Linux Desktops Send NASA Rovers to Mars

joestar writes "It's not a secret that Linux has been used at NASA for a long time, and it appears that they have been using it quite extensively on the desktop. From the article: 'At the JPL, it is common to see Red Hat Inc., SuSE or Mandriva Linux running on users' desktops alongside Windows. [...] that's still a lot of Linux on the desktop.' More surprisingly, they seem to be reluctant to use Linux on servers: 'Our personal view is that Linux, period, is only for the desktop. We don't run our main servers on Linux, because there are too many flaws in main Linux kernel.'"

Comments

What we do not know

[SirCyn]

Our greatest strength is to know our flaws. I think any OSS appplies here.

Re:What we do not know

[SatanicPuppy]

I think the biggest issue anyone could have with Linux is that it comes with too much stuff by default.

For a mission critical app with a zillion dollars worth of hardware riding on it, I might be more comfortable putting my faith in a much tighter, more easy to audit OS. Not that there aren't Linuxes like that, but they're usually not supported by the big Linux support companies, and that is the second reason why I might go with IBM or SUN, for example.

Re:What we do not know

[tomhudson]

We don't run our main servers on Linux,

... you'd think they could find someone to run linux on their servers ... its not like it takes a rocket scientist

oh, right, thie is JPL ... :-)

Re:What we do not know

[Malor]

The biggest issue anyone could have with Linux is that it fucking breaks.

NASA has this dead on. When you're dealing with failures that can cost millions, the 2.6 kernel is simply not reliable enough. Hell, if you're dealing with failures that cost thousands, it's not reliable enough... and most server failures cost at least that much for midsize and larger companies. Downtime is really expensive. And you're entirely likely to have it with 2.6.

We in the open source community have this collective groupthink that Linux is extremely stable. It ISN'T, not anymore. 2.2 was incredibly robust... in my opinion, one of the best pieces of software ever written. 2.4 was problematic but eventually mostly stabilized... it still has occasional issues with unusual hardware combinations, but by and large it's pretty solid. 2.6, on the other hand, has been a complete nightmare from the point of view of pretty much any professional sysadmin. Constant regressions, constant bugfixes, and they won't fucking leave it alone and let it stabilize.

It takes YEARS to shake the bugs out of a piece of software, but they refuse to commit to backporting bugfixes to anything older than a couple of months. They just wave their hands in the air and expect 'the distros' to fix their coding errors, instead of doing it right in the first place. So everyone else has to scramble around and backport bugfixes, or else adopt a pile of new features every couple of months. Then we get the bugfixes for the new code, along with MORE new code, with yet MORE bugs. Rik van Riel has stated, I kid you not, that's he's perfectly okay with only one in three 'stable' kernels actually being, you know, stable.

So of COURSE NASA doesn't use it on servers. Linux is not being written for reliability. It never was, it just happened by accident. It was ALWAYS intended as a desktop Unix, but it was so amazingly robust in its early, simple incarnations, that it was pressed into wide server duty. And instead of realizing why Linux became so popular, the devs seem to have stayed with their desktop orientation... and in fact have changed the development process so it's more fun for them. It's a nightmare for everyone ELSE, but now they don't have to deal with the boring, nasty grunt work of making sure the code actually works in every single case.

I can't find the quote now, but at one time, Linus said something along the lines of "Hardware is inherently stable; there's no reason why software can't be written to the same standard." But he seems to have forgotten that completely. Linux has turned into the Windows of Unix.... lots and lots of features, not so hot on reliability. You KNOW it's a problem when Ars Technica, one of the most competent geek websites anywhere, switched back to Windows for _stability_. The Linux dev team should be completely ashamed of themselves for that one.

I've been using Linux since late 93 or early 94. I put it into real production service in business in '98 or so, and relied on it for years. All we had back then was ext2, which lost data if the box crashed... but it didn't matter much because it never crashed.

That is SO not true anymore.

Re:What we do not know

[Neil Watson]

I agree that 2.2 was super stable and that 2.4 was less so. However, I've found that 2.6 is very stable. Certainly more that 2.4. A lot of stability issues come down to how you build your kernel. Build only what you need. Don't use things marked experimental. Research your kernel/hardware combination before hand.

Linux

[taskforce]

Linux's kernel may be flawed, but the GUI is perfect, right?

Pot calling the kettle black?

[Anonymous Coward]

Any organization that manages to screw up metric and imperial on a several billion dollar project has no right to comment on "flaws in the kernel".

Re:Pot calling the kettle black?

[ACME Septic]

Sure they do. Haven't you ever heard the phrase, "It takes one to know one!"

Re:Pot calling the kettle black?

[Limecron]

Lockheed, not NASA, provided the Imperial units causing the loss of the $125 million Mars Orbiter project.

http://edition.cnn.com/TECH/space/9909/30/mars.met ric.02/

Linux desktops?

[tehshen]

I thought they used space shuttles to send things to other planets. Oh, the things you learn...

Re:Linux desktops?

[Dunbal]

I thought they used space shuttles to send things to other planets.

      Naw, the shuttle is just to put things into low earth orbit. To get to the planets you need the desktops...

Ironic, isn't it?

[grasshoppa]

It's ironic, isn't it, that most companies and corporations find the exact opposite to be true.

Says something about Nasa, don't it?

Re:Ironic, isn't it?

[MightyMartian]

Yes, that they've been using Unix and variants for decades, and while they may not feel confident of Linux on servers, they work in an environment where *nix is extremely common place, and very likely desirable as compared to alien operating systems like Windows.

Re:Ironic, isn't it?

[MightyMartian]

I'll wager that the vast amount of NASA's management and engineering software is written in C and for *nix platforms. Now, if you're responsible for deciding what operating systems to put on workstations, and the choice is between an operating system that is in many respects foreign in architecture and development tools, and an operating system which is very similar to what's running on your servers and offers identical or nearly-identical tools, which one will you pick?

This is what Microsoft has never really understood, and because it's never put that much effort into getting *nix software to easily port over (they did have good intentions with NT 3.5), there are a huge range of applications, particularly at the high end, which will likely never be found on a Windows machine.

Interesting article

[coastin]

I have also seen Linux being used extensively by NOAA in the last few years for weather data tracking and forecasting. I run a combination of Linux, Mac and one XP PC in my lab at Texas A&M Agriculture Program where we run a network of crop-weather (Crop Weather Program for South Texas) stations and an extensive on-line decision support system for cotton growers along the Texas coastal plains. The servers are Linux along with my desktop and notebook, there are four Macs counting one notebook and one MS XP machine to run a Campbell Scientific application that communicates with the weather stations. If Campbell Scientific were to offer a Linux build of LoggerNet I would not need the XP box at all.

RTFA?

[tgrimley]

Brack's team instead runs Sun Solaris 8 for its main servers. He cited the OS's more stable, reliable, and longer lifecycle as one of the key reasons for this deployment.

That's the sentence after the one you are talking about.

Just Ask Yourself

[ackthpt]

Which would you prefer to be used for mission critical applications, where failure can cost hundreds of millions of dollars in time and material, not to mention lost opportunity.

if bill gates' wife was admitted to the hospital and put on life support managed by one particular OS, which OS do you think he'd actually trust?

Re:Just Ask Yourself

[ackthpt]

Bill Gates has a wife? Maybe I won't be a virgin for the rest of my life!

Are you a billionaire, too?

The name of the game is Linux

[digitaldc]

In Brack's group however, the name of the game is exclusively Mandriva Linux. In fact, it should be noted that this Mandriva deployment is the largest in the world.
"In terms of [Linux] distros for the overall lab though, we actually run more Red Hat Linux," Brack said. But, regardless, that's still a lot of Linux on the desktop.

So let me get this straight, the name of the game is exclusively Mandriva Linux, but they actually run more Red Hat Linux?
Is Mandriva really exclusive to the game? or is that actually Red Hat? I'm so confused.

Re:The name of the game is Linux

[Anonymous Coward]

There is an aspect about IT at JPL that does not come across in the article. There are something like 5500 employees at JPL and there are roughly 3-4 computers for every employee. Now if you consider that most employees only have one computer, it is more like 6 computers for every scientist or engineer.

This means that we have a diverse and decentralized IT make up. Certain core services are within a particular group. But every engineering group is left to themselves regarding how to apply computer resources to projects. So the group that Brack provides administration for (roughly 200 users I think) exclusively uses Mandriva Linux (and only on workstations). While JPL as a whole uses Red Hat for most installs and JPL as a whole does not have a problem with using Linux for servers. In fact you will find almost every type of server OS represented (yes, even the *BSDs).

Having said that, our relationship with Sun is largely historic. Since JPL is run by CalTech we have always gotten that incredible education pricing on Sun hardware and since it ran so well too it was used A LOT in every sector of IT at JPL. Sun has lost some ground to Wintel, Lintel and Mac OS over the years. But it is still highly respected at JPL and heavily used.

Rumor Control

[flood6]

More surprisingly, they seem to be reluctant to use Linux on servers: 'Our personal view is that Linux, period, is only for the desktop. We don't run our main servers on Linux, because there are too many flaws in main Linux kernel.

They mentioned Windows in the summary, so to head off the "so they use Windows servers over Linux???" comments, TFA said they run Solaris on the servers because they have found it to be more stable, reliable, and have a longer lifecycle. I'm not saying I agree, just clarifying a summary I can see leading to pointless comments.

No need to ask.

[ScentCone]

if bill gates' wife was admitted to the hospital and put on life support managed by one particular OS, which OS do you think he'd actually trust?

No doubt you're implying that he'd opt for one of the heavily scrutinized Linux distros with native support for emergency room cardio equipment? What, Red Hat hasn't done that yet? No widespread testing yet for Hoary Hedgehog, EKG Edition?

If I were Bill, I'd probably choose Win2K... but that's not really the issue. It's the application, the drivers, and the comm interfaces letting the machine talk to the life support stuff. I'd want to be hooked up to whichever of those has seen the most hours of use in the most places under the most circumsntances. And if the O/S that happens to have been the platform on which all of that use-time was racked up happens to be Bill's, then so be it. Win2K is very, very stable - especially when you're not surfing to Russian pr0n sites, installing free casino software, or trying to overclock under a beta video driver for maximum frag resolution.

Re:No need to ask.

[panthro]

Most of that "life support stuff" has been running on low-level embedded control systems and, in more complicated cases, proprietary UNIX variants, since before either Win2K or his wife were a twinkle in Bill Gates' eye.

Systems like that, used in medical, industrial and military applications, make Win2K look as stable as an overweight donkey on ice skates. Windows, like most general-purpose things, is a clumsy, plodding hack that does a mediocre-at-best job of a variety of things instead of a really good job at one. Linux, as the term is used most of the time, falls under the same category, albeit perhaps somewhat less clumsy and plodding. I wouldn't trust a desktop PC to run my toaster.

Generally, control devices used in critical applications like life support machines are rock solid. There are some PLCs at the plant I work in that have been running continuously for years in a harsh environment (aluminum foundry) without incident.

Flaws at linux?

[bubulubugoth]

No doubt, but... where, and which are?

There is a lot of NASA contrib at networking, drivers, etc, but the kernel flawed, that is interesting.

TFA also says that the NASA is a SUN shop, and they are still using Solaris 8, and they have no doubt to switch to Solaris 10. So this means that they have 6 years old hardware? Becose, I dont think that new SUNs hardware is supported by Solaris 8...

I wonder, do they buy comodity hardware? Becose, if you are planning to roll a massive linux installation, the first thing you do, is check for hardware compatibility...

The article, actually isnt very useful, to help for or detract the linux usage at servers or desktop. It would be nice, that this kind of public funded enterprises, to had their methodology at public access, so we can learn more about that kind of stuff...

This is so confusing...

[Spy der Mann]

Linux dominates the server market, and the NASA says it sucks. But they use Linux for desktops, where the market is dominated by Windows - which sucks.

Now I'm confused! :-S

Scientists and Linux go way back

[i_should_be_working]

In physics and math departments at universities and national laboratories around the world it's not a strange thing to see people using Linux.

Ready for desktop since...

[just_another_sean]

1993 according to this.

I quote:
During 1992 and 1993, the Linux kernel gathered all the necessary features it required to work as a replacement for Unix workstations, including TCP/IP networking and a graphical windowing system (the X Window System). Linux also received plenty of industry attention, and several small companies were started to develop and distribute Linux. Dozens of user groups were founded, and the Linux Journal magazine started to appear in early 1994.

Just one of several examples of doing a google search on Linux History. I personally have bene using Linux on my desktop and servers since I discovered Slackwarein 1996. (Thanks Patrick! :-)

ESC

[drewzhrodague]

I keep pressing the escape button, but I can't seem to get off this crazy planet. What am I doing wrong?

In NASA...

[menkhaura]

In NASA Linux is only for desktops!

Solaris, Tru64, Win2k3 Server...

[Anonymous Coward]

I work as a sysadmin in Houston for the aerospace company that operates as NASA's prime contractor for ISS. We work very closely with the tech people over at the Johnson Space Center on the Station and Shuttle contracts and perform tasks for them that include large-scale analysis and number crunching (we recently handled the foam debris analysis for STS-114). We use a mix of systems on the back-end, but the breakdown generally is Windows 2000 AS & 2003 Enterprise Server for misc. (non mission-critical) application hosting and e-mail and printers and general office automation stuff, and Solaris or Tru64 or VMS(!) for anything flight- or vehicle-related, and dedicated mainframes for large (or legacy) tasks.

They don't run their servers on Linux, eh?

[pscottdv]

We don't run our main servers on Linux

Oh, really?

So explain this guy (www.top500.org).

Re:They don't run their servers on Linux, eh?

[snarlydwarf]

That's a desktop machine, of course.

For someone with a Very Big Desk.

I work in Mission Control and...

[EccentricAnomaly]

this article is BS.

I've worked in mission control at JPL for several years and I've never seen Linux used as a Desktop OS by more people than I can count on one hand. In fact JPL has a contract with Lockheed Martin to supply Desktop computers that makes it very hard to run anything other than Windows or MacOS.

But, I have seen many workstations running Linux, and many servers running Linux. In fact, I think virtually all navigation is now done from Linux servers. And when workstations and Servers don't run Linux they run Solaris. There used to be some HPUX machines around, but you don't see many of them anymore after the crap HP put people through with HPUX-11 (what the hell was HP thinking by dropping fortran-77??)

Anyway this article is complete BS. Much like one MacWorld ran a while ago claiming JPL used Macs for everything.

Re:I work in Mission Control and...

[mdman]

But I read it on Slashdot! you mean its not true! Oh the shame! ;)

Re:I work in Mission Control and...

[Arandir]

But, I have seen many workstations running Linux...

To most people, a workstation is a desktop. "Desktop" itself is a very nebulous term, originally meaning a computer small enough to put on your desktop, but now meaning any client system you directly interact with. You also have the problem of many people using "desktop" to refer to a GUI operating environment. A "workstation" however, comfortably fits into all of the above. Workstations are desktops.

Re:I work in Mission Control and...

[Perl-Pusher]

Well I work at NASA Langley and have worked at Goddard. In peoples cubicles you will find mostly RedHat Linux or a few have SGI & Sun workstations. We have two Clusters running RedHat. We have about 25 Origin servers (about 300k each), 20 or so Sun & SGI workstations and a very large AMASS tape storage system (5 silos).

Most cubicles have a windows machines in addition to the others, that windows machine is the only thing supported by the CONITS contract. JPL mission control is not indicative of all of NASA.

Most developers I work with have Linux desktops and or laptops, some dual boot with windows. System admins around here seem to prefer SGI's, they scarf up many of the used SGI workstations as they get upgraded or bid on pallets of discarded ones. Some have Linux Boxes, and a group of them are using FreeBSD! I had a compact Alpha running Redhat until about 6 Months ago. Now I'm using FC4 on a AMD64 system, and I have a company bought powerbook.

Among the scientists it's about 65% Windows with Linux making up almost all of the rest. Windows Laptops were running almost 100%. But every meeting I see an new Powerbook on the desks. Last Science team meeting I attended had about 5 powerbooks and the same number of windows laptops. I remember 3 years ago I had the only powerbook in the room. Mine is still the only one that dual boots Yellow Dog. It's my uber geek badge ;)

I worked at NASA once

[rkanodia]

A few years ago, I had an internship at the Air Traffic Control lab at Ames Research Center in Mountain View (technically Moffet Field is its own city, but whatever), California. The people were nice and pretty good at what they did, and the desktops were mostly Red Hat, but the IT system there was pretty weird. I sat in a cubicle next to a one filled with unused desktop machines and monitors. Pentium 2/3, G3/G4 Macs, 17" CRTs, all kinda of stuff that was just 1.5-3 years old. Even so, people who were coming in would get new computers. Why? Because you can't just take one of the computers from the storage cubicle; you have to fill out forms and it needs to go through a bunch of processes to make sure that it works, that the hard drive is wiped, and a clean install of the OS is performed.

Obviously, the IT department would rather just open up a new machine than spend a bunch of effort refurbishing an old one, so they made the paperwork to have an old machine put back into service much more complicated than the paperwork to order a new machine. Furthermore, there was a tactical element involved: I ended up with a brand new, top of the line machine because my boss wanted one, but wasn't due for a new computer for a couple of years. If I remember correctly, because I was an intern, he was able to justify the purchase for 'a new employee' on the accounting side, while keeping the ownership rights from IT's perspective - so when I went back to school, he took the machine I'd been using and - you guessed it - dumped his old one in The Cubicle.

NASA has orbiting brain lasers?

[Rhinobird]

This was the first thing that popped into my head:

http://www.ubergeek.tv/article.php?pid=54

I doubt it...

[Ice Wewe]

When I went and got a tour of JPL, from a staff member, not a tour guide, I looked in on their server clusters, they were running Fedora Core 4 smp. Now, perhaps this is just an isolated case, but everywhere I looked, there were computer's running Mac OS X, Linux, and Windows. When we got a tour of the main control room, where they had terminals displaying the data being received from the space probes/landers/craft, they were running linux. Therefore, I tend to doubt whoever said that the linux kernel wasn't 'stable' enough for their purposes. Perhaps they're just trying to keep Microsoft happy, because when I was there, it [Windows] certainly wasn't the majority OS.

Fine, how about this one?

[pscottdv]

JPL Visualization Supercomputer

I expected that.

[jd]

I worked at NASA Langley in the late 1990s, and there was a move towards the desktop and away from X terminals. However, the desktops they gave everyone were Windows based and did not have Cygwin installed, making them damn-near useless as all the applications were X.

As I recall, I was one of the first there to really kick up a fuss about Linux, and since that time I'm very glad to say that most of the computational fluid dynamics code (ie: the stuff they use to simulate aircraft and jet engines) almost universally supports Linux. Not quite - the stuff for migrating CAD to grids and back isn't Linuxified - but everything else seems to be.

One of my really fun tasks, whilst there, was to migrate FROM Visual Basic to X/Motif. Yeah, sure, Motif wouldn't have been my first choice either, but I got the interface to work many times better under that than it did under VB.

About the only thing I really hated about Nasa Langley was their insistance on using rsh for all network connections (even over the Internet) and their use of .rhosts files on all internal machines. It was a major hole and I can remember expressing my displeasure to the chief of network security at Langley. Strangely, I was sacked shortly thereafter. Since then, I've learned rather more tact, but I guess my core complaint hasn't changed a lot. It's all fine and good, talking about "bugs in the Linux kernel", "FIPS-180", etc, if it gets the organization to do better than they would otherwise. When it is used to cover their ass because they know what they have is crap but they don't want to risk change, then I regard their excuses as little more than the Peter Principle in action.

It sounds, from what I'm seeing today and what the article and others are saying, that NASA has largely come out of cryogenic storage and is showing signs of a fully functional intelligence.

Only signs? Sure. Donald Becker (who also worked at NASA) didn't just complain about problems with the network drivers - he wrote his own damn drivers, and it took a very long time for anyone to come close to writing drivers even a fraction as good. Nor did he complain about the lack of clustering capability, he wrote his own - bproc - and the supporting tools that collectively became known as Beowulf.

And the rest of NASA's problem is...? Sure there are bugs in the kernel. And NASA has a small army of programmers fixing inconsequential bugs in old Fortran code that has been in solid use for 20+ years. Let's say that NASA held a 2 month bug-squelching fest. It might still not get Linux to the point where Goddard or JPL were willing to use it on production servers in general, but I'll bet you anything that:

• It'll mean the Fortran codes running on Linux boxes will run more reliably, for less effort, than could have been achieved by continuing to fix the Fortran for the same length of time
  
• It'll inspire the regular kernel developers and may even encourage those on the fringes to become kernel developers
  
• As most servers don't need the full range of capabilities, NASA will be able to produce a rock-solid "micro Linux" designed specifically for specialized servers
  

NASA has made a big difference to the software available for Linux (at least, if you're interested in moving objects), and in the distant past made a revolutionary difference to Linux networking. They could make a revolutionary difference again, if they loosened up on the distribution of their Open Source and/or got another Donald Becker to get some critical segment of the kernel working absolutely perfectly. I'm not holding my breath, but there is so much potential there that they'd be foolish to ignore it.