sup nerds rate my cluster design

the goal is a truly no single point of failure homelab deployment

I'm optimizing for low power and quiet operation. These will be running talos linux for a bare metal kubernetes deployment. The control planes simply need fast storage and CPU ensuring etcd doesn't get out of sync. The GPU workers are getting NVIDIA T600s for on the fly transcoding, should it be necessary. The ceph cluster will hold 2 OSDs(disks) each. I'm shooting for 24TB usable in the beginning then expanding it later on. I'm putting SSDs in these but I want to go NVME at some point

I might go 100g networking if I can find a good deal on optics. I have a few Juniper qfx 5200s sitting around and I've gotten them pretty quiet fucking with the fans. I know they're not low power.

this will be running Plex, NAS, blah blah blah, typical homelab shit. Its just me and my wife using it but it will have horizontal pod autoscaling to meet any burstable demand and geographic redundancy since once this is up I'm spinning up 2-3 replication clusters if I can get some rack space at a good price

Looking for recommendations on low power, high performance platforms. The skus in the diagram are suggestions