I’m having a NUC i5 for some non-HA projects. Now I’m thinking to build up a Docker platform on this NUC. And to be honest I’m not that experienced with docker, have used it some times but I’m definitely not an expert yet.
Does this make sense to move my HA there or would I have to expect a loss performance? Now HA is running von a Raspi4, 4GB and with a USB3 Samsung T7 SSD and I’m quite happy with the performance.
The idea is to get rid of some of my raspi and move these to Docker images (like Pi hole, some Python envs, maybe HA and 1, 2 more).
I’ve run HA on Oracle VirtualBox on a Windows 10 i5-based machine for some time and the performance was just fine. (Much better than my Pi3 although I hadn’t added an SSD at that point.) I’ve since upgraded to an i7 box because I also want to run Blue Iris and that’s resource intensive.
The VirtualBox setup allows you to “pass-through” USB connections to the virtual machine and it works with my ZigBee/ZWave (Nortel) stick just fine.
This is a pretty good how-to on setting up the virtual machine: Migrate your Hass.io from a Raspberry Pi to a Windows Virtual Machine – The (thesmarthomehookup.com)
and here’s a good bit on making sure USB pass-through is enabled: VirtualBox USB passthrough guide | Connect USB to VirtualBox (eltima.com)
I often read in this forum that Raspis don’t have a that good reputation, but mostly because of SD. But as wrote, I’m running it on a SSD and its from my opinion very fast and responsive (it’s a huge difference compared with running it on SD)
Nevertheless I want to replace my 5 raspis (1xHA, 1xPihole, 3xSpark/Hadoop Test-Cluster inlcuding Apache Zeppeling and some Python stuff) by the NUC.
So the question is, is Home Assistant inside a docker image on a Ubuntu NUC (i5, 16GB Ram, Evo 970 SSD) as fast as native on a raspi4 on SSD, even faster or might there be a worse performance?
The Sysbench CPU test puts the Pi4 at about 394 and an i3 about 4210. (The i5 would be even faster.) Add in memory speed and somewhat better USB handling and you can really tell the difference between running on a Pi and running on a NUC.
Even though you’re talking about using one NUC to do the work of 5 Pis, the numbers look like you’ll still come out ahead with a NUC.
My experience running HA on a virtual machine would tend to bear that out and you’ll be running a Linux distro instead of the more cumbersome Windows that I’m running.
Haven’t knew the hughe difference in memory speed.
2 of my 5 raspis I won’t need anymore in summer. I’m study part-time and they are for some big-data exercises.
In the end on the NUC I see 4 containers in the first place (Pi hole, Home assistant, Python, Apache zeppelin).
And if I summarise the thread, running Home assistant in a Docker env shouldn’t be a problem regarding performance?
I’m now running it since 3 days on the Intel NUC and I’d say it’s even faster. Not as much as the difference between Raspi on SD Card and Raspi on SSD, but maybe 15-25% faster.
The Fibaro commands are now instant (before there was a very small delay, maybe 1/4 to 1/2 of a sec ← for sure no problem at all, but it’s a difference now) and Grafana loads faster and overall it feels even a bit smoother. So in the end it’s no gamechanger, and on Raspberry4 with SSD it worked also very well, but my goal was to get rid of the many Raspis without a performance loss and this I could achieve with this Intel NUC.
