William Morgan, Buoyant | KubeCon + CloudNativeCon Europe 2023
Buoyant CEO and Linkerd creator, William Morgan, provides a preview on the new features being released in Linkerd 2.13 including: dynamic request routing based on the Gateway API, circuit breaking for overloaded services, a FIPS-compatible version or government users along with new security and reliability features integrated for Buoyant Cloud users.
Transcript
This is texturing TV. Hello and welcome to kubecon plus Cloud nativecon Europe and we're talking with William Morgan and CEO of buoyant about service meshes and specifically Linker D. William.
Welcome the show. Thank you very much. Great to be here.
I think by now most people know that people get started with service meshes, usually around managing apis and it's kind of where they began and I think there's some familiarity with that. Do you think though that people fully appreciate the implications of that on how networking is managed how security might be delivered going forward and it seems to be a lot of Downstream impact here that is not been teased out all that. Well, yeah, I think like any You know category of software.
There's this educational process that has to happen, especially. Over time so I think at this point I would say everyone that we talked to I think understands the value of a service. They were they understand it's something they're gonna need if they're adopting kubernetes don't need it.
They understand like exactly why or only all the implications probably not, you know, it's probably a big Spectrum there from the experts to the like, I don't know. I just know I need it. So who's driving adoption of the service mesh, then is that the developer and then the networking and Security Guys discovered later or are you starting to see more of the networking and security guys show up at the table in the first place?
Yeah great question. So it's usually not the developers themselves at least from the you know linkerdy which is a project that I work on. We try and isolate the developers as much as possible from this, you know, they're the ones who are supposed to be writing the application code.
They should be focusing on the business logic and like all this stuff that the service mission is doing around. Security and reliability kind of like, you know decoupled from from them. But so our big drivers of adoption, you know, historically it has been SRE teams a platform owners, you know, the folks in the organization, who are Building the internal platform in service of the developers more recently.
We are seeing adoption. Coming from security teams coming from compliance, you know, especially in larger organizations and even a little bit coming from Finance teams, you know in the modern era of everyone being very concerned about how they're spending their dollars. Right.
Now what differentiates one service measure from another from your perspective because it's seems like we went from nobody knowing what a service mesh is to a an abundance of choices. Well what differentiates them? Well one is really really good.
It's linkardy and all the rest are terrible. So next question. All right.
No, no. Yeah, you know, I think at the macro level they're all trying to accomplish the same. Things, you know, it's it's like okay.
I'm trying to get from point A to point B. I can take a cruise ship. I can take an airplane.
I can take a car right at the macro level they all. Get me from A to B, but the implementation is quite different and the trade-offs are quite different. So the particular angle that linkerty has has taken which is kind of.
A shocking one is we try and focus on Simplicity above all else. So especially operational simplicity. So as soon as you take linkerty into your environment, you know, how do we minimize the cost?
You know, not just the machine cost but really the human cost to operating this piece of software, which under the hood can be quite complex. That's you know, that's that's kind of how linkerdy differentiates differentiate itself in the market. There's other trade-offs that you might make you might you know, you might want a cruise ship you might want, you know, something where you want every possible feature regardless of you know, and you're willing to take on the cost to do that.
That's our that's our differentiation. How big is the Linker D Community now because there's always concerns when it appears that there's only one company driving something. And where are we in the development of the community?
Yeah. No, that's really interesting. The you know, it's hard.
It's always hard to know the exact size of a community because you know, we're not tracking them or whatever. You don't have to register to download link or do you kind of just show up and then use it and like maybe you talk to us and maybe you don't I'd say, you know based on the metrics that we do have around, you know, kind of usage and look at the website and stuff. It's actually grown, you know, it doubled in size and 2022 so it which Was kind of shocking to me because it's fairly mature project at this point.
You know, it's the first I think the only service mesh so far to achieve graduated status and the cncf which is kind of the top level of maturity. So usually when you get really really mature, you know adoption, you hit the kind of S curve and adoption but it's continues to grow so, you know, I think that's a function of kubernetes itself. Of course growing.
Yeah, the fact that there's one company buoyant behind it. You know, I think in the early days was scary for people, but I haven't had That question for for years, you know, I think people are used to that that kind of relationship where you see it with hashicorp or or other companies and in some ways it makes things a lot easier because when someone's comes to us and says, oh we like linkerty, but we need it to do X, you know, there's one team that we have to talk to right? There's not Seven different teams pulling in different directions working in different companies with different sets of priorities everyone who works on linkerdy is Unified.
So in some ways it makes things a lot easier and allows us to be very agile. By now you've seen a lot of people Implement Linker D. Are there any patterns that you've seen or the things that you see people doing that maybe they shouldn't be doing?
Yeah. We definitely have seen a lot of people adopt it and you know, we have a whole business around helping people adopt it. So yeah, there's a lot of patterns that we see I think the the people that we see do this the most successfully and what we try and encourage everyone to do is they follow a couple kind of A patterns I guess it's the right way to say it.
So number one. They will have a platform team, you know, they'll have a team which sometimes they call the devops team. Sometimes it's called the kubernetes team, but they'll have a team who's responsibility is.
Is building that platform and linkerty is a component of that and those are the ones who are responsible for choosing. Here's the service smash. I'm going to use here's kubernetes, you know, here's the ci/cd system.
I'm going to use and they own that kind of end-to-end and they treat themselves as like a, you know, it's a customer service organization. For the developers. So having that organizational kind of alignment is really important.
I think another thing is You have to be prepared when you bring the service mesh in but you know oddly enough sometimes linkardy can uncover problems with your application. Your application was you know behaving this weird way and it's kind of being papered over and now we add this, you know layer of infrastructure that's optimizing stuff. It's you know, routing requests based on latency.
It's doing all those fancy stuff actually can uncover holes in your application. So you have to be prepared. You know, someone says hey this thing no longer works.
You know, it might not be linked reviews fall. You might just uncovered this, you know kind of flaw, you know in the process of adding Lickety. So you have to be prepared for prepared for those conversations.
I'd say stuff like that, you know, if there's some tactics around that and then you know, the big one of course is you, you know, we need a way to show to demonstrate the value of linkerty to the rest of the of the company right? It's the the curse of infrastructure software is You know and being in those roles is when things go right? No one remembers your name and when things go wrong you get into trouble, right?
So like it's all it's all sticking. No carrot. Okay, so if you really want to be successful and this is another thing we help our customers.
Do you want to try and get close to the business, you know the business value and you want to use the service mention some way where you can demonstrate to the business owners. Hey, look here's how we're improving business metrics, right if you can do that and you have a much easier time, you know. We've seen kubernetes itself get a lot easier at least to provision and deploy in the last year.
Do you think that maybe we're at the beginning of some sort of simplification movement of the whole Cloud native stack and Because there's a lot of moving Parts above it that are heavy. Yeah, and so the question becomes, you know, are we gonna see some effort to streamline the whole stack? Yeah.
It's been interesting to watch things evolve, you know, I think with any of these projects and or these kind of areas of Technology, there's this initial movement of like, you know thousand flowers blossoming like all these different things. We're trying and then as a market matures, okay, we figure out the ones that work and you know, that part becomes less interesting to people so it becomes more boring both because it's it's, you know more the choices are limited but also because people are like, okay we know how to do that. Now, let's move on to the next thing.
So I think that's what we've seen with kubernetes early on there are a million distributions and a million ways of doing it people rolling their own people. We're compiling kubernetes and now that stuff is like You know a it's unnecessary B. It's like less interesting because there's new exciting stuff, you know to go look at so you see kind of consolidation, you know and and maturity happening there.
Same thing is happening with the service fish the early adopters of linkerty, you know where people who are very enthusiastic about the technology. They wanted to get into the weeds. They wanted to like or you know, write the code and they wanted to you know, really get their hands dirty with the technology more recently them Market has matured to the point where we have people coming and saying well we want to buy linkerty.
I'm like, okay, you know really how it works. But you know what they're expressing, you know, when they say that what they're expressing is I want linkardy but I don't actually care about how it works. I don't want to operate it.
I just want it to work in my environment. I want to service mesh and I want to work in my environment and that of course is, you know, a giant lead up to exactly everything we do in buoyant which is you know, if you come to us and you say hey we want linkardy and we want to just work. Well, that's exactly the software that we provide.
We automate everything about linkerty. We automate the monitoring. We'll automate the upgrades we these security policy analysis features and tools so everything that you as a company need to adopt a linkerdy into your you know organization without having to have a higher a team of six service mesh experts or whatever it is.
So that kind of interplay, you know is only possible because the market itself has matured not just for kubernetes, but for the service specialists So where do we go next? I mean is the mission accomplished or is there still work to be done here? I just got to finally relax and you know, I never have to think about it any of the stuff again.
No, no, there's always more. There's always more, you know, the directions that I'm particularly interested in is, you know, and I kind of talked about this earlier is in getting closer to Business value, you know, how do we get the service? Mesh, how do we get linkerdy to the point where you know, it's not just solving engineering concerns for the solving business concerns.
So a lot of the you know, I'll give you an example a number one reason for adoption of Lincoln not the only reason but the number one reason is mtls. Okay, great Mutual tell us why do you care about that? It's usually compliance right?
It's security often, but it's usually compliance. Okay. So if you're adopting if you're going through all the trouble of adopting a service mesh into your organization to fulfill some Regulatory Compliance, you know requirement.
Well, you know, what can we do to actually make that easier for you? So we've been building all these tools around. Okay.
Let's give you let's give you a report. Let's give you like an audit that says hey, you did all this stuff you did, you know layers having policy here and mtls here and let's draw you the map and let's give you something you can actually print out and show to an auditor and say hey look you might not understand kubernetes. But here's our report.
Here's our analysis. We're 7% compliant one exception is here and like doing stuff like that is that's really exciting to me because I'm always trying to drive towards business value. I'm trying to get out of the you know, the horrible world of being an infrastructure engineer and just sitting in the dark getting yelled at and try trying to get some of the some of the Limelight.
as we go forward does the service mesh itself sits on top of some proxy software. Does that all start to collapse because we have Ingress controllers and we've got service meshes and we've got the proxy software and there's a lot of moving parts. So can that all be kind of condensed in some way?
Yeah. I mean, I have some opinions there because linkerty has taken a very different tack from the rest of the service mesh world and one of the things yes about differentiation before one of the things that differentiates us at the technical level is we built our own proxy and we built it in Rust and we built it, you know, very specific for this purpose of being a sidecar proxy for a service mesh, which gives us all sorts of advantages that you know other service Mission from the patients don't have but I think to the to the customer, you know, that's just an implementation detail, you know really care whether it's a sidecar or something else. You're really care with this proxy or that so I think those details do kind of get abstracted away.
It's like, okay, what's you know, what's the CPU in your iPhone? Don't really care. What I care about is like I can scroll through tiktok really quickly or you know, or whatever it is.
So yeah, I do think that stuff gets, you know kind of Fades into the background which again brings us right back to like, okay. Well, what's the business value that you're delivering? You know, how do you make linkerdy a critical component of your business and not just solve a technical?
All right, folks you're hurting here. If you don't know what a service mesh is. You should check it out for sure because it's gonna be an important part of the stack just be aware that it goes Way Beyond apis There's No Escape.
Thanks for being on the chair. Absolutely. Thanks for having me.
All right, and we'll be back in a minute.





