I live in an off-the-grid tiny house. Solar and batteries are my only power source. As such, I want to minmise power consumption. Especially devices running 24/7 (Fridge, Router,…) have quite an impact.
I am now wondering, if I can let the ‘main’ Home Assistant (that used to run on my RasPi3/4 while I was on-the-grid) on Cloud infrastructure (I have access to Virtual Machines) and just have a very slim HA hardware component. That should only forward information and sensor inputs. It should accomodate either the Zigbee USB dongle (or have it onboard, which I doubt).
Can I just run a RaspiZeroW with the said very low-compute setup here 24/7 and let more heavy-lifting be done by a cloud instance? I read that running HA only on one PiZeroW isnt’ fun.
How would I set that up?
So far, I don’t want to run one or several IP cameras, but would even that be possible with such a slim-local/beefy-cloud setup?
It’s something I have been looking into when I want to look into setting up an instance to monitor and such at my parents place through a spare rpi3 I have.
Thanks for your fast reply.
If I get that right, the ‘main’ instance would be the HA server running in the cloud.
The ‘remote’ instance would be the HA that is running on my low-power-consumption device in my house.
But how would a PiZeroW cope with the bare minimum installation? And does anybody have some idea how much a PiZeroW vs. a Pi3/4 would really consume while executing HA?
Correct the instance that is to be attached to the main one will be setup with extra lines as per the instructions to be the remote instance.
As for how a pi zero will cope, someone that has worked with one for their setup or played around with one will be able to comment on it, that said overall power consumption should be negligible. I would be placing it as well as primary modem, routers and networking gear on a UPS anyways to have some backup power when needed to keep things working for the core network.
You will induce a lot of ‘cloud lag’ without a lot of benefits.
If you run automation compute logic locally to avoid cloud round trip lag (twice for every call I might add) then you do the actual automation on the pi zero… Which WILL quickly feel underpowered if it has enough chonk to work at all. It will not feel responsive
If you run the automation in the cloud. Lag. Lots of it. At least 20-30ms each way per call. Your system is now dependent on your internet being up 100% and it will feel not responsive in a different way than above… And less predictable because you’re at the whim of your ISP.
Personally I don’t see where the power you MIGHT SAVE after considering the network gear you have to have and the complexity and risk that you add.