What is good hardware based on your experience to run Node-RED?

Currently I am running Node-RED on 3 raspberry PIs (model B version 2).
I am very pleased with the stability. Never had hardware issues that I couldn't fix.

So is there other hardware that can be used to run Node-RED and that you have good experience with ?
I am especially interested in devices with similar low power consumption as the raspberry pi family but with better hardware specs.

1 Like

Beaglebones are very solid also - have quite a bit more IO - and the repo maintainer tracks the Node-RED releases well to keep it up to date.

I started off with a bunch of RPi's B ver 2 as you. Now, for the time being, they are all stored away, awaiting an upcoming trip to our recycle station. Today, I have only RPi3's running, I think 7 in all in our "production system", i.e. home automation & video monitoring & vpn server to give access from outside

For development & testing I have some more RPi3's mounted in a rack to make them easy accessible etc

The RPi3 is a noticeable upgrade from RPi2 in terms of performance. And it's nice with the WiFi and BT integrated. But as one of my old sw developer once stated: "a computer can never be too fast"

It also depends heavily on what you like to do. When handling real time critical video analytics with OpenCV or presenting heavy charts with many data points in the dashboard, well then it's getting tough also for the RPi3, you start to hope for the next, like RPi5 or so

So instead, for heavy charts, I store and save the collected data in a Windows PC (Gigabyte BRIX with a SSD disk) using sqlite3 as database, then using Highchart for making nice charts

I also looked at ODROID XU4 but so far I did not buy one. It looks to me as it is around 5-6 times faster than the RPi3 but at a 3 times the price.

Other thoughts I had was to upgrade a retired laptop with Linux but it did not have a Nvidia GPU with CUDA support, a requirement if you want OpenCV to run much faster

While it is true to say that a computer can never be too fast, it is also true that there is no point having one faster, or more expensive, than is required for the job. I have several Pi Zeros which run simple tasks, for example one collects data from my weather station and sends it back to the main server over wifi. In addition it runs a simple node-red dashboard. It is easily able to do the job and is much cheaper and smaller than a Pi 3.

1 Like

100% correct & true, select the appropriate hw for the job to be done

It's when you start to elaborate, try & test various things, you tend to try to improve performance (without overclocking) things, more as a challenge. I have spent substantial time on video analytics & DNN and in that particular case, high performance is nice. Even though the RPi3 does it "fast enough" for my application, it would be nice to run faster but that seems to require me to make another hw upgrade. So that decision is currently put on hold for a while

Walter can you please share a picture of such a setup? And can you please share a link to where you bought it.

Indeed my Raspberry is not very suited for video monitoring. Have been told that I should go to some Intel-based boards for such tasks... Although every developer should build its nodes on Raspberry, to make sure his code is performing well on a system with less resources :face_with_raised_eyebrow:

Probably overkill, but I have several instances running daemonized using PM2 in an ESX clustered environment. All production instances thusfar are used for systems monitoring and automated alerts for things like servers and medical devices (including a quick glance at uptime on life support equipment and logging network errors to a timescale database.) Rock solid so far: I've had a couple incidents where the host VM auto-migrated (vmotion) and there wasn't even a missed tick in the database.

It's been so successful for improving incident response times (previously we relied entirely on clinicians reporting undesired equipment behavior) that I've been asked to look into other systems we can make more "real-time" via node-red instead of the mishmash of powershell scripts currently in the environment. Next step: mail bot to route alerts to pagers.

1 Like

I'm running Node-RED on 3 PiZeroW around my kitchen all with cheerlights - one acting as master controller - one has PIR sensor to detect movement) (as well as a Pi3 and my main Win10 PC on my desk)

Doing in that way shares the workload out.


This is how it looks like, maybe stack is a more correct name, and I bought it here (maybe this shop is not available in your country)

Related image


Wow, well that is for sure impressive. But for such applications, medical, I assume in a hospital or similar, you have to have very high reliability (and redundancy I think).

For device monitoring, it's more than the application requires, but i'm treating it as practice for a project i'm working on to augment vitals alarms (identify that a patient is quickly trending toward a crash and send notifications to a nearby clinician in order to decrease the time between an incident occurring and a code blue being called, and then automatically send notifications by several means when a code is called in order to get every qualified person moving immediately. That one will need to be a 3 or 4 nines type system.

It speaks to what I love about this project: it works wonderfully to do small things like flip on a light to wake you up in the morning and also big things like identify a potentially life-threatening condition and rapidly notify life-saving personnel. Hell, it even works better than some very expensive commercial products in my testing.


Yes, stability is fantastic for such a cheap device as long as you follow the Golden Pi rules :slight_smile:

  • Always have a good, overspec'd power supply. Preferably, also use some kind of battery backup to allow controlled shutdown on loss of power.
  • Always use a good, reputable SD card type. Preferably much larger than you need (to allow lots of room for wear levelling.

You might want to check out Pete Scargill's blog. He regularly tests out alternative single board computers.

1 Like

Haha! Like your style. Of course, for live use I'd be very cautious about the use of a Node.JS & Pi based system where people's lives depended on it. If for no other reason that Node is not always stable on updates & also isn't security hardened. Still possible, I'd just want multi-redundant systems (like on an aircraft) with slow rolling upgrades (e.g. different versions of Linux, Node.JS, Node-RED, etc running in parallel).

But for prototyping and testing, it is hard to beat the cost and capability of a Pi with Node.JS and Node-RED.

Looks like a large majority uses RaspPi platform(I presume with OS on a SDcard), my questions from a long term operations perspective are:
1.How long does the sd last with continuous 24/7 rw operations? How would it compare to say a laptop or desktop (50%? or less?)
2. Has any one experienced their pi failing due to "worn" sd?

Hey have a look at Peter Scargills blog - he recently spoke at length on this and went through some of the issues with the Rpi alternates.

he also spoke about his experience with different SD cards and backing them



It looks like we are diverting from the original question raised in this topic. I think it is better to raise such follow up questions under a separate topic. You can always embed a link to this topic.

My Pi2 has now been running continuously for at least 2-2 1/2 years with no problems at all. You would normally expect a laptop to last 3-5 years of intermittent, daily, heavy use. Not sure about 24x7 use. Thing is that computers are generally VERY reliable as long as they have a good power source. They tend to fail on arrival or shortly thereafter but then maintain reliability for a long time. HDD's are common failure points followed by hinges, batteries and screens on laptops but that is about them being (un)packed and moved all the time. Oh and the DC power ports on some laptops are prone to failure for the same reason.

PC's in one place, especially when connected to a filtered socket or UPS will last many years of continuous operation. I've no reason to suspect that a Pi wouldn't either. They appear to be amazingly reliable.

Yes. If you buy small, unbranded cards and have a standard installation, they will fail often within 12 months. If you buy good branded cards like the Samsung EVO or EVO Pro ones and also buy oversized ones - e.g. 32GB which gives additional room for wear levelling - they should last years.

Key things for card survival on the Pi: space for wear levelling, reliable and strong power source (2-3 amps from a good USB power block) and never let the power die (always shut down properly). The power cut issue is well known and can corrupt a card. I have both my active Pi's attached to a couple of PC UPS's along with my switch, router, NAS and WiFi AP.

My SD card is being written to, on average, once per second. If you are really worried about it, you can use the SD card just for booting and use a USB SSD drive for everything else.


And if you are not using the USB ports for anything else you can use 4 quality USB Sticks for redundancy. I boot from USB on my Pi3B and Pi3B+. It must also be possible to do a RAID configuration with 4 USB drives too. Hmmm................off to have a look.

Well, yes - though with the caveat that USB sticks are likely to be LESS reliable than SD cards since they may not do wear levelling.

Hmm, not sure about that. And it would have to be software RAID so probably not very performant on the Pi's USB 2.0 interfaces.

Also, the thing about RAID that people forget is that, while it makes the system more recoverable without a rebuild, it actually makes the likelihood of a disk failure HIGHER (since you have more devices to fail). RAID is certainly not the panacea that many people think. RAID5 in particular is not what people think. For example, on my NAS, I use RAID0 not RAID5, though I could. That is simple mirroring which would save a difficult rebuild if my disk failed in certain modes. Using RAID5 would have meant at least 3 disks instead of 2 so extra costs and also makes disk management harder. I've been bitten that way before.

But for a Pi, as long as you have backed up your Pi, NR and other application configurations and your flows, etc. it is really easy to rebuild on a new SD card. So RAID is not as likely to be useful.

I keep a small stock of the 32GB Samsung EVO (moving to EVO Pro) micro-SD cards as they are used in the Pi's, camera's, phones and some laptops.

1 Like

Yes Julian, I think that the projects feature of Node Red really has made a difference in that respect. I just searched and did find some solutions, but of course only software.