It's worth the hassle. If you are obsessed with power consumption, consider the fact that Pis are very power inefficient. They draw less but also produce way less than other CPUs. Performance per Watt is shit. They were conceived as thinkerware. They are not reliable the way everyone here wants them to be.

Just to put your puny power concerns into perspective. My homelab draws around 130w on average. One big box with lots of HDDs, 2 Lenovo Tinies, networking.

My laptop draws more power in 24h than my homelab. If you have a desktop, the discrepancy is even bigger. Nearly anything in your household blows your setup out of the water. I live in a country with relatively high power costs and I can assure you, your concerns are nothing more than a thought experiment.

I would get the N100, get 2x2TB SSD as well as some smaller SSDs for VM and LXC storage. Create a hypervisor with the appliances you need - NAS, Docker, heck if the board has a free PCIe slot you can put a 4x1GbE NIC and spin up a networking VM too. This way you can streamline the maintenance - updates, backups, etc. And the best part is that this single box will be waaay above anything you have now.

I built my first thing from scratch. I say "thing" because it was neither a NAS, nor a server, or a hypervisor. It had storage, standalone services, containers and VMs.

If you are unclear on what you need, DIY is probably the best approach.

I've invested a lot in my DIY machine. It's a node in a proxmox cluster. The other two nodes are mini PCs. If I was starting from scratch now, I would probably go with a prebuilt NAS and a bunch of mini office PCs for running VMs and other things.

DIY gives a lot of flexibility but you are managing every aspect of it. Borking the storage, borks your whole setup. It's a lot of fun but you need to know what you are getting yourself into.

Getting only a NAS will be insufficient once you start experimenting. Which means you will need to get another machine for hosting services/VMs. It can get expensive quickly.

If you go the DIY way, start witha hypervisor and virtualize everything else. That way you will have a more stable setup.

Everyone jumps to parrot "Frigate" but Frigate is shit for live viewing and interactions. Live viewing is rudimentary, there is no PTZ control (as far as I know) and no voice communication.

It's strictly a smart NVR and as such it performs exceptionally well. It recognizes, records and marks events so that you can come later and find what you are looking for, very easily.

If you want interaction - live viewing, PTZ, Voice, you sholud look elsewhere. Shinobi supports PTZ, so does Blue Iris. Both can record audio but I'm not sure of broadcasting audio back to the cameras. Zoneminder is an option too, it supposedly supports them aswell but I have never tried it.

When I was setting up my CCTV, it was Shinobi vs Frigate. I had configured Shinobi and planned on deploying it but decided to try Frigate on a whim and never looked back.

In short, Frigate covers most needs perfectly but people shouldn't just parrot the same thing without answering within the context of the question.

If you own the hardware, sure, use it. It will suffice.

If you plan on buying this ancient hardware, do not do it!

There is some really anecdotal evidence from the other comments.

Check out Haswell capabilities https://en.m.wikipedia.org/wiki/Intel_Quick_Sync_Video

It decodes h.264 and that's about it. Yes, plex does not transcode for free but hardware is hardware. Spending money on obsolete, lacking hardware is not wise.

If you are buying, go for 6th gen, 8th gen or 11th gen. These are the sweetspots that are either cheap, widely available or very capable.

Kudos to you sir. I'm first to jump against RPi in homelab posts but this is on a whole other level. I think everyone would love a detailed explanation on it.

The Compute blade comes to mind and I'm drawing parallels between them. AFAIR, the compute blade does power and management over the front ethernet port. Which requires the PoE stuff to be there too. Does your backplane simplify the boards (and make it the project cheaper)?

I scrolled through the discussion and remembered why I love Docker. The only people who see Docker as a hurdle are those who haven't used it. Especially with these microservice apps.

I've have sites on Jekyll and Hugo. I prefer Hugo.

One of my Hugo sites is very heavy on images and although it's not a photography portfolio, it's a portfolio of progress photos. Almost 1500 images. Hugo builds the site and generates the thumbs in under 3min which is impressive, compared to other SSGs.

I use Go templating in other places so Hugo templates come a lot easier than Jekyll's. It's a bonus in my book.

TechnoTim on wildcard certs

Instead of using Duck DNS, buy a cheap domain, register it with Cloudflare, set up "cloudflared" which dynamically updates your DNS records - a DIY DDNS. You could even call it DDDNS. That way you are in control of the whole chain.

The others gave you a solution but note the following:

Jellyfin in a VM means that once you passthrough the GPU to the VM, you will lose graphical capabilities on the host. That could be a dealbreaker. Segmenting the GPU is inefficient.

I host my Jellyfin in a LXC because I can passthrough the GPU without reserving it to the LXC.

Here is my config. I use the i5-7500 and can say that it's pretty good. It's in the main node of my proxmox cluster that hosts my NAS VM and main Docker instance with lots of containers. Average power consumption is about 65-70W.

I started out with just this node as a standalone server so it has everything. I got a FD Define R5 case because it can hold lots of drives. Define R7 is recommended too. R6 isn't because its generation is the only one using its own HDD trays (IIRC). Overall, they are the best NAS tower cases. Sound-proofed and every HDD is mounted with gromets. I use a Seasonic Focus PSU because they come with a lot of HDD power connections out of the box and I don't need to daisy-chain.

This is the 4th iteration of this machine and the item I've carried over since v1 is the HBA. (RAID card in IT mode). It removes the requirement of the MB to have a high amount of SATA connections. The whole card is passed through in my NAS VM.

What I'm planning to do is to add a NIC. Not sure 4x1GbE or 2x10GbE. Or maybe both. With them, the power usage might go in the mid 80s but that's fine for a machine that can be a hypervisor, a docker server, a NAS and a network appliance all in one. The downside is that if you reboot it, you lose the network for a while.

Additionally, I got two Lenovo Tiny 920q that run my experimental stuff. They are purely for compute and for HA failover in case the main node fails.

If you can find DDR3 for free and your power is cheap/renewable, stick with it.

In any other scenario - upgrade to newer hardware. Don't buy e-waste.

It's worth it. At some point you might enconter a service that requires SSL to work even on LAN. I treat them like pipes. The fewer pipes i need to pipe traffic through, the easier it is.

I use split DNS to access services locally, over the internet and via VPN. Everything is behind a Traefik proxy that uses wildcard certs. It enforces SSL for everything and I have just one pipe to think about.