Author Topic: High Reliability SBCs  (Read 2867 times)

0 Members and 1 Guest are viewing this topic.

Offline OnboardG1Topic starter

  • Newbie
  • Posts: 3
  • Country: gb
High Reliability SBCs
« on: July 27, 2020, 09:32:56 am »
Hi Folks,

I'm on the hunt for a very reliable SBC for a future project. We're developing a general purpose DAQ system for some monitoring hardware that's going to be in the middle of nowhere. As a result we can't really have a device that occasionally frags its kernel storage like the Pi does. I've done some of my own research and found both VersaLogic's Owl and Advantech's designs. I think these might be a little overpowered for what we need, although the VersaLogic ticks the reliability box. Realistically we need the file system, SPI capability and the ability to run a PTP/NTP daemon. We've also considered an RTOS but since we're not power constrained or building in volume a little overkill is okay to make development and debug a little easier. I'm not very experienced with SBCs, so I suspect I've missed some obvious candidates. If anyone has good (or bad!) experiences with vendors who provide SBCs for long term operation I'm keen to hear them.
 

Offline Psi

  • Super Contributor
  • ***
  • Posts: 12783
  • Country: nz
Re: High Reliability SBCs
« Reply #1 on: July 27, 2020, 09:55:32 am »
One thing i would look for in a high reliability SBC is passive cooling with a heatsink under 55C even at continuous 100% CPU usage.
Too many SBCs run the CPU at a temp that is not compatibly with long term reliability.
That said, sometimes you can change the power profile and keep them running at a slow clock for less heat.
« Last Edit: July 27, 2020, 09:57:40 am by Psi »
Greek letter 'Psi' (not Pounds per Square Inch)
 

Offline MasterBuilder

  • Regular Contributor
  • *
  • Posts: 89
 

Offline tggzzz

  • Super Contributor
  • ***
  • Posts: 23122
  • Country: gb
  • Numbers, not adjectives
    • Having fun doing more, with less
Re: High Reliability SBCs
« Reply #3 on: July 27, 2020, 10:14:26 am »
Reliability is only partly a component issue. Reliability is a system design issue.

You can start by defining what you mean by "reliable" for your uses, if you haven't already done that.

Often it is helpful for you to define what isn't needed. That allows you to concentrate on essentials, and often suggests system-level techniques to increase your system's value.
There are lies, damned lies, statistics - and ADC/DAC specs.
Glider pilot's aphorism: "there is no substitute for span". Retort: "There is a substitute: skill+imagination. But you can buy span".
Having fun doing more, with less
 

Offline OnboardG1Topic starter

  • Newbie
  • Posts: 3
  • Country: gb
Re: High Reliability SBCs
« Reply #4 on: July 27, 2020, 11:13:00 am »
Reliability is only partly a component issue. Reliability is a system design issue.

You can start by defining what you mean by "reliable" for your uses, if you haven't already done that.

Often it is helpful for you to define what isn't needed. That allows you to concentrate on essentials, and often suggests system-level techniques to increase your system's value.

Yeah that's a fair point. We need it to run constantly for multiple-year timescales without requiring hardware-level intervention since that means a long trip out to the arse end of bugger all to make a repair or replacement.

Nice shout on the "don't need" list. I'll have a think about it.

Quote
One thing i would look for in a high reliability SBC is passive cooling with a heatsink under 55C even at continuous 100% CPU usage.

Yes that's also a fair point. Keeping the heat down is definitely a priority.
 

Online nali

  • Frequent Contributor
  • **
  • Posts: 901
  • Country: gb
Re: High Reliability SBCs
« Reply #5 on: July 27, 2020, 01:08:48 pm »
Do you actually need a SBC? For example as you've mentioned NTP it sounds like you'll have this connected by some means, if that's by a cellular modem you can get modules or terminals with high level scripting (Python, Java etc) to run applications. Or use a MCU with an IP stack?

Reliability is only partly a component issue. Reliability is a system design issue.

^^This. I spent some time working at a systems integrator where they would download & stuff a Debian desktop distro on to a SBC, tweak it a bit and punt it out because to build a custom kernel "would be a lot of work and we're very busy". Needless to say "busy" involved time fixing things...  :horse:


 

Offline tggzzz

  • Super Contributor
  • ***
  • Posts: 23122
  • Country: gb
  • Numbers, not adjectives
    • Having fun doing more, with less
Re: High Reliability SBCs
« Reply #6 on: July 27, 2020, 01:17:37 pm »
Reliability is only partly a component issue. Reliability is a system design issue.

You can start by defining what you mean by "reliable" for your uses, if you haven't already done that.

Often it is helpful for you to define what isn't needed. That allows you to concentrate on essentials, and often suggests system-level techniques to increase your system's value.

Yeah that's a fair point. We need it to run constantly for multiple-year timescales without requiring hardware-level intervention since that means a long trip out to the arse end of bugger all to make a repair or replacement.

Nice shout on the "don't need" list. I'll have a think about it.

Don't forget reliability is probabilistic, not absolute.

Do understand the difference between "high reliability" and "high availability". The telecom network is based on standard components configured as a highly available system.

Do understand the graph of reliability vs time for a triple-mode redundant system.
There are lies, damned lies, statistics - and ADC/DAC specs.
Glider pilot's aphorism: "there is no substitute for span". Retort: "There is a substitute: skill+imagination. But you can buy span".
Having fun doing more, with less
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf

 

- - -
-->