text stringlengths 0 2.67k |
|---|
**Justin Garrison:** Is wasm still typically function-based? I mean, it's short-lived runtime. Because I always hear the fast startup time as like this huge deal for WASM. It's like "Oh, it's great. We can start in microseconds." I'm like "Well, it doesn't matter if I only start it once." If I run a traditional web ser... |
**Taylor Thomas:** Yeah, that's a great thing to point out, Justin, and that's something that we always talk about. So if you want to run Lambda functions as a service, we're great at that. But we also are pretty much architecture-agnostic. We personally have ran microservices on top of it, we've done long-running thin... |
**Justin Garrison:** But it's still event-driven. How does a long-running process with event-driven -- or is this just like forwarding the requests? It's always there accepting requests, but then as they come, it just triggers the event? |
**Taylor Thomas:** Yes. |
**Bailey Hayes:** Inside wasmCloud, yes. There's a very popular protobuf API for long-running requests that we've sort of taken inspiration from and modeled within the wasmCloud ecosystem for that type of thing. However, I will say that one of the key benefits and one of the reasons why people are so geeked about serve... |
**Justin Garrison:** One of the things I was trying to get to is I have been defining the WASM ecosystem, or WASM specifically -- WASI specifically for system interfaces, to myself. This isn't talking about actually WebAssembly in browsers. For the system infrastructure, microservices style. So I've been defining it fo... |
\[32:22\] So now, my definition for myself has always been like "Oh, well, now we're doing that at one level up" and saying the application's a little more portable. I don't care what database I'm calling, or what event bus, or those sorts of things. And I can say, "Oh, I can change my architecture." But you could do t... |
**Bailey Hayes:** It's so close. I would say that wasmCloud is the Kubernetes Lambda that you're looking for, because what all of these other projects that exist in the ecosystem that help WebAssembly run well on Kubernetes - which are awesome and cool - are missing one key aspect here, which is the WASM-native orchest... |
So with our operator that we run, that is all done over interfaces. And we natively scale this out. And actually, I feel a little silly saying it, because Taylor here actually created what we call WADM, the WebAssembly wasmCloud Deployment Manager. |
**Justin Garrison:** WADM? This is great. |
**Bailey Hayes:** WADM! Yeah, we say it like a punch. Yeah. Go ahead, Taylor... |
**Taylor Thomas:** You say it like a punch. You can see it in the logo. If you look at the project, it looks like a punch from a \[unintelligible 00:33:53.06\] |
**Autumn Nash:** I love whoever did the naming for WebAssembly. Whoever did the naming - chef's kisses. Like... |
**Taylor Thomas:** We have a little too much fun with the w's, we'll just say that. |
**Justin Garrison:** Can you describe the backend of wasmCloud? When we had Fermion on, they let us know it was like EC2, it's AutoScaling Group with Nomad. And it's like "That's great." They shove everything in that, and they can run thousands of functions or executions per node... How does that look for wasmCloud? Wh... |
**Taylor Thomas:** Actually, this is like our bread and butter about what we love, and it kind of backs up what Bailey was saying about why we think wasmCloud's really that next Kubernetes. Because an important thing to note here before I dive into that - because this provides important context - is that Kubernetes is ... |
And so we tried to learn a lot of things from what Kubernetes did well, and what it didn't do as well we created this. So how wasmCloud works under the hood is we're built on top of another big CNCF project called NATS. NATS is essentially our networking layer. And the best way I've ever heard this explained was actual... |
**Justin Garrison:** It sounds like an enterprise service bus. |
**Taylor Thomas:** Yeah, kind of. I always hate saying enterprise service bus, because it makes everybody cringe, like "Oh, no!" |
**Justin Garrison:** How much SOAP is there? |
**Taylor Thomas:** \[36:04\] Yeah, how much SOAP is there? Is there an XML envelope? Anyway... So the whole thing with NATS is that it has those smarts, because everything -- you don't have to worry about service discovery. You're using a message bus to address things to where they need to go. And so we build on top of... |
So that's how wasmCloud itself works. Now, in and of itself it doesn't have any smart routing, but what we didn't want to do is have the same problem that Kubernetes has, where the Kubernetes API -- I mean, I know there's separate components underneath. We can always dive into that. But essentially, the Kubernetes API ... |
Then on top of that, for integrating in with things Kubernetes - so that in and of itself, those things you can spin up and run on plain VMs, you can run it on Raspberry Pi's, you can run it on whatever you really want to. You're not required -- our big things we're compatible with, but not dependent on any previously ... |
So you can use it from the wasmCloud tooling, or the Kubernetes tooling, and then it integrates in with things using like endpoints slices to be able to automatically create services to bind these together. So we're actually -- we've demoed this several times, and we can always demo it, we'll always be demoing it again... |
So there's no longer this thing where you have your DevOps workflow where you have to have 20 different copies of the thing being deployed to US East and US West, and then this region, and then this backup policy. You can just manage it as a single cluster depending on what you're doing. And those clusters can then aut... |
So that's kind of the overview from the low-level to the higher level of how this all integrates in with what we currently have and anything else you want to run. |
**Bailey Hayes:** I'd describe it as a single WASM-native control plane. |
**Autumn Nash:** \[40:01\] That's awesome, because I feel like everybody wants to be cloud-agnostic right now, so they can move back and forth... And I feel like out of all the conferences I've been to, it's either the same startup that nobody wants over and over again about AI, or it's like this, where it's something ... |
**Justin Garrison:** Autumn, I thought we were gonna make a whole episode about AI being mentioned. I was so close... \[laughter\] |
**Autumn Nash:** But it's true, though. It's true, though. It's th same -- like, I've never been so horrified on conference boards, and I'm like "You did what?" |
**Taylor Thomas:** Being in the startup world, it's the "sprinkle AI on it" strategy. And so that's what we call it. |
**Justin Garrison:** That's not a startup thing. That's not just a startup thing. Go look at anything any FAANG is doing. |
**Taylor Thomas:** I know, everybody's doing it, but it's also -- that's like the strategy. People are like "Oh, we weren't sure, so we just sprinkled some AI on it and got our next round of funding", or whatever, you know. |
**Justin Garrison:** Yeah. The only difference is Google's and Microsoft's and Amazon's sprinkles are a lot bigger, right? |
**Taylor Thomas:** Yes. |
**Autumn Nash:** But it's wild... It was just like "Let's think of the worst idea we can come up with. And then let's put a bunch of money in it..." I'm just like "This is wild." So anyways... |
**Justin Garrison:** But back to single architecture-free application. Is that a responsible architecture? Is that something that you would recommend or should be doing? |
**Bailey Hayes:** I got this... \[laughs\] |
**Justin Garrison:** Having an app that spreads regions, spreads cloud providers, spreads -- and trying to make it as easy as possible... Like, I look at some of the really large outages in the past for any company usually comes from this "I have one interface for everything, and I changed the config, and automatically... |
**Bailey Hayes:** I would say that it's not particularly a WASM problem, but specifically in the wasmCloud space. We designed it with resiliency in mind, and global scale in mind. And the way this essentially would work for you as doing a configuration rollout is setting a label for your specific zone that you want to ... |
**Justin Garrison:** Yeah, at some point you do want to take responsibility for how this works. The first time I was using WASM was I was building on Cloudflare Workers, and I think it was great -- and Taylor, to what you're saying before, I didn't care or know that I was writing WASM at some point. It was just like "O... |
**Bailey Hayes:** Yeah. So to answer Autumn's original question, we were pushed in this direction because of our users and our customers. And what they needed is a way to react to this pendulum that keeps swinging back and forth. It swung hard towards the public cloud, it is now going back very hard in the other direct... |
And now I will also say that most of the folks that are our customers, they're banks, they're telcos, they're people that you wouldn't expect using new, innovative technology, that are really all-in because this is the only thing that really solves that problem well, with the level of resiliency, scalability and really... |
**Justin Garrison:** \[44:14\] New, innovative technologies also feels very old sometimes, right? This feels like a Kubernetes cgi bin. I mean, that's what function as a service and event-driven architecture sometimes feels like, but making it abstracted and putting some smarts at the lower level. Like, put that at the... |
**Bailey Hayes:** What's old is new again. That is definitely the case. I mean, look, hey, I didn't say it this time, but I could have said "WebAssembly, it's your write once, run anywhere technology." You might've heard that before. |
**Justin Garrison:** Yeah, that's why we were talking about it with Java in the last interview we had about WebAssembly... Because it was -- it feels very similar, but with newer smarts at different layers. We've taken some of what we learned from enterprise service buses, we're taking some of what we learned from Kube... |
**Bailey Hayes:** And we've broken it down to first principles, and went back to the root of the problem, which was effectively "How do you describe what an application is? And what does an application need?" And then built a platform around that. |
**Justin Garrison:** And wasmCloud is something anyone can run themselves? This is open -- |
**Bailey Hayes:** Open source, CNCF project... Yeah. I think what I would tell people to do is to use the wasmCloud operator. We've got a Helm chart, it helps you deploy NATs, and then wasmCloud on top of NATS... It runs really well inside of Kubernetes... It doesn't have a lot of the pitfalls that some of the other WA... |
**Justin Garrison:** Do you know how that compares to other event-driven runtimes that are more container-native, like Knative? Knative gives me that sort of Lambda-style execution inside my environment, but it's all container-based. I think they have some abstraction of how you get executed... I haven't run it for a w... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.