinction between serverless platforms and container platforms is going to emerge to be artificial, I was hoping I could bring that up.

43:18 - Great. And I so so what I believe is is happening is I think Kubernetes has a limited useful life sorry, Kubernetes community in which the emergent patterns that we are getting to now with Kubernetes are going to bake into something that is the probably fractures into two or three emergent patterns, that if you segment that group into a thing, it becomes a much more contained thing, right.

43:49 - And then that becomes a new product, what I would consider a fourth generation and I’ll give you the generations, a fourth generation platform, based on where what we learned from Kubernetes. And so I actually think it’s a mistake to think Kubernetes I think Kubernetes will be around for a long time. But I think it’s it is not a final generation. And there’ll be another thing and that is reinforced to me if you look at like Mesa sphere, which like was huge and then was gone.

You know, there’s there is nothing, you know, nothing that says we are going to pivot quickly from Kubernetes. In part because the fundamental things that we’re doing to build Kubernetes are not Kubernetes. They’re containerizing applications. They’re building microservices, their CIC pipelines are observability and for the purpose of usability and and kind of adoption in production. I think you’re absolutely right. You called it a fourth generation platform.

I’m not sure what you meant by that. So when I look when I look at my Starting from pass, talking about Platform as a Service, you know, for your 10 years ago now, yeah. Right, I think we had a generation, like with Cloud Foundry and very Platform as a Service things. I think, I think we had, like the Docker swarms, and sort of the first generation container management pieces that Kubernetes sort of came in, on top of, and displaced.

And so and we have serverless stuff happening in parallel, we also had, like the mesa sphere that I was referring to going on, I feel like there’s, there’s enough uncertainty with that, and there’s enough use case drift that we’re going to, there’s another thing that’s gonna come out of this, it’s probably going to be simpler.

45:44 - Well, I argue that we don’t have any choice if if enterprises are actually going to own a significant portion of infrastructure into the foreseeable future, if digital transformation is an indicator of that possibility, and the upcoming looming importance, greater importance of, of the bottom line and top line value of operating and owning infrastructure, effectively, if any of those things are truths in the future, then the ability for enterprises to more effectively deploy and manage applications is is a no brainer, I don’t think the that we can get there with Kubernetes.

46:28 - Because I think we’ve already been trying to make Kubernetes, perfect for five years. And in order for it to be really perfect, it’s going to be even more complex. And it’s going to have more tools to support it, or somebody is going to try to build those capabilities in to Kubernetes, which will even make it harder and more difficult to manage. And I just think that we’re even in Kubernetes, I think we’re still building for an evolution of an app deployment and infrastructure management rather than a revolution and how applications are deployed and managed.

Well. Good, rich. Well, I think both what Rob said, and what Mark just pointed out, kind of leads me to think about the What does the next turn of the crank look like? And you mentioned serverless, you made a made of, I think you made a really good point about that, um, you know, if what we what gets delivered by the industry by someone in the industry turns out to be a much more declarative, here’s what I want. Make, please make this happen.

Don’t Don’t, don’t make me deal with all of the, the the fight, where we’ve got a dog fight going on here. Don’t make Don’t, don’t put on me Don’t put on my operations organization, the necessity of being, you know, such experts and all of the, the the fine points of Kubernetes. And this, you know, enormous tool chest of, of specialized tools. I think that’s, you know, that’s arguably the place that the market is going to drive it and someone’s going to curate, to your point earlier, Rob, you know, the right selection of tools, they’re gonna make it, you know, work from a point of view of a customer or consumer where they basically say, I declare this to be what I need.

Tell me or actually just go off and do it. manage the orchestration on my behalf. Based on my own my.

49:02 - My role? I think I literally was talking to the as your CTO for APAC yesterday, take Kubernetes led equation that they don’t have enough talent right now to run the existence next. Exactly.

49:16 - Yep. So so there’s a separate issue that we keep picking up on, but let me start a different path.

49:22 - Okay, right. We talk about Red Hat or VMware and that stuff. I think it’s more likely, if you want to talk about 2030 that you’re talking about is your you’re talking about AWS, those become the curated stacks. I’m not sure redhead exists in 10 years.

49:42 - Well, what was the last comment, john? I don’t believe that whatever. I think we were thinking kind of historically here. You look at Red Hat, they’re rolling everything up, IBM rolled everything up. VMware is rolling everything up. But 10 years from now, who’s really doing the aggregation and drive To start from what i think i think i think you’re right, it’s going to be, it’s going to be the CSP.

50:06 - And I’d like to also point out, we should be looking for the Kubernetes. Air, the next gen Kubernetes. Because five years, Rob mentioned, we’ve been doing Kubernetes for five years, that’s when OpenStack Kubernetes came, appeared on the OpenStack. Horizon at the five year point. So Kubernetes has Kubernetes has an adoption issue.

50:36 - Right, they have to get to enough penetration initially, but I think once again, looking 10 years down the road, right kind of stuff, I think we’re also focusing on the library. Right? If we can’t curate, so I don’t care if you curate the stack or not, right, you can’t find the operations people to do this. So the important part about normalizing the staff is about being able to automate it. Right, automate the operations, automate the deployment, it’s about applying AI and ml and all these various components to it so that you can automate the heck out of this stuff.

And that’s where the future is, right? That’s where you’re gonna really find the value, it’s not going to be in curating the stack is going to be an automating the operations of this.

51:17 - I buy into that by wrongly agree with what you’re saying, I actually I love the way you said that. And I it’s personal for what we do with racket, right? We are, we’re literally doing composable automation to make the the need for vertical stack lab less from that perspective, it’s an AI that I hadn’t thought about it the way you described it, which I really like, if you can build an automated stack, then maybe you don’t care about a lot of this stuff.

And actually, I would say that might take away your whole argument about having to be owned by Azure and msps.

51:55 - Right. So you have the architecture, right, so that the stack is is usable, and generate generalizable? So I think we’re getting all these experiments with all these things to learn where the hard parts are, where the what can be thrown out. And so to be able to, well, it’s turning into the general cloud with the switches to customize, as opposed to having 10,000 stars.

52:34 - So let’s come up with a few models that work and then have switches to make things we need to simplify it, actually, through software, of course, which would make it more usable, and less needless training. Go ahead, Mark.

52:51 - Oh, no, Rocky, I apologize. I didn’t mean to try to cut you off. I Rob, if I could be so bold.

52:58 - And since we’re running short on time, I would love especially considering the title of of this weekly events being cloud 2030. I would love a couple of weeks worth of conversation on how we foresee AI and ml impacting how people utilize infrastructure and deploy applications and maybe maybe even a little bit about how AI may displace as many applications as they create new.

53:34 - I would love to we need to find some, some credible experts would be would be helpful for that. Yeah. Not not to put anybody here down.

53:45 - But I would love to see somebody who’s really no, I agree. Bring bring in some opinions on that. So yeah, let’s let’s we can put Sorry, I there’s also a precursor to that mark, which is you have to generate the data on the operational spectrum and the development spectrum to actually be able to apply it.

54:04 - Absolutely. We’re talking about Alright, so I can explain some of the stuff I’ve been working on in that area. But you know, that was the first part of it instrumented? So you can actually apply the tools.

54:14 - Yeah, no, I mean, and john, I don’t, I’d love to hear more about it. And I don’t know for sure if this is the direction you’re going. But I have to make the assumption that if in 2010, that service mesh, we thought it was a good idea to have graphical policy management for multicloud.

54:32 - Two years before the market thought it was even a necessity, that it’s not about whether it was the right or wrong thing to do. But it was the wrong way to pursue the eventual right thing.

54:44 - And, and so I and what I mean by that, is that in the future, or actually let me back up one more time. How often have we implemented solutions in an enterprise as an example That are individually policy driven based on the user updating their policy. And how often does that get maintained and how often is successful, almost never, in my experience. documentum is a great example. In enterprises, it’s almost never successful.

If we’re making an assumption, that the only way we can manage distributed infrastructure and distributed deployment of applications and failover, and multi tenancy on everything from 5g to data sets to IoT devices, etc, etc, we can’t assume that there’s going to be somebody out there, every time a new app gets deployed, actually writing policy into how that app utilizes infrastructure. AI has to be effectively knowing your organization and knowing your demands, and allowing you to deploy applications as if you’re just adding another drop of water in the bucket.

And it gets distributed and deployed and operated at the performance levels required with the compliance levels required. I mean, it just it just, I realized I’m making it sound overly simplistic, but I just we can’t we can’t continue to build a bigger and more complex infrastructure and assume that, that we can manage the intricacies of how applications are deployed by YAML files or policy with.

56:39 - I mean, this is this is the problem. I know, we’re we’re kind of at a time, but I think this is the root of the issues we’re not we should probably spend more time talking about that very issue, which is what are some of the challenges that people are having with this, that enterprises are having with it is an indicator of where, where they might go based on how they might solve those problems. And I think the other pieces, the other conversations about architecture and infrastructure, and you know, whether it’s serverless or whatnot, come up in that conversation will naturally come up in that conversation, or, in some cases, I think some of them will fall by the wayside, because we’ll quickly realize that you know, what, it’s kind of irrelevant in that five to 10 year timeframe.

57:23 - And I’d like to reiterate on mark, the, the fact that we’re looking at the policies, and we need to extend the policy out to the operational policies, where they have, if they’re not integrated, we’re going to keep on with this same problem of come complexity, and inability to actually get a handle on automation. Without the integration of operational policy into the business policy, and the user policy. It’s, we’re going to keep thrashing.

And that’s where Kubernetes is operational. And we keep dealing with the operational and stuff close to the hardware, and we keep ignoring the compliance and the security and the networking, where it has to where you have to have it available. And stuff along those lines.

58:28 - Can I throw out a and I know we’re short on time, but we’re out to an image. And that is, if I were to look over the next five years, maybe not 10, for the place, the alternate location for this to all come together in a way that is customer or user centric, I’d asked, I’d ask you to consider a new generation of MSP that is that is in the business of curation, simplification, packaging for a particular community of use.

And that may be an on at least at this point, a less considered but a very good source for the best solutions and solutions that will hit some will tap the right buttons for the for the consumer for the for the buyer, rich, rich, hitting my favorite. Let’s kick this off for the next topic. While we look for some AI speakers. what you just described to me is is an ecosystem.

59:46 - Exactly right. And I would I would I think we could talk an hour on what it would take to actually rebuild an ecosystem.

59:55 - I think. I think it’s worthwhile. Yeah, I think that’s that’ll be for next week.

60:00 - I agree but put these two things we’ve covered before Rob, I think one, I think we’re in the weeds. I think we’re missing the constructs, we need to think of the abstraction layers that allow us to actually function better. We don’t care whether it’s Kubernetes, or something else.

60:14 - I think as we kind of talked about in the previous call, I also think we’re hung up on complexity. We’re trading organizational complexity for software complexity, and one has a lower cost. It’s inevitable. Right? That’s where the AI and ml comes into.

60:29 - This is This to me is is the Jevons complexity paradox question where I, I think that we need to consider the complexity doesn’t have the same cost that it did. But your your, I desperately want to talk about that more.

60:45 - I like it. Yeah, I think it would be a really good, good topic.

60:49 - Let’s do it. And they’re interconnected. This is great. Thank you. Wow. These conversations are really powerful to me. And the longer we talk, the more we open up really important considerations for the next years, there are major forces shaping the industry and creating opportunities for us. And we’re starting to suss them out, I think in really material ways. We’re looking forward to hearing from you please join us at the 2030 Cloud and be part of the dialogue.

Thanks.