Computer Servers 'Stranded' in Space (bbc.com) 89
A pair of Hewlett Packard Enterprise servers sent up to the International Space Station in August 2017 as an experiment have still not come back to Earth, three months after their intended return. From a report: Together they make up the Spaceborne Computer, a Linux system that has supercomputer processing power. They were sent up to see how durable they would be in space with minimal specialist treatment. After 530 days, they are still working. Their return flight was postponed after a Russian rocket failed in October 2018. HPE senior content architect Adrian Kasbergen said they may return in June 2019 if there is space on a flight but "right now they haven't got a ticket." The company is working with Nasa to be "computer-ready" for the first manned Mars flight, estimated to take place in about 2030. The company is also working with Elon Musk's SpaceX.
I bet it says "Shipped with Amazon" (Score:5, Funny)
Shoulda used UPS, not Amazon's own shipping.
Re: (Score:2)
Space qualified hardware is not trivial. In the old days what was qualified would be one or likely two or more generations earlier than current which was a big deal when each generation typically brought significant performance improvements. Nowadays I assume it's more bulk. Still I'd think they'd leave the system up till it stopped working and then schedule to bring it back for analysis.
I believe the hardened hardware is still around a decade old, performance-wise. It's just that performance hasn't changed much in a decade. I do wonder what they do about cooling. You don't have to underclock a modern server processor by much to dispense with the fan, but you still need some airflow over a large surface area heat sink. Maybe they just solved the problems related to shaking/vibration well enough to support a sizable heat sink.
Or of course you can just use mobile components everywhere, bu
Re: (Score:2)
Even with a large heatsink, there would need to be fans involved, because heated air wouldn't rise away from the heatsink in microgravity
Re: (Score:2)
Good point! Though it seems there is case airflow through heatsink to the station.
I just wonder about the resiliency of cooling fans after launch - normal components really aren't designed for heavy vibration at multiple gs.
Re: (Score:2)
I do wonder what they do about cooling. You don't have to underclock a modern server processor by much to dispense with the fan, but you still need some airflow over a large surface area heat sink.
What utility is a fan in the vacuum of space? The real solution is not blasting to space a CPU that generates a lot of heat. That means they'll be "space hardened" SPARCs, or PowerPC, ARM, or Pentium grade chips.
Re: (Score:2)
It's not a vacuum system. It runs in the pressured area if the ISS, and can dump heat to the ISS via a heatsink.
The eventual trip to Mars will be similar, I think.
teammasters want $2500/hour for space runs! (Score:2)
teammasters want $2500/hour for space runs! and that is just the workers pay.
Re: (Score:1)
Somebody probably wants to inspect the boards, chips and other components for radiation and mechanical damage.
Re: (Score:3)
They probably want to study them to see if there has been any damage from radiation or other issues, but leaving them running longer may help them to find more useful information.
Re: (Score:3)
Running in low Earth orbit won't expose them to nearly as much radiation as they'll get in interplanetary space.
Re: (Score:1)
True, but it's more than at home.
Of course, I expect what they're really testing is the marketing advantages of having computers in space.
Re: (Score:2)
True, but it's more than at home.
There are room-sized irradiation machines used to sterilize food [wikipedia.org].
It should not be too hard to set up a test rig here on earth.
Do SSDs use depleted boron-11 as a dopant? That is a cheap and obvious first step to making them rad-hard.
Re: (Score:2)
They probably want to study them to see if there has been any damage from radiation or other issues, but leaving them running longer may help them to find more useful information.
That's just what they said about Spirit [xkcd.com].
Re: (Score:2)
Didn't theFukushima and Chernobyl robots give up this information quite quickly?
Re: (Score:2)
they need them back to do an tear-down
Re: (Score:2)
Re: (Score:2)
The point isn't to have a supercomputer in space; it's to test the long term effects of space on modern computer hardware.
The ISS's *built-in* computers run on a late 80s 80386SX processors, which were produced with 1000 nanometer semiconductor technology. The processors on this computer have minimum feature sizes of 45 nm or less, and are thus more likely to be radiation sensitive. The processors for these computers were taken right off the Intel assembly line with no special hardening.
The effect of spa
Re: (Score:2)
The effect of space operation is particularly of interest in specifying the mission computer for a manned Mars mission
Except that in LEO, the computers are protected by the Earth's magnetic field. Sending COTS parts to Mars would be a bad idea, especially if the crew's life depends on it.
Re: (Score:3)
I don't think anyone is suggesting non-hardened COTS electronics is a good idea for a Mars mission.
The radiation in LEO is not the same as beyond the van Allen belts, sure, but it's not the same as at the Earth's surface either.
Re: (Score:2)
The radiation in LEO is not the same as beyond the van Allen belts, sure, but it's not the same as at the Earth's surface either.
People keep saying this... has slashdot been overrun by folks applying for bullshit research grants?
Re: (Score:2)
https://doi.org/10.1016/j.zeme... [doi.org] among other sources.
Re: (Score:2)
non-hardened COTS electronics
All COTS is non-hardened.
Re: (Score:2)
No hot swap?? (Score:2)
redundancy power supply as well as some of the redundant solid-state drives.
For an long mission they need to be hot swap or have say 4 PSU's that only needs one.
For disks maybe an 3-4 disk raid 1 setup.
Backup up disks
FULL RESTORE IMAGES
FULL LIST OF ROOT / ADMIN passwords maybe hard coded ones.
BIOS RECOVERY DISK / USB.
ETC
Re: (Score:1)
redundancy power supply as well as some of the redundant solid-state drives.
For an long mission they need to be hot swap or have say 4 PSU's that only needs one.
For disks maybe an 3-4 disk raid 1 setup.
Backup up disks
FULL RESTORE IMAGES
FULL LIST OF ROOT / ADMIN passwords maybe hard coded ones.
BIOS RECOVERY DISK / USB.
ETC
Uh, the article was essentially bitching about how specialized hardware being tested for ruggedness had not been returned on time, which is about as exciting as reading about a late library book, and tells you how much of a non-story we really have here.
Your rant covers some hardware but mostly software DR planning, which essentially had little to do with a physical experiment.
Re: (Score:2)
"Well!", said the PHB, "the wings didn't fall off the plane on the first flight, let's not look any closer for stress fractures. Guys, make 10k more of these now"
Empty Dragon (Score:2)
SpaceX Dragon is going up there real soon with no crew.
Belt it in to an empty seat and bring it back.
Re: (Score:2)
This.
Re: (Score:2)
Is this a common practice in the Uk?
Re: (Score:3)
nO. eVERYTHING iS fINE oVER hERE
I've had HP loaners... (Score:2)
And they rarely call to get them back. Unless somebody else needed them....they were on permanent loan.
Strange experiment (Score:4, Insightful)
I am not sure what they are trying to prove with this experiment. The environment inside the space station is hardly any more daunting than sitting on a table in Cupertino. The real issue is when you get above the Van Allen belts and get a few zaps from solar flares.
I am sure they know this, but, also, processors for the space environment are also a perfectly well-known quantity. Analysis alone should get them a very reliable answer on life time, upset rate, etc.
Re: (Score:1)
Re: (Score:2)
It's not only NASA that wants to know how these things perform.
Re: (Score:3)
When there's a hull breach, the loss of capacitors is probably the least of their worries.
Re: (Score:2)
When there's a hull breach, the loss of capacitors is probably the least of their worries.
During a hull breach, yes, assuming the loss of capacitors isn't happening in a form that's adding to the ongoing disaster. After a hull breach has been dealt with, I think any survivors will be quite interested in having all systems critical to survival working...
Re: (Score:3)
Analysis alone...
I would prefer that we do more than just analysis before building a $100 million device and sending it to Mars. The next logical step is to use the ISS that we have. If the result of the experiment is "What happened is exactly what we expected to happen" then great! It's the occasional "Oh no! It should have lasted years, but it lasted 2 days because we failed to consider *insert unexpected phenomenon here*"
Re: (Score:3)
People have been doing that experiment, and validating it with flight data, since 1964. 1985, "we predict a double-bit memory error once every 28.5 days". Cut to 2018 - observed upset rate = 28.496 days.
There essentially no question what space does to computers. And in any case, we still send things costing far, far more than $100 million into space (that is, *almost everything launched* is much more than that) and every one of them already has one or more computers in them, they follow the predictions ple
Re: (Score:2)
we predict a double-bit memory error once every 28.5 days". Cut to 2018 - observed upset rate = 28.496 days.
That's GREAT! We ran an experiment that validated our predictions! DO NOT attack studies that aim to test predictions or reproduce existing results. That's the bread-and-butter of science. We have a big problem in science today with funding not going to reproducibility studies. It's more exciting to fund new research, but yet we constantly see headlines about studies not being reproducible.
There essentially no question what space does to computers.
As a consumer of science news, you might think this is well understood and boring and should not be done. But the
Re: (Score:2)
Running tests on the ISS isn't going to be very useful for when you want to go to Mars. Cosmic radiation gets much worse, and when a solar flare comes your way, there's no place to hide.
Re: (Score:2)
That's not how engineering works. You don't run only in the target environment. Suppose you need to design a computer to run for 5 minutes at 200 degrees C. The engineers would not simply test it for 5 minutes at 200 degrees C. They would run "characterization testing" which means they test it at 100, and 150, and 250. They run it in high humidity and low humidity. They run it at 150 for 5 minutes, then 200 until it dies, to see if running at the lower temperature reduced the longevity. The run it at
Re: (Score:2)
I am sure they know this, but, also, processors for the space environment are also a perfectly well-known quantity.
I would just hire the folks who made the Mars Rovers to build the space computer.
Being how long they lasted, they seem to know what they are doing.
Re: (Score:2)
Mars rovers have special rad-hard technology: https://en.wikipedia.org/wiki/... [wikipedia.org]
Very expensive and rather slow compared to off-the-shelf computer stuff.
Re: (Score:1)
JEDEC memory scrub rate recommendations change already significantly within troposphere to reach some theoretical reliability number. The Mars mission needs locally accessible high performance computing capability, and this should be a first step to investigate the weaknesses in machines that are deployed rarely even at very high altitude environments due to increased costs of cooling and cosmic rays. Lowest bar and all that.
Re: (Score:2)
IMO, it's the ride on the rocket that's the interesting test. Most electronic components deal very poorly with extended heavy vibration, even at 1g. (If you ever move cross-country, you'll find a lot of consumer electronics mysteriously stop working soon after the move.)
Repurposed (Score:2)
2017 (Score:2)
Re: (Score:2)
Re: (Score:2)
If your server purchases only have a useful life of two years, you're doing it wrong.
Re: (Score:2)
The point is to recover the hardware so they can analyze it to see what failed and why.
Ob (Score:2)
Hoy them out of the airlock.
They'll come down sooner or later.
Complaining about a 3-month delay? (Score:1)
Talk about uptime! (Score:1)