MarkDH102
Posts: 377
Joined: Fri Feb 13, 2015 3:18 pm

Re: Interesting interview with Eben in Physics World

Thu Nov 07, 2019 7:20 am

In terms of using the Pi in an industrial environment, I cannot believe it would be in any kind of "critical" control type of operation - just monitoring as stated in the article. The SD card solution is just too unreliable for 24/7 running...and each Pi would need a UPS plus some really good noise suppression on the incoming mains...

Heater
Posts: 14437
Joined: Tue Jul 17, 2012 3:02 pm

Re: Interesting interview with Eben in Physics World

Thu Nov 07, 2019 7:49 am

Perhaps so, but there are billions of non-critical applications that can be helpful. As the article points out.

SD card reliability is hugely improved by using it as a read-only device.

The power supply issue you mention is true no matter what SBC you use.

I have no particular reason to believe the Pi board and the components on it are any less reliable than any other electronics, if one is prepared to accept and cater for the "domestic" specification rather than demand wide operating temperature range, radiation hardening, and tolerance to other environmental extremes.
Memory in C++ is a leaky abstraction .

User avatar
Gavinmc42
Posts: 4293
Joined: Wed Aug 28, 2013 3:31 am

Re: Interesting interview with Eben in Physics World

Thu Nov 07, 2019 8:00 am

The SD card solution is just too unreliable for 24/7 running...and each Pi would need a UPS plus some really good noise suppression on the incoming mains...
There are solutions for those.
I have had Pi's running for years,
One email this morning, "up time - 148days, 47mins".
That would be since last power out, it comes back up when power is restored.
Does nothing critical except monitor temperature and send an alarm if over temp.
Very simple circuit and mostly shell script code.
Industrial controls would be fancier.
I'm dancing on Rainbows.
Raspberries are not Apples or Oranges

Ernst
Posts: 1272
Joined: Sat Feb 04, 2017 9:39 am
Location: Germany

Re: Interesting interview with Eben in Physics World

Thu Nov 07, 2019 9:40 am

The SD card solution is just too unreliable for 24/7 running...and each Pi would need a UPS plus some really good noise suppression on the incoming mains...

Code: Select all

[email protected]:~ $ uprecords
     #               Uptime | System                                     Boot up
----------------------------+---------------------------------------------------
     1   204 days, 03:00:21 | Linux 4.9.80+             Fri Jul 20 14:20:38 2018
->   2   181 days, 03:11:48 | Linux 4.9.80+             Fri May 10 08:25:31 2019
     3    83 days, 17:00:17 | Linux 4.9.80+             Fri Feb 15 13:35:56 2019
     4    61 days, 03:51:42 | Linux 4.9.80+             Mon Apr 30 07:23:38 2018
     5     8 days, 00:49:18 | Linux 4.9.80+             Tue Jul  3 15:11:41 2018
     6     5 days, 20:00:14 | Linux 4.9.80+             Sun Apr 22 22:44:54 2018
     7     4 days, 19:12:08 | Linux 4.9.80+             Sat Feb  9 17:06:24 2019
     8     3 days, 12:00:15 | Linux 4.9.80+             Wed Jul 11 20:14:08 2018
     9     3 days, 05:07:34 | Linux 4.9.80+             Wed Apr 18 17:11:09 2018
    10     2 days, 11:19:06 | Linux 4.9.80+             Sun Jul 15 08:29:04 2018
----------------------------+---------------------------------------------------
no1 in    22 days, 23:48:34 | at                        Sat Nov 30 10:25:53 2019
    up   563 days, 15:45:27 | since                     Wed Apr 18 17:11:09 2018
  down     4 days, 02:40:43 | since                     Wed Apr 18 17:11:09 2018
   %up               99.276 | since                     Wed Apr 18 17:11:09 2018
Still using the same SD card running fail2ban with currently 8250 IP addresses banned in the last 181 days.
The road to insanity is paved with static ip addresses

jamesh
Raspberry Pi Engineer & Forum Moderator
Raspberry Pi Engineer & Forum Moderator
Posts: 24945
Joined: Sat Jul 30, 2011 7:41 pm

Re: Interesting interview with Eben in Physics World

Thu Nov 07, 2019 9:52 am

MarkDH102 wrote:
Thu Nov 07, 2019 7:20 am
In terms of using the Pi in an industrial environment, I cannot believe it would be in any kind of "critical" control type of operation - just monitoring as stated in the article. The SD card solution is just too unreliable for 24/7 running...and each Pi would need a UPS plus some really good noise suppression on the incoming mains...
So many definitions of critical. I wouldn't use a Pi to run an airliner (Although perhaps Boeing would), but the are less 'critical' applications that it could handle, with the right work done on it. The CM3 board with eMMC for example, would be appropriate in some critical environment.s

But as others have said, there is a colossal amount of non-critical industrial applications out there. We sell a lot of kit into that sort of area.
Principal Software Engineer at Raspberry Pi (Trading) Ltd.
Contrary to popular belief, humorous signatures are allowed. Here's an example...
“I own the world’s worst thesaurus. Not only is it awful, it’s awful."

jahboater
Posts: 5031
Joined: Wed Feb 04, 2015 6:38 pm

Re: Interesting interview with Eben in Physics World

Thu Nov 07, 2019 10:06 am

MarkDH102 wrote:
Thu Nov 07, 2019 7:20 am
The SD card solution is just too unreliable for 24/7 running...
Really?

All my Pi's run 24/7 and have done for years.
Never a problem, and SD card corruptions simply don't happen.

User error, or unreliable power source.

Compared to most PC's, the electric bill for the ARM powered Pi is tiny, so its not worth shutting them down when not in use.

Heater
Posts: 14437
Joined: Tue Jul 17, 2012 3:02 pm

Re: Interesting interview with Eben in Physics World

Thu Nov 07, 2019 11:03 am

Boy do I hate this ongoing SD card reliability debate:

Always some poster points out that SD cards are unreliable and/or that they have had SD card failure or file system corruption on SD (not the same thing)

Always some other poster offers their anecdotal evidence that they have Pi running for years without SD card/fs problems.

Neither of which is particularly useful information.

Me, I had an SD card get very hot and start emitting smoke after a few months of continuous operation. The Pi was fine when the card was replaced. No really. What do you make of that?

SD cards do fail. File systems do get corrupted on random power down. How frequently that happens and if it is more than an inconvenience for a users application is up to the user. If it might be an expensive bother then take steps to minimize the risk.
Memory in C++ is a leaky abstraction .

User avatar
rpdom
Posts: 16087
Joined: Sun May 06, 2012 5:17 am
Location: Chelmsford, Essex, UK

Re: Interesting interview with Eben in Physics World

Thu Nov 07, 2019 11:10 am

Heater wrote:
Thu Nov 07, 2019 11:03 am
SD cards do fail. File systems do get corrupted on random power down. How frequently that happens and if it is more than an inconvenience for a users application is up to the user. If it might be an expensive bother then take steps to minimize the risk.
So true. I have had two SD cards fail in the 7 years I've been using them. Both of those were full size cards in a Pi 1B and 1A. Of course I had backups which I could build new cards from.

I've had more USB memory sticks fail in PCs (with much less use) than I have SD cards.

jahboater
Posts: 5031
Joined: Wed Feb 04, 2015 6:38 pm

Re: Interesting interview with Eben in Physics World

Thu Nov 07, 2019 11:39 am

Heater wrote:
Thu Nov 07, 2019 11:03 am
File systems do get corrupted on random power down.
Isn't that true of any of type computer system disk?
Its not SD card specific (its an OS problem).

The only difference perhaps is that SD cards are tiny and cheap, so maybe people don't take as much care of them as they did for expensive large spinning rust drives.

laurent
Posts: 328
Joined: Thu Jul 26, 2012 11:24 am

Re: Interesting interview with Eben in Physics World

Thu Nov 07, 2019 11:51 am

SD card may fail for some reasons, but solutions or workarounds exists:
- Mecanical: could be glued (could be applied to camera connectors also)
- FS corruption: Read only rootfs on a good quality UHS I SD card
- Unexpected power loss: UPS HAT or powerbanks

If you just install a regular Raspbian on a compute module's eMMC, it will also fail for the 2 last reasons above.

The Pi Foundation/Trading not only made an affordable SBC, they opened the competition with new rules.
Although they made wonderful products (the last Pi 4 is really awesome), the Pi Zero is clearly a huge achievement of the foundation.

tpyo kingg
Posts: 724
Joined: Mon Apr 09, 2018 5:26 pm
Location: N. Finland

Re: Interesting interview with Eben in Physics World

Thu Nov 07, 2019 12:26 pm

jahboater wrote:
Thu Nov 07, 2019 11:39 am
The only difference perhaps is that SD cards are tiny and cheap, so maybe people don't take as much care of them as they did for expensive large spinning rust drives.
I've had only two microSD cards go bad so far, out of ten or so. One just stopped working and couldn't be rewritten. Another seemed to have cracked. The cracked one was in the same machine, untouched, for days (or weeks) and then one day would simply just not power on. When I removed the card, I noticed the slightest of bends at a weird diagonal angle. I have no idea how to achieve that angle since it didn't match any of the ways one can put the card into the machine either correctly or incorrectly.

But be that as it may, when I got a Raspberry Pi for the first time a few years ago and figured out that the whole system could be swapped out simply by switching cards I was stoked. That was a most excellent design decision which greatly helps learning and experimentation. Contrast that with the Beaglebone which, although is also an excellent board, one has to spend 15 to 30 minutes to wipe, reinstall, and restore systems when moving between incompatible projects. So having swappable microSD cards opens a whole new way of experimenting.

The inclusion of GPIO, referred to in the article, was another excellent choice. Overall, the Raspberry Pi brings many of the good things that were possible with the old Apple II, II+, and IIe series back in the 1980s albeit in a packet close to the size of a deck of cards. It was possible, with the right tools, knowledge, or help, to build and add in peripherals or use or write software (in a choice of languages) or anything in between. The point I see with that whether hardware or softwre, is that unlike with most other computers, they allowed one to follow a comfortable, gradual curve from casual use to intensive programming or hardware projects. The Raspberry Pi achieves that handily.

By the way, the same magazine has a second article on Eben Upton about his background, beginning with his start in physics:

https://prod-physicsworld-iop.content.p ... index.html

Obviously it also covers a bit about of the Raspberry Pi.

Heater
Posts: 14437
Joined: Tue Jul 17, 2012 3:02 pm

Re: Interesting interview with Eben in Physics World

Thu Nov 07, 2019 12:59 pm

@jahboater,
Isn't that true of any of type computer system disk? Its not SD card specific (its an OS problem).
Yes of course. Except I would say it's not an OS problem. Perhaps an OS configuration problem. Fifteen years ago I was involved in rolling out almost a thousand Linux based embedded systems across three cities in Scandinavia. They operate in harsh conditions and are very expensive to repair/replace if they fail. I was pleased to be in meetings with the engineers that use those systems recently and hear them praise the reliability of those units.

@tpyo kingg
I've had only two microSD cards go bad so far, out of ten or so
Exactly. These anecdotes don't help much. A 20 percent failure rate is appalling in an industrial setting where down time and repair is expensive.
when I got a Raspberry Pi for the first time a few years ago and figured out that the whole system could be swapped out simply by switching cards I was stoked. That was a most excellent design decision which greatly helps learning and experimentation. Contrast that with the Beaglebone which, although is also an excellent board, one has to spend 15 to 30 minutes to wipe, reinstall, and restore systems when moving between incompatible projects. So having swappable microSD cards opens a whole new way of experimenting.
That is all very true. And I love it.

In the context of industrial use it does not help much though.

Note: I'm old school, when I think of reliable micro systems I think program storage in PROM and EPROM, or big old FLASH chips. Just now I have been advising on interfacing with or replacing such systems that have been running 24/7 for getting on 30 years!
Memory in C++ is a leaky abstraction .

jahboater
Posts: 5031
Joined: Wed Feb 04, 2015 6:38 pm

Re: Interesting interview with Eben in Physics World

Thu Nov 07, 2019 1:22 pm

Heater wrote:
Thu Nov 07, 2019 12:59 pm
@jahboater,
Isn't that true of any of type computer system disk? Its not SD card specific (its an OS problem).
Yes of course. Except I would say it's not an OS problem. Perhaps an OS configuration problem.
Yes, its not a problem as such, its just that any modern large OS will have a disk cache which is vulnerable to unplanned power outages. Windows, MacOS, Linux, UNIX, spinning rust, SSD's, SD cards. Its just the way things are.
My point was that its not a problem specific to SD cards only - as many here seem to imply.

I'm convinced people think they can just yank the power cord on a Pi, but would never do it to a $2000 tower PC!

The real difference for the Pi is that its quick and simple to put things right after a disk problem!

Heater
Posts: 14437
Joined: Tue Jul 17, 2012 3:02 pm

Re: Interesting interview with Eben in Physics World

Thu Nov 07, 2019 2:28 pm

Personally I think one can expect ones data to go missing on an abrupt power down. Or at least very recent versions of it. But if the operating system won't boot after such an event that is a bug. A design flaw. In the same way I expect my car to restart after running out of gas when I put some more in.

It's a shameful bug we have all lived with on 2000 dollar PC's and such for so long that we think it's normal.
Memory in C++ is a leaky abstraction .

pfletch101
Posts: 569
Joined: Sat Feb 24, 2018 4:09 am
Location: Illinois, USA

Re: Interesting interview with Eben in Physics World

Thu Nov 07, 2019 3:48 pm

Heater wrote:
Thu Nov 07, 2019 2:28 pm
Personally I think one can expect ones data to go missing on an abrupt power down. Or at least very recent versions of it. But if the operating system won't boot after such an event that is a bug. A design flaw. In the same way I expect my car to restart after running out of gas when I put some more in.

It's a shameful bug we have all lived with on 2000 dollar PC's and such for so long that we think it's normal.
I don't think that your two statements are consistent. I agree that it is unreasonable to expect an electronic device as complex as a computer system - be it a Windows PC or an SBC running Unix - to recover with all its data intact from an unplanned power outage. However, unless the OS itself runs completely read-only, which has consequences that most PC users (at least) would find unacceptable, it is inevitable that data which it needs to boot will occasionally be corrupted under these circumstances. In fact, my experience is that modern Windows systems (I can't speak for Apples) do a remarkably good job of recovering from unexpected power losses. I have my main systems on a UPS, so this almost never happens to me now, but an associate had the habit of shutting down her computer and then immediately powering it off. It took a while (and many uneventful subsequent startups) before an investigation of user data corruption on the system revealed the problem.

W. H. Heydt
Posts: 11466
Joined: Fri Mar 09, 2012 7:36 pm
Location: Vallejo, CA (US)

Re: Interesting interview with Eben in Physics World

Thu Nov 07, 2019 6:47 pm

pfletch101 wrote:
Thu Nov 07, 2019 3:48 pm
Heater wrote:
Thu Nov 07, 2019 2:28 pm
Personally I think one can expect ones data to go missing on an abrupt power down. Or at least very recent versions of it. But if the operating system won't boot after such an event that is a bug. A design flaw. In the same way I expect my car to restart after running out of gas when I put some more in.

It's a shameful bug we have all lived with on 2000 dollar PC's and such for so long that we think it's normal.
I don't think that your two statements are consistent. I agree that it is unreasonable to expect an electronic device as complex as a computer system - be it a Windows PC or an SBC running Unix - to recover with all its data intact from an unplanned power outage. However, unless the OS itself runs completely read-only, which has consequences that most PC users (at least) would find unacceptable, it is inevitable that data which it needs to boot will occasionally be corrupted under these circumstances. In fact, my experience is that modern Windows systems (I can't speak for Apples) do a remarkably good job of recovering from unexpected power losses. I have my main systems on a UPS, so this almost never happens to me now, but an associate had the habit of shutting down her computer and then immediately powering it off. It took a while (and many uneventful subsequent startups) before an investigation of user data corruption on the system revealed the problem.
All of which is why every computer (including all the Pis) in the house are behind UPSes. That vastly reduces the chances for abrupt power drops.

Heater
Posts: 14437
Joined: Tue Jul 17, 2012 3:02 pm

Re: Interesting interview with Eben in Physics World

Thu Nov 07, 2019 7:15 pm

W. H. Heydt,
All of which is why every computer (including all the Pis) in the house are behind UPSes.
Anecdote:

A couple of years a go we had a PC running Linux in the office that was not seriously critical but would be inconvenient if it went down. So we gave it an UPS.

After a year or so I arrive at the office in the morning to find the UPS is screeching from it's peeper and the Linux PC is down.

On investigation I find the UPS is dead. I plug the PC to the mains and try to start it. It will not boot. Turns out both it's hard drives are now unreadable.

I have no idea if something killed the UPS and the PC drives at the same time. Or if the UPS committed suicide and took the PC drives with it.

Make of that what you will...
Memory in C++ is a leaky abstraction .

Return to “General discussion”