📺 https://youtu.be/GlcZu6RGfoI
- Make the case for on-prem
- I want to live in a world where on-prem is the priority
- Cloud computing is fucking expensive – especially ML
- Need on-prem to really learn infrastructure engineering
- Here’s the hardware I have laying around
- 14 Mac Minis
- 4 MSI Tridents (GPU Nodes)
- 30-ish assorted Raspberry Pis
- 12 low-power laptops
- Streaming rig
- 1 Digital Ocean
- 5 old working phone
- Let’s talk operating systems
- No cloud-ish operating systems
- Find a place for OpenBSD for network
- Enterprise-y
- Fedora (RedHat)
- Debian
- Ubuntu
- OpenSuse
- Alpine
- Raspbian
- K3OS (RancherOS)
- Sidero / Talos
- Arch
- nixOS
- Slackware
- ContainerOS
- Compile your own (but only later)
- LFS (grow our own)
- Gentoo
- General architecture
- Ceph network storage
- Emulate two availability zones (data centers)
- Tangent into chaos engineering (and John Deere generator)
- I feel a split in my learning strategy
- Time for raw infrastructure (networking, LFS, netboot, etc.)
- Time for hardware and node setup
- Time for CKAD development (operators, redundant pods, ingress, etc.)
Related: