---
title: "Beyond MicroserVices: Running VMS, WASM, and AI WOrkloads on Kubernetes"
weight: 1
tags:
 - wasm
 - vm
---

<!-- {{% button href="https://youtu.be/rkteV6Mzjfs" style="warning" icon="video" %}}Watch talk on YouTube{{% /button %}} -->
<!-- {{% button href="https://docs.google.com/presentation/d/1nEK0CVC_yQgIDqwsdh-PRihB6dc9RyT-" style="tip" icon="person-chalkboard" %}}Slides{{% /button %}} -->

This is more of an "overview" talk and less actual new knowledge or specialized stuff.

## Baseline

We all know

- Deployments
- Statefulsets
- Functions
- and so on

## Strange new World: VMs on Kubernetes

- Why VM? Legacy! (and VDI and some testing envs)
- The cool thing: VMs are basicly Pods with virtualization powered by kvm/qemu/libvirt
- Demo: Kubernetes on GCP with KubeVirt installed and deployment of a vm with guest tool access
- TL;DR: Kubevirt makes the vm management ux pretty good

TODO: Steal vm vs container vs kubevirt layers illustration

## Kind of a different universe: WASM

- WASM: Low level typed intermediate machine code
- WASI: System Interface for externel functions (fs, network, ...)
- Pro: Secure, Portable and performant
- Con: Bleeding-edge, complex, and not feature-complete

### Now on kubernetes (with spinkube)

- Still executed on a node with a pod, but this pod does not contain a container, but a Spin which contains the service as a wasm container (via the containerd-wasm shim)
- Up to 10x faster spin up than a traditional container

## And how about ai?

- Goal: Host yourself or at least in the EU
- Simple quickstart: Ollama
- Challenge: Cost planning