Raspberry Pi Forums

Without getting too specific with my final product, what do you think would be the likelihood of a Raspberry Pi running 24/7 for several years? Essentially, I have a product in mind that I'm thinking about selling, and am concerned that the Pi would have a chance of malfunctioning after an extended period of time. Before rushing to conclusions, here are some details of what it would be doing:

- Being in an "idle" state for the majority of the day
- Around 30 minutes per day of light use
- Interaction with a screen running a lightweight Java/C++ GUI
- Receiving and processing analog inputs frequently (~1 Hz) during that 30 minutes/day

I have heard concerns about SD cards having a limited lifespan (failing after a finite amount of rewrites), so I'd likely just put the bootloader on the SD card and having the processing done on an external flashdrive (if this is possible/reasonable). Let me know if this will be an issue.



This may not be easy to answer something rather vague, so I'll draw up a scenario similar to mine:

I want to create a dishwasher that has a GUI that displays water consumption, run-time, and a few other features. I don't know a lot about circuit design, and the Raspberry Pi is a cheap but powerful device that could do this (with an ADC). Would it be reasonable to use a Pi as a part of a final product that I would sell to consumers? Kitchen appliances seem to run for 10-20 years, so I'm concerned that the Pi might have a shorter life expectancy than that, even for light applications. What would be my alternative options? I'm a lone entrepreneur, so at the moment, I don't have the resources to design a circuit board for the application.

Any help would be greatly appreciated!

This sounds like an application for the Raspberry Pi Compute Modules (which is designed to be used in embedded systems). You'd need to build your own carrier for the CM (which uses a stock SODIMM connector). There's no reason why a RPi shouldn't run for years especially if you move to a read only Linux root filesystem.

As dougie says, it sounds like an ideal application for the Compute Module, however as producing your own PCB isn't currently an option, you're best sticking with the Pi.

Having a read-only filesystem will protect the SD card from wearing out, but you will need to store some information somewhere. You can either have another partition on the card and write logs to that, or a USB memory stick that can be removed for diagnostics/analysis, or even a small flash RAM chip on the i2c bus. That has the advantage that no info will be lost if the SD card fails, just put a new default card in. Ok, you may need to get a small PCBdesigned and built for that, but it isn't a complicated one and you should be able to get it done relatively cheaply.

One thing that I'd be concerned about in something like a dishwasher is vibration. The Pi has a lot of connectors. Some are more robust than others. I see a lot of hot-melt glue or epoxy resin in your future. The SD card interface, and the cards themselves, in particular has been a fairly common point of failure. That's where the CM has an advantage by using a permanent eMMC chip instead of the socketed SD card.

Time. Are you going to use an internet connection/gps/rtc or just not worry about the timestamps? For "embedded" stuff like this you might not care what time of day it is, just as long as you have something that can count seconds elapsed.

Just my thoughts, don't let them put you off your exciting project

Personally I'd think a micro-controller would be a better fit for a dishwasher than a Pi. You could develop it on an Arduino and then condense it into a single chip when you have it working. Its a lot less powerful than a Pi, but has ADC built in and when you have it to the single chip stage, its a fraction of the cost of a Pi and will withstand the inside of a dishwasher much more reliably. Its powerful enough for what you want I think.

http://arduino.cc/

http://www.ebay.co.uk/itm/1Pc-ATmega328 ... 2ecfd9964c

Arrrgggggghhhhhh, burn the heretic at the stake....

pluggy wrote:Personally I'd think a micro-controller would be a better fit for a dishwasher than a Pi.
<snip>
Arrrgggggghhhhhh, burn the heretic at the stake....

Agreed.
Unless he's building an Internet-Of-Things Dishwasher, with Cloud Connectivity, that's user-programmable over the internet via a dedicated Smartphone App, and ... <insert your own buzzwords here>


Personally, I'm not quite sure how "I have a product in mind that I'm thinking about selling" and "I want to create a dishwasher" and "I don't have the resources to design a circuit board" fit together?

My 512MB pi has been running for over a month now, but has a few reliability concerns which you should beware of.

If you plan to depend on anything at all hooked up to the usb, that is bad news for long term. Mine loses its /dev/ttyUBS0 at random intervals which resemble Poisson(2) days and when it reconnects changes that to /dev/ttyUSB1, which of course is unknown to my program. Then it refinds /dev/ttyUSB0 about half a day later, so I have about 80% coverage of logging data this month; hardly the sort of reliability that anyone could sell. I'm seriously considering modding my program to explicitly test for the wrong usb in ls /dev but that is such a nasty kludge that I'd rather fix the real problem if possible.

R61zt wrote:My 512MB pi has been running for over a month now, but has a few reliability concerns which you should beware of.

If you plan to depend on anything at all hooked up to the usb, that is bad news for long term. Mine loses its /dev/ttyUBS0 at random intervals which resemble Poisson(2) days and when it reconnects changes that to /dev/ttyUSB1, which of course is unknown to my program. Then it refinds /dev/ttyUSB0 about half a day later, so I have about 80% coverage of logging data this month; hardly the sort of reliability that anyone could sell. I'm seriously considering modding my program to explicitly test for the wrong usb in ls /dev but that is such a nasty kludge that I'd rather fix the real problem if possible.

Have you tried the very latest 2014-06-20 version of Raspbian yet? Apparently it contains significant USB fixes.

R61zt wrote:My 512MB pi has been running for over a month now, but has a few reliability concerns which you should beware of.

If you plan to depend on anything at all hooked up to the usb, that is bad news for long term. Mine loses its /dev/ttyUBS0 at random intervals which resemble Poisson(2) days and when it reconnects changes that to /dev/ttyUSB1, which of course is unknown to my program. Then it refinds /dev/ttyUSB0 about half a day later, so I have about 80% coverage of logging data this month; hardly the sort of reliability that anyone could sell. I'm seriously considering modding my program to explicitly test for the wrong usb in ls /dev but that is such a nasty kludge that I'd rather fix the real problem if possible.

My RPi is running day and night for about 18 months now. No USB problems at all; about 10 months ago I've moved the root FS to an USB HD and it works like a charm. I would look for power issues in your case.

R61zt wrote:Mine loses its /dev/ttyUBS0 at random intervals which resemble Poisson(2) days and when it reconnects changes that to /dev/ttyUSB1, which of course is unknown to my program.

You need to write udev rules to create automatically a symbolic link /dev/mydev->/dev/ttyUSB? .

hampi wrote:You need to write udev rules to create automatically a symbolic link /dev/mydev->/dev/ttyUSB? .

Even if you have a consistent symlink, every process using it will need to be restarted whenever the underlying device changes. They might as well use the real device. Better to fix the USB disconnect issues.

I have had a few Rpis around here that have been in constant use for several months except for the occasional loss in power due to severe storms.. No issues. They just work.

Thank you all for your kind and constructive remarks! I’ll respond to each of you individually.


DougieLawson:

Good advice – I actually hadn’t heard of the Compute Modules before reading your message. Certainly seems to be a good application since it’s designed for embedded systems. My one concern is that, cost-wise, it may not have much of an advantage over the Pi itself if you consider the cost of manufacturing the carrier card ($30 Compute Module + Carrier Card vs. $35 Raspberry Pi + $10 SD Card). If there would be significant advantages as far as computing power and longevity when using the Compute Module, then maybe that’d convince me more to learn circuit design (I’m going to at some point anyways, but this would just cause me to speed up this process).

So, to pose a specific question, what are the pros and cons of using the Raspberry Pi Compute Module rather than the Raspberry Pi for an application similar to the dishwasher scenario?


rpdom:

Same question, what exactly are the benefits of using the Compute Module over the Pi? I see that you mentioned the eMMC vs. the SD card, which is certainly an advantage, although vibration may not be a huge issue in my actual application. Space isn’t an issue at all, so the extraneous peripherals of the Pi aren’t a problem.

Hmm, I’ll definitely take that into consideration. I do have experience with circuit analysis, just not circuit design, so that may be a possibility.

I really appreciate your last comment. It’s really easy (as an aspiring entrepreneur) to let my emotions either overexcite me or discourage me, so it’s nice to stay grounded.


pluggy:

I should have been more clear in my original post – a large part of the end product is the GUI, something that I don’t think an Arduino would be able to create (can an Arduino create a standalone interface at all? I’m actually not sure). My original plan was actually to combine the Raspberry Pi and Arduino, but I realized that the only point of the Arduino in that plan was to use the analog inputs, which an IC could take care of.


AndrewS:

Heh, actually, more than one of those things may be involved in the final product… I feel a bit embarrassed :X. Again, I should have been clearer my original post, but I’m actually not designing a dishwasher; that was just an example (I’m a bit protective/paranoid of revealing too much about a project before development is far). The Smartphone App is definitely a possibility. I might make the main GUI capable of streaming video (inputted from the App). That’s all up in the air at this point.

Yes, I can definitely see how that could seem difficult to accomplish with limited knowledge of circuit design, but it’s easier than you might expect. This (https://www.kickstarter.com/projects/pi ... ne-can-use) is a good example – the 3D printer’s circuitry is essentially just a Sanguinololu paired with a Raspberry Pi. Although, even they admit this solution is “ad-hoc”, and I believe are currently looking for electrical engineers to design boards for this specific purpose. If you really disagree and think that learning circuit design is crucial at this stage, please let me know.

The compute module appears expensive and it would be if you buy one with the $200 developer kit then buy ones for production one at a time. But if you're buying hundreds or thousands or tens of thousands the unit cost will be a lot lower than $30.