Best recommended platform to run HA

Configuration
1

Hello HA community.
I have been experimenting on and off with HA for the past few years. I have run it on a PI 4, a VM, Docker, and currently a supervised version on a Debian machine.

Not sure what platform worked best, but the Debian machine has been the least amount of fussing.
Trying to optimize all the different machines I have running at my home and want to consolidate by adding HA to one of them from a stand alone machine. Currently at my home I run the following machines 24/7 and would like opinions as to what might be the best option to run HA simultaneously.

X86 7th generation Intel machine running Windows 10 and Blue Iris (my sole purpose security NVR). Could run HA virtualized on this machine - option 1

Synology NAS (actually epenology ) running on a 4th generation Intel cpu. Could run HA with Docker on this machine - option 2

Debian Bullseye running on an older HP mini computer (current HA setup). Sole purpose is for HA. Option 3

OpNsense router running on a mini PC (cpu J3160 - 4 cores). Not sure if this is an option to run HA one. Option 4

Raspberry PI 4 - Currently reserved for other projects, but could use for HA. Option 5

Older Pi3- Option 5.

So many choices. The goal is to save power and simplify the IT headaches. Now that the family has a taste of everything HA does, when it goes down there is a bigger price to pay. thanks for the input.

2

NUC or equivalent is low (volt/ma) power yet good cou/ram/expandable

2 Likes
3

I have it running on an old Gigabyte Brix J1900 and it idles around 8-12% with occasional spikes to the 20s… I just grabbed it out of a drawer since I was having issues with USB passthrough not living through reboots on my VM machine. Just about any of those will be enough.

If it were me I’d probably keep it on the HP mini PC assuming the CPU is within the last ~10 years but run HAOS since it’s sole purpose is HA. The J3160 would be a great option, but if it’s being used as a router I’d just keep it as just a router since it’s not a very powerful CPU. I personally don’t like using a Pi because SD card storage isn’t reliable enough for my tastes but a lot of people have done it successfully.

4

Plus 1 to the NUC - with native HAOS install - not supervised / docker / VM etc.

5

I’m hosting mine on Synology DSM 7’s VMM.

I’m facing issues when using the SkyConnect. It works fine with ZigBee2MQTT, but doesn’t work at all for multiprotocol.

So if you happen to be using SkyConnect, you might wanna stay away from Synology DSM related setup until further information can be found.

6

WITH docker … allows to also install non-HA containers…one must like/appreciate docker though

7

Tend to agree on this point and prefer option 3.
RPI not ideal due to SD cards and the potential failures that come with them, not sure about Docker setups as I’m not comfortable in knowing if all will work as easily with them.

Currently running on an older mini-PC (based around an intel i3-4010u) running DietPi, which is Debian with some extra scripts to make it run nicely lean. Running Supervised and has been flawless so far, and I only started using HA a couple of months ago.
Option 4 sounds interesting in the way that it reduces the amount of machines you have running, just not sure who easily Opnsense and HA can coexist…

8

IMO, Option 1 using VMWare if your system has enough resources.

It’s the easiest path especially if option 1 is already up and running. Definitely the easiest path for full recovery.

9

Reconfigure machine 1 to run proxmox. Then install windows/blue iris in a VM, haos in another. Or use bullseye plus supervised in the second vm (since you are familiar with it).

Downside, backing up blue iris and reinstalling win + bi. Upside, the future …

10

Option 5 (RPI 4B variant) - I’m running on a Raspberry Pi 4B 4GB in an Argon One M.2 case (only SSD, no SD needed, includes cooling) on standard RPI HA OS image: average processor usage 6% and memory usage 27% (app. 1 GB). I have about 80 sensors/switches on it, app. 50 automations in parallel and I am not running any heavy database-configuration, it fits the job very wel. It consumes about 11W (inclusive WiFi, RFXTRX, Z-Wave and ZigBee communication). Which allows me to run it with its own little 15 Wh USB-C UPS. The problem, nowadays, is of course being able to get an RPI 4B.