Nearly every supercomputer (of the 500 most powerful) runs Linux

[Commie Pinko Dirtbag]

470, to be exact.

And no, the other 30 aren't Windows. Well, 5 of them are. The other 25 are Unix.

http://www.itwire.com/business-it-news/technology/39471-nearly-every-supercomputer-runs-linux

Just as with the previous list of the Top500 Supercomputers, the Oak Ridge National Laboratory's Cray XT5 'Jaguar' leads the pack with a sustained performance of 1.76 PFlop/s. However, a new Chinese entrant called Nebulae has achieved a higher peak performance of 2.98 PFlop/s (compared to Jaguar's peak of 2.33 PFlop/s) and is placed second overall.

Of the 187 new entrants, all but one are running some variant of Linux and in fact 470 of the Top 500 run Linux, 25 some other Unix (mostly AIX) and the remaining 5 run Windows HPC 2008.

The Chinese systems Nebulae and 7th placed Tianhe-1 are both based on a hybrid design with each node comprising a pair of Intel Xeon processors attached to a pair of GPUs (NVidia for Nebulae and AMD for Tianhe-1) as accelerators. These two systems helped China to take second place for total PFlop/s (9.2%), still trailing USA with 55.4%.

Australia has just one entry in the list, the Bureau of Meteorology / CSIRO HPCCC system occupying position 112 and running the CentOS operating system. Down one from the previous list, New Zealand systems hold 7 places – 6 of which are located at Peter Jackson's WETA Studios (the machine to fall off the list was a lesser system also at WETA). WETA has five identical supercomputers which fell from position 193 in November 2009 to 279 in the current list.

[no limit]

1. Yeah, labs love to use linux because of license costs

but that doesn't really tell you anything about the efficiency of windows vs linux.

[Commie Pinko Dirtbag]

3. Wow.

You really DO have an axe to grind.

Well, maybe you work at Microsoft. Then it's understandable. If my livelihood depended on appearing to be a fanboy I'd be a fanboy too.

[zipplewrath]

10. Well, a strange comparison

Super computers tend to be focused around a single activity, with other capabilities being added as "needed" or as they are useful. Windows is intended as a general purpose operating system intended to be able to do a wide variety of tasks. I'm no windows "defender" and think it can be easily demonstrated to be an inefficient operating system. But Linux and Unix generally are great operating systems for people who want to maintain an operating system. But if you just want a computer.....

[johnaries]

32. EXACTLY! The OS's were designed for 2 different markets.

Windows is designed for maximum user-friendliness. It performs a lot of functions in the background for the user. But, as the old saying goes, "the more complicated the plumbing, the easier it is to stop up the drain."

Unix is absolutely the most stable OS there is. But, the license fees are outrageous. Linux is basically a Unix-clone with a few changes. Linux is actually a little easier to use. For example, if you are setting up multiple printers Unix there are separate files for each printer you have to edit separate files for each. In Linux, you can set up all of them in /etc/printcap). But Linux is not quite as stable as Unix. Almost, but not quite.

But, with both Unix and Linux you have to know what you're doing. They were designed for "geeks". Windows and Mac were designed for the general public.

Comparing Windows/Mac to Unix/Linux is apples to oranges.

[HopeHoops]

45. I don't know. NT 4.0 was a pretty fucking good server (SP4 at least) and not all that "friendly".

I really don't think they've come up with anything since (or before for that matter) that could touch it for reliability and ease of administration. I'm holding out hope for Win7. Despite my early concerns with it, I'm fairly impressed so far. The stability thing remains to be seen. I've gotten six months out of an NT 4.0 box without a reboot (may be longer, I'm being conservative), and that's great for a M$ OS. In contrast, I had a dual 486 box running SCO that went two years without a reboot and the only reason we had to do that was to install a full-room UPS system.

[no limit]

48. Our windows 2003 servers stay up for months on end

any reboots are the result of power failure (really shitty power grid here and our UPSes are low power because of budget constraints) or because of scheduled maintenance.

I can't remember the last time our windows servers crashed on their own.

[HopeHoops]

103. I had a Win98 SE box go at least a year without a reboot. Had to move it.

That's usually the case. You're going for an uptime record and "Shit, I've got to move the damn thing."

I've got UPS boxes all over the house, and not just for computers. All three kids have one for their laptops and cell phone chargers, same with our bedroom, and I've even got one in the kitchen for the two cordless phone systems (and a fish tank). If the power goes out and you don't have a corded phone handy, you had BETTER have the cordless base on a UPS! As unthinkable as it is, we had a situation where NONE of our cell phones were charged when the power went out. The land line was available (Mother's Day call to my mom was the issue). There's one for the entertainment system, but mostly for the surge protection. The only thing on the battery are two desk lamps. That's only an APC 350, I think. I use APC 500s for almost everything else - good price for the performance.

My main zone (home) is run by an APC 1300 and three APC 500s. The 1300 has an LED graph and it barely hits 30% capacity with five systems - even when they are all active. This laptop is technically the sixth, but it never hits the meter one way or the other, probably because of its own battery. The monitors and external drives run off of the 500s.

Actually, I don't trust any other brand than APC (no, I have no affiliation with the company other than customer - not even stockholder). I had too many of the others fail in the work place where the bean counters ordered shit because it was cheaper (and sometimes by as little as $3 per unit). The sub-standard crap isn't worth the cost savings if you have downtime or it doesn't protect the equipment in a thunderstorm. You can buy a SHITLOAD of APC 500s for the cost of a high-end workstation!

[HillWilliam]

75. /etc/printcap has been out of style for a few years now

Linux printing is even easier to set up than Windoze printing these days. Point and shoot, and it actually works. Even the wireless HP printer that never did work right on XP or Vista works first time every time on Linux.

Fedora's probably the "geekiest" of the lot, but I find it's still more reliable and easier to configure than Windoze. It's all I run at home. I've tried Ubuntu and for your average user it's probably a lot more stable, forgiving, and easy to configure than any MS product so far. (And yes, I even liked NT... it may have been dinky, but it worked.)

I may have to use Windows at work (fight with the bugs constantly in a development environment -- I'm a software developer for a living). I don't want to fight my machine at home -- I just want to plug it in and have it "go". That's why I've used Linux for years. The less I have to fiddle with my machine, the more work I get done. The more work I get done inside, the more time I have to go play in my gardens and dig in the dirt :)

[Recursion]

77. Amen

For all the talk of user-friendliness in Windows, for me the whole point of Linux is that you plug stuff in and it just works. The less time I have to spend worrying about how my machine is set up (that is, the more sane policy decisions that have been made by a distribution maintainer) the better.

Hmm... Has anybody ever thought of setting up a Windows Distribution? Make a sane set of default policy decisions for the OS and a key set of apps and deploy them?

[HillWilliam]

90. Step One would be to unscrew the inscrutible

and I don't think even Redmond has sweet-FA idea what's in their own OS. Having also had my ass burnt off a few times from MS patches, I think it would be one nasty chore to set up a Windows distro then another nasty chore to fend off the flotillas of attorneys. Just thinkin' out loud over here ...

[Titanothere]

94. I don't know what distro you're running...

but for me the adage has always proven true "Linux is free if your time is worthless".

I can't tell how many times I've been in XFREE86config or whatever trying to change my displays or, and this is the current state of affairs, get my split spacebar keyboard to be properly configured.

If there's one thing Linux doesn't do to for the non-programmer, it's "just work".

Standard answer from Linux guy "Well, just recompile your kernel with these obscure libraries after editing these random makefiles and you'll be good to go! You do have the latest libc & gcc right?"

Pass. I have a Linux box at home purely as something to recycle my old hardware on.

[Recursion]

99. Xfree86config?

1999 called. It wants its windowing system back. I haven't touched an X config file in 3 years.

[Titanothere]

108. some wounds take a long time to heal

[no limit]

12. What did I say that is not accurate? I have no axe to grind, people like you do

as soon as you hear microsoft you assume it must be bad without knowing a thing about what you are talking about.

No, I do not work for microsoft, I wish I did. But I deal with microsoft products on a daily basis. So when I point out that what linux or mac fanboys are saying is inaccurate its not because I have a axe to grind, its because these people are talking out of their collective asses.

[Commie Pinko Dirtbag]

16. See #8. -nt

[no limit]

18. I already responded to #8. #8 was posted long after you posted

The person that posted that atleast made good points, you didn't. All you did was attack. So what did I saw that was not true?

[Commie Pinko Dirtbag]

19. Well, gee, sorry for not refreshing the thread every 5 seconds.

:eyes:

Pathetic.

[no limit]

22. I am still waiting for you to tell me what I said that wasn't true

you were making some really wild accusations about me. So you must have had a reason for saying those things. What did I say that wasn't accurate?

[mkultra]

55. That windows and linux are equally efficient

which is actually kind of ridiculous to even propose. S

[no limit]

59. I said they were equally efficient? hmm...you might want to reread my post because I never said that

there are things windows is more efficient in and there are things linux is more efficient in.

My simple point was that the fact that all these labs use linux as the platform for their super computing has to do with a lot of factors and in no way means that windows is automatically more efficient. Which is 100% accurate on my part.

[mkultra]

61. You are making a tacit accertion

that the use of linux in clustering and supercomputing is, in majority, due to licensing instead of speed and is not indicative of a performance gap. This statement is patently untrue. The driving force behind clustering and supercomputing IS performance.

The only thing windows does better the Linux is the one thing Linux doesn't do. Push a GUI out to the user. The cost of managing a Linux cluster can be just as expensive because what you give up in licensing fees you must pay for in expertise. Windows allows lower skilled professionals the ability to manage the system while linux has a steeper learning curve requiring more talented administration.

Frequently, Apple or Microsoft will offer assistance to a university and subsidize much of the cost to be part of the project. In the end though, the only one that can compete with Cray or IBM is linux.

[no limit]

64. What you are saying is simply not true.

You are trying to brush aside licensing as if it doesn't matter, yes it does and it matters a huge deal. You don't know what considerations these people looked at before deciding on linux yet you are making the assumption with nothing to back this up that it was about efficiency. You simply don't know that. That was my point and it continues to be my point.

When you said I somehow suggested that windows is more efficient than linux you were making that up, I never said such a thing. windows can be more efficient for certain applications while linux can be more efficient for others. The fact that the top 500 super computers out there don't use windows tells you nothing about efficiency, it doesn't tell you anything about anything. What you have to do is look at the main forces behind the decision of going with linux over windows. And I'm sure cost isn't the only issue but I gurantee you it's a big one when you consder how many thousands of computers are required.

[mkultra]

68. see, and this is clearly just a matter of judgment

or rather, lack of experience, on your part.

If you dont wish to suggest that windows is faster, fine. I can easily say that windows is NOT faster for ANY application. windows is EASIER.


I'm not brushing aside licensing, im saying that with windows licensing, you get support. The support you do not get with Linux, results in higher pay for your administrator due to expertise. Its not quite a wash as Linux still turns out cheaper, but its that cost swap is often ignored by inexperienced IT professionals.

Most clustering environments are prepared to spend a chunk of capital. They are after performance as the goal though and achieving that goal becomes an act of balancing the cost performance ratio.

Even if money was not an issue, I would go with SUN before windows which has similar licensing costs.

[Recursion]

70. On SPARC, I hope?

I love Solaris on SPARC; OpenSolaris on x86 makes me want to gouge my eyeballs out.

[no limit]

74. Windows is not faster for any application?

Wow, what an amazing statement to make. I'm sure you have evidance to actually back this up? What benchmarks have you done to make such a statement?

[mkultra]

102. I dont need beenchmarks

I have worked with both extensively and have never seen a case where win was faster. If you consider the architecture alone, you could easily perceive why that is true.

The industry wide perception agrees with me. Do you have any benchmarks that show otherwise?

[no limit]

105. I don't, I just wouldn't make that claim until I saw benchmarks

I have seen benchmarks for file I/O which show linux beating out Windows. But based on that benchmark I can't simply assume that Linux is always faster no matter what. That will all once again depend on application.

[Recursion]

65. Compiz?

The only thing windows does better the Linux is the one thing Linux doesn't do. Push a GUI out to the user.

I dunno. The new compiz-enable distros seem to do that pretty well, and put Aero to shame. I usually stick with an older environment like GNUStep because I'm used to it, but the new ones are very eye-candy-ish.

[mkultra]

69. well, i was really just being generous to his point

[Renew Deal]

23. Microsoft doesn't do "supercomputer operating systems."

This really has nothing to do with Microsoft. You might be able to call him a fanboy and he might say you're clueless. :shrug:

[LLStarks]

47. Yup. Windows HPC is quite a miserable supercomputer operating system. nt

[Xithras]

51. Actually, they do. And they're quite successful. They're just not used for the massive clusters.

Linux based supercomputers demand fewer cycles for operating system overhead, providing more cycles for your research calculations. When you're doing atomic modeling or advanced physics work, every cycle counts. Licensing costs are also a huge factor...to increase power, you typically increase the number of physical computers within the computing cluster. Particularly large research supercomputers can contain thousands of individual computing nodes, and the Windows licensing for that would be horrendous. That's why Linux dominates the high end research world.

Most supercomputers in the business world don't have to work nearly that hard, and handle much simpler tasks (like global sales forecasting, or accounting modeling). The computing clusters rarely exceed a dozen physical servers, and supercomputing projects typically have a lifetime measured in days. Microsoft HPC clusters tend to do well in that environment. The consistent API's and turn-key environment mean that a single programmer can write out a calculation, throw it into the HPC environment, and close out the project with a resultset in just a few days. Linux supercomputers, on the other hand, typically require that the calculation software be written from scratch, and it's not unusual for researchers using those superclusters to spend weeks or months just setting up a single run.

It's apples and oranges. The two types of products serve radically different ends of the supercomputing spectrum. No research institution would ever do subatomic particle modeling on a Microsoft HPC cluster. No Fortune 500 company would ever do a 5 year global sales growth by market model using a Beowulf cluster. It has nothing to do with one platform being better than another, and everything to do with using the right tool for the right job.

Fanboys from any camp are a pox on technology.

[no limit]

53. Thank you for that information, great write up

Wish I was smart enough on clusters to write something like that up ;).

[Recursion]

67. The lightweight process model and IPC do lean more towards clusters than the VMS/Windows model

Sort of the converse of why heavily-threaded games have performance troubles on Linux: making a thread is the same thing as making a process (actually it can have a little more overhead depending on your open descriptors). But that also means making a new whole process is orders of magnitude faster on Linux.

[entanglement]

107. True.People familiar with capacity computing (web/database) often don't realize how different HPC

(capability computing) is. I'd say apples and sea-horses :)

Also, science / engineering / computational biology workloads are typically extremely floating-point heavy. An average web / business application hardly comes close.

[L0oniX]

118. IMO ...that poster is a MS shill. MS is not beyond planting shills on blogs.

[hughee99]

7. Yes, labs in general love linux because of the license costs, but

the sort of people who would buy a supercomputer aren't usually looking to save a few nickels on the OS, if it's not as good.

[HopeHoops]

8. The license cost IS attractive!

It would be interesting to see performance stats on identical super computers (Linux vs. Windows), but when they're spending that much for a machine to get the highest performance, it is a fair bet that the OS cost is low-grade noise on the balance sheet and performance wins the day.

I can say from my own experience that identical machines running various flavors of UNIX (not Linux in particular - SCO, AIX, etc.) easily outperform Windows. I would expect the same to be true (and perhaps with a growth curve) as the power of the machine increases. That said, for the home user, there are some apps that just don't have an equivalent in the Linux world. That "some" is a continually shrinking quantity, and that's a good thing. I use OpenOffice on Linux, Mac, AND Windows and have no trouble bouncing files around on thumb and external hard drives between the three. I've also had very few issues with opening, changing, and saving M$ Office files in OpenOffice. Curiously, I run into FAR more problems getting M$ Office to work with its OWN files from earlier versions (except for Excel).

[Commie Pinko Dirtbag]

11. Your post has sense.

Therefore, it makes Nordic Redmond Baby Jesus cry.

[no limit]

13. I would love to see such benchmarks too

I'm curious what applications do you see liunx running faster than windows?

[Statistical]

17. Exactly bash Microsoft all you want but to pretend that license cost isn't relevent is silly.

The top supercomputer in the world for example consists of 25,000 computers and 50,000 processors (250,000 cores).

HPC 2008 runs $475 per node. Not sure if Microsoft considers a discrete computer or a discrete processor a node.

Anyways thats $12 million to $24 million in software licensing.

http://www.microsoft.com/hpc/en/us/how-to-buy/volume-licensing.aspx

[charlie]

15. Source, source, source

I'm sure having your hands on the source counts for more if you're looking for maximum oomph from your rig.

[HopeHoops]

21. DING! DING! DING! DING! - We HAVE A WINNER!

The key to the kingdom is in the source. That's the beauty of Linux in general. If there's a bug, someone posts it on a forum, someone figures out a fix and posts it, the SysAdmin downloads it, checks the code to make sure it makes sense, and applies the patch - MOST of the time without causing downtime.

Now, as YOU mentioned, if you're running a highest-end cluster, Linux is like having your own machine shop for your late-60's muscle car. Does anyone feel like waiting around for Microsoft to figure out a "fix"? Given their track record on home machines, I would expect every "fix" creates three new problems.

[no limit]

25. Aaccess to source is nice, but you are suggestin applying a OS patch in these enviroments is trivial

it's anything but.

Engineers working on these systems aren't going to apply some patch off an internet message board without serious testing. Yes, windows patches might take longer to put out but atleast when they are put out you know they went through extensive testing by a very large and well financed testing team. Obviusly in these cases these labs and universities decided that's something they can deal with by doing the testing themselves but that can be a problem with linux.

[Recursion]

31. Are you really saying Microsoft's patch record is better than Red Hat's?

OK, you've clearly never been a sysadmin. One Microsoft patch once took down two of my database servers and shifted everybody's appointments on the shared calendar to the wrong time zone.

[no limit]

35. I don't know if Microsoft's is better than Red Hat's but red hat isn't free

specifically because of what I said. They support their linux distro so they test patches.

I doubt these places are using red hat just because of the cost associated with it, I could be wrong on that.

Does fedora have a better track record on patches than Microsoft does? Are those patches officially supported by Red Hat? I ask because I honestly don't know.

And yes, I have been a sysadmin. Just not as a linux administrator, I work with windows based clients and servers.

[Recursion]

41. No, but Debian is, and it's marginally better than Red Hat

And their release cycle is designed specifically for systems administrators.

[no limit]

43. Is debian really better at releasing official patches that Microsoft?

Is there anything to back up that claim?

I just find it hard to believe that Microsoft who has a huge team worth millions of dollars behind their patch effort would be outdone by a community of non-profit developers.

[Recursion]

49. Roughly a decade of being a sysadmin is my back up for it

I just find it hard to believe that Microsoft who has a huge team worth millions of dollars behind their patch effort would be outdone by a community of non-profit developers.

I certainly don't find it hard to believe. Debian is developed and maintained by people whose purpose for developing and maintaining it is having the best possible operating system they can. Debian never broke their tzdata with a patch; Microsoft did in 2008. Debian didn't break Gnumeric's math with a security patch; Microsoft did break Excel's. Debian has not broken GRUB with a security patch, Microsoft broke some versions of XP's bootloaders with a patch this February.

This is what I meant about Debian's release cycle. They don't have a "patch tuesday", and they don't try to cram as much into the OS as they can to meet an arbitrary release date. This causes no shortage of whining among the upstream package maintainers, but it's the Right Way To Do It.

[no limit]

50. These patches are not officially supported are they? And in your decade of being a sys admin..

how many of those years were spent on windows enviroments?

My problem is I don't have a lot of linux experience, but I have about a half a decade worth of experience dealing with windows server enviroments and I've been playing with client versions of windows since around 1998. I've had nothing but good luck with Windows patches over those years. And you might have had great luck with Debian patches. But I don't think either of us are qualified enough to make the judgement that one is better than the other at patches without actual specifics. So if you have specifics I would love to know about them.

I can not think of a single nightmare I've had in the last 5 years with a windows patch. They had some issues, sure. But I'm sure the same can be said about Debian or any other linux distro. Overall Microsoft is timely with critical patches and when they do come you can be assured they were fully tested by a very professional and very well finananced team (that's not to say you don't do your own testing for major patches).

[Recursion]

52. There have always been Windows machines in every shop I've been at

I can not think of a single nightmare I've had in the last 5 years with a windows patch.

Well, as I alluded, we lost a major deal because after a timezone update in 2006 everybody's meetings appeared at the wrong time on the calendar.

I've only been lucky enough to have one shop absolutely free of Windows servers, though that one still had Windows workstations (I came to a rough and grudging peace with Windows 2000 as a desktop OS).

Anyways, the reason I jumped on this is that regressions in Microsoft patches are specifically the thing I think of most when I complain about Windows as a business platform.

I'm not sure what you mean by "officially supported". Nobody buys support from the Debian foundation. But Debian "officially supports" its current stable and immediate previous stable release, and tests patches and updates to isolate and mitigate regressions from them.

[no limit]

54. That's why you have to test these types of patches before you apply them

same goes for linux I'm sure. I don't know what solution you were using for your calendars (exchange?) but microsoft can't always test their patches against your enviroment. I do know that we have used exchange for a long time and I haven't ran in to a timezone problem in the production enviroment. Although in testing the timechange congress enacted was a pain in the ass when dealing with active directory.

Just because you had shops with windows doesn't mean you are an expert on using them. Lets be honest, on a percentage in the last decade how much work did you do on linux vs windows? 90% vs 10%? 50/50?

I am not qualified to say much about linux so I don't go around saying that windows is better than linux simply because I don't have a huge understanding of linux.

When I mean officially support is that you know it's been fully tested prior to release and if any problems come up you have a number to call for official support (or other avenues of official support). Having a community to help support you is nice, no doubt. A lot of times I prefer community support over official support. But it is not a substitute for official support in mission critical applications.

[Recursion]

63. Well, no, you really *don't* have to test Debian patches. That's the point.

Assuming you've kept to the Debian Stable environment, they're tested against regressions.

In some senses it's apples and oranges: Debian packages not only your kernel and userland but also your web server, database, application server, mail server, desktop environment, etc. The maintainers spend an amazing amount of time and energy on this.

Funny you should mention the time split. Let's take my last big shop: 80 Linux servers and 14 Windows servers (+2 OpenBSD servers but we can leave them out for now). I spent half of my work time maintaining the Windows servers. And I'm actually pretty good at Windows administration, which is how I got that job. Some of that was patch regressions. Some of that was an application server (not Microsoft's fault) that refused to release memory from its VM once it had been allocated. Ever. Some of that was just the random shit that always seems to plague Windows (every two days or so an ODBC client will just drop all named pipe connections and refuse to reset them without restarting the entire subsystem; the boss's Outlook pst gets corrupted because he keeps 4 gigs of mail; the receptionists' computer absolutely refuses to relinquish its IP address and no amount of ipconfig /releaserenew will convince it to play nice -- and, ok, the last two were desktops so I won't count them; the Terminal Server that will not display the users mice unless the KVM has that server selected at the time; the domain controller that has decided to stop doing WINS until you reboot it; the SQL server that hangs on a full backup, crashes, and restarts with different collation rules, breaking all our search functions... this is why I don't manage big shops anymore)

Now, I haven't administered any Windows system later than 2K3. I'm sure it's improved. I just wasn't going to wait for it to.

But it is not a substitute for official support in mission critical applications.

I want to avoid profanity so I won't start my story of the appallingly lame "official support" Microsoft offers and how useless it was when I needed it.

[no limit]

72. You really don't test patches on your linux machines before applying them?

I think that will be a huge problem for you. You saw this problem with windows when you didn't fully test a day light savings patch with your applications, I simply can not believe that linux is so magical you never have to test patches before deploying them to a production enviroment, I say that based on my many years of experiance working with computers in general.

I also haven't done any significant work with server 2008, it's still kind of in its infancy stage. 2003 hasn't give me any problems. I run exchange, IIS, MSSQL 2005, DHCP, DNS, and data protection manager for backups spread out over multiple servers. I have not run in to any of the problems you are having. Our enviroment isn't large enough to depend on WINS so I can't comment on that.

The pst file issue is well known and fully documented by microsoft. If you have PSTs that big there are different ways of dealing with that.

And I haven't had much need for official MS support in the past and I admit the times I did try them turned out to be a total waste of time. But we are small fish. The labs I've worked with actually have access to ondemand support reps from microsoft that will come down and address their issues which is nice for mission critical applications that cost $10,000 each hour they are down.

[Recursion]

76. Of course I test patches

I tested the timezone update for obvious breakage. Slight movements of meetings on a shared calendar is not an obvious breakage.

And, yes, I do know about the PST issues. I just don't see why I should have to when no similar stumbling block exists for Evolution or Thunderbird. (And, incidentally, no matter how big you get, there's really no excuse for using WINS at this point; this was a legacy application that required it.)

You said upthread "Microsoft can't test for your particular environment", and this is true. Debian can because they control every aspect of the distribution: web server, application server, database, desktop environment, text editor, etc. And if you leave that particular, slow release cycle "stable" setup, even for one application, you're out on your own. That's the downside. If you're on Windows, and your programmer says "we need to upgrade to Ruby 1.9", well, you upgrade to Ruby 1.9. If you're on Debian, you don't.

But, in terms of how Debian gets such high levels of QA, there's nothing magical about it: they have a whole lot of people who volunteer a whole lot of time testing every change. I'm one of them.

[no limit]

79. So debian actually supports your webserver, your SQL databases, etc?

What about additional applications you run? We like to use something called EPICS that we generally install on linux, this is a custom framework we use, Debian would have no way of testing against that. I'm sure there are thousands of other applications for linux that have this problem. Do many people really just run the applications that come with debian or do they run many additional custom applications which debian has no way to test against?

Maybe I have too much ignorance when it comes to linux but I just don't see how debian can be a full blown OS and at the same time support in it's releases every single third party application out there. Seems to me that if EPICS base releases a new version I would have to go the the epics web site and update from there, Debian won't automatically patch this for me.

[Recursion]

81. Yes, that's exactly right. Look at the "apt" network some time.

That is the entire point of running a Debian system.

[no limit]

82. Very interesting. So any linux application you want to run on debian must be in a debian package?

So for example with EPICS someone would have to keep a debian distribution of EPICS on some server for you to be able to use it?

Sorry if these are dumb questions and they might be good for another discussion, but I find this concept interesting.

[Recursion]

84. Well, it depends on what you mean

Debian comes with gcc, so you can compile whatever you want. Once you install something that is not a Debian package, it's not "Debian" anymore. It's a GNU/Linux system based on Debian. EPICS has a Linux client, but its installation makes it an example of something that's "not Debian anymore". But as long as you keep it in /usr/local, Debian promises not to break ABI compatibility for it until the next "stable" release; if it likes the libraries now, it will as long as you keep the system.

[no limit]

86. But then Debian can no longer assume that patches won't break your application

can they?

I am aware of gcc but how the application runs once compiled depends on the OS. So if Debian changes how they deal with time zones (as unlikely as that is Im just using it as an example) your application might very well have to be modified to address that change. No?

[Recursion]

89. Debian never said patches won't break a third-party app. That's why they are discouraged.

Well, not "discouraged" so much as "you take your life in your own hands". They do, as I said, promise to keep ABI compatibility. And the inherently orthogonal design of UNIX-like systems does make testing and mitigating breakages easier.

I mean, seriously, this isn't just fanboy-speak: the reason so many sysadmins like UNIX so much better is because it's much more predictable to administer.

[Cronus Protagonist]

109. A patch did that all by itself?

Does the sysadmin not take any of the blame for not testing it against important database systems before making it live?

[HopeHoops]

40. Agreed - that's the machine shop for the muscle car - and a team effort (with review and approval).

Keep in mind, a one decimal place error in some Fortran code cause a space probe to re-route itself to some void back in the 70's. I would NOT want the responsibility for giving final approval to a change in a supercomputer's OS!!! Talk about heart-burn!

I've had to give the go-ahead on a lot of critical systems, but nothing of that magnitude.

[L0oniX]

119. + Ding

[Recursion]

29. +1

Also an important part of keeping a machine secure.

[Recursion]

28. Err... I think it has more to do with their parallelization support

http://www.beowulf.org/

(Incidentally, Enterprise Red Hat or Novell installs are actually more expensive than Windows)

[no limit]

38. I haven't heard of beowulf but from what I've seen in this thread Windows has similar implementation

clustering was never my strong suite.

[mkultra]

57. Its clear you just dont have the experience to even compare the two

Most people who start computing in the windows world think its the cats meow. If you operate in a multi OS environment as i do, you come to learn very quickly which system is the fastest and most stable.

[no limit]

60. No, I don't have experiance to compare server clusters as I clearly said in the post you replied to

what's your point?

So do you work with server clusters? How many? What kind of clusters? What platforms do you use?

[mkultra]

62. I have worked with them before but do not now

My current role is in an environment that contains Hp nonstop, OS400, MVS, Linux, SUN , and Windows. I work on all of these platforms. I can tell you for sure, that windows is second from the bottom in preference only to OS400.

[no limit]

66. MVS is better than windows? Sun is better than windows? you don't seem to know what better means

better for what applications exactly? Have you tried running a domain controller in a linux enviroment? Sure, it can be done but doesn't mean it ought to be done. Does that mean because windows has an advantage in this area windows is automatically better? Absolutely not.

I would also say that exchange is much better for collaboration, messaging, and schedules than any platform Sun or linux have to offer. This is just one application.

On the other hand I always prefer Apache running on linux over IIS running on windows when ASP.NET is not in use. That's one area where linux is better in my personal opinion.

So again, better is a relative term. So when you tried to compare MVS to windows and then said MVS is better I really have a hard time taking you seriously.

[mkultra]

71. well, again, im talking about actual computing

Not that kind of user support bullshit.

[no limit]

73. Computing of what exactly?

You are ignoring the application and simply saying linux is better. That is true in some applications and not true in others. I guess that boils down to your lack of experiance (see how easy it is for me be a total dick too?)

[mkultra]

100. well, my appologies

I wasn't really trying to be rude although i guess it was. Perhaps its better for me to state the positive and just say that i have what would be considered quite a bit of experience working in varied environments compared to other IT professionals. Hopefully that doesn't sound to boastful.

Most of what windows is good for is infrastructure computing. Meaning email, domain, SAN, and such. To me, windows is a good solution in this respect due to its ease of administration.

When i say real computing i mean processing data. Collection information, transferring data, supporting a database, doing mass calculation, etc etc.

These are the type of duties that are the real needs in most IT based businesses and they require speed.

[no limit]

104. No apologies needed, these types of discussions tend to get heated for some reason

and I agree with you. There are many server applications I prefer windows for, there are other applications I prefer linux for. In the end it boils down to whatever you are trying to accomplish. So simply looking at the fact that most super computers run linux doesn't tell you much about anything.

[HopeHoops]

2. I ordered a 10.2" netbook from HP that came with Linux INSTALLED!

It also had Firefox, Thunderbird, Open Office and a lot of other goodies ready to rock. No hard drive either - all solid state. I get about 6 hours out of one charge. Kick-ass little machine.

[Recursion]

30. I've been using Linux and BSD for about 12 years now

I really, really don't miss Windows at all.

[HopeHoops]

36. I used to work with BSD and System V at the same time. DAMN was that weird.

I'd have a bunch of windows open on a VT-1200 or something and it was easy to get confused as to which was connected to which OS. So many of the switches are different. I could run either straight and could just fire off commands full of switches, but put both on the same monitor and I KNEW I was going to screw up every other time. Sometimes multi-tasking isn't so great.

[Recursion]

44. I'm too lazy to set $PS1, so I just use different shells

tcsh means I'm on Linux; ksh means I'm on BSD.

I've heard from the old timers that that was the real reason GNU caught on so widely: install it and no matter the platform you have the same switches.

[HopeHoops]

96. Yeah, but then KDE came along,....

Well, you know it from there.

On Edit:
By the way, "GNU" is a self-referential acronym - "GNU is Not Unix". And then there was YACC - "Yet Another Compiler Compiler". UNIX geeks may be geeks, but at least we have a sense of humor.

[L0oniX]

112. FreeBSD here. Wifes box uses XP which has over 90 security updates.

...and open source can be a wonderful thing.

[Warpy]

4. I'll be a happy camper

when somebody makes Linux drivers for Linksys. Until then, my Linux box remains disconnected from the net, OK to play with and learn the system but incredibly limited.

I was hoping the new Ubuntu would recognize it, but no such luck.

[Commie Pinko Dirtbag]

5. PCI wireless card, or what? -nt

[HopeHoops]

14. Um, my HP netbook came with their flavor of Ubuntu and it was as easy as any M$ box to connect.

And yes, I have a Linksys wireless router. I use WEP security and M$ XP, Vista, 7, Mac OS X, and Linux all connect with ease. I didn't have to download drivers - they came installed with the machine.

[Recursion]

33. Err.. Linksys what?

Most of their hardware has been supported in the mainstream kernel since like 2.6.26 I think?

[Statistical]

37. Really depends on the chipset especially for wireless.

Sometimes it is luck of the draw. There are some older wifi chipsets that aren't supported and likely never will be.

Good news is going forward support is better for new hardware. Many companies routinely collaborate with Linux to ensure good support in distros.

[Recursion]

39. There was a year or so that OpenBSD was very strong on wireless

I'm not sure what happened; they seem to have fallen behind Linux since then. Nothing for Atheros from the 5k series on. Broadcom essentially frozen since 2006. It's irksome, because that's my favorite platform.

[hobbit709]

34. Linksys/i.e. Cisco always wants to be difficult.

Ndiswrapper works for that problem.

[Warpy]

56. Couldn't get the damn thing to install

Remember, I have to download into a Windows box, burn it to CD, and then cart it over to the Linux box.

[hobbit709]

97. Then dump the Linksys card and get a Netgear

They're cheap and work fine in Linux.

Why burn a CD, all you need is a flash drive

[cbdo2007]

6. Everyone I know who uses Linux spends most of their time

trying to get it to do some simple function similarly to how Windows already does it for them.

My windows works great for me 99% of the time and already does way more and better than anyone with a Linux machine that I know.

[hobbit709]

42. My Ubuntu 10.04 lets me do anything I do in Windoze

Surf the internet, word process, graphics, video editing, ripping CDs and DVDs, burn CDs and DVDs, play my video and music files. and it found ALL my printers and prints just fine including the HP that Windows 7 does not play well with.
And it took less time to set it all up, including the OS install than XP or 7 did just to get the system and drivers up and running, much less all the apps.

[HillWilliam]

83. Same here

I haven't run Windows in the house in years except on one ancient Thinkpad that just lately died. I'm not missing a thing. I can do more (and more reliably) than I can with the Windows boxes I have to put up with (and continually fix) at work. Every time I have to install Windows, figure on at least a couple of days blown getting things back in order. Even then, nothing ever comes back right and each iteration has some annoying "personality quirk" you have to battle from then-friggin'-on. With Linux, I fling in a DVD, finish my configs once and for all (with GUIs, mind you) plop my backup in and within an hour or two, I'm rockin' again.

I've been a software developer and sysadmin for 30 years. In small shops you have to wear all the hats. I've worked on everything from pocket PC's to monster mainframes. For ease of setup and ease of use, give me some form of Linux. It used to be that after a long day at work fixing everyone's Windows headaches I had to come home and to that all evening, too. SCREW THAT. I threw Windows out of the house about 12 years ago and now everything just works, leaving me with some actual time to play fiddle or scratch in the garden. I like that ideer!

[no limit]

93. Can Ubuntu run photoshop? How about play call of duty?

It's cool it does everything you need it to do, but for many people that's not the case.

[hobbit709]

95. GIMP works just fine.

could care less about Call Of Duty

[no limit]

98. It probably would be fine for me if it supported CMYK

Even then not having inDesign would be a deal breaker for me.

[hobbit709]

101. Check this out

http://www.linuxjournal.com/article/7054

[no limit]

106. Interesting

Im not about to relearn everything since I'm fine with using windows but I'm glad Linux is making strides in these areas. Maybe one day it truly can replace windows.

[Recursion]

116. It does

And, not saying you are in this situation, but every graphic artist who has said that to me turned out to be using a printer that only accepts RGB colormaps. (If your printer doesn't fill up a room, it doesn't do CMYK either. Even if its toner cartridges are cyan, magenta, yellow, and black.)

[no limit]

120. I use CMYK for designs that go to 4 color presses

[eShirl]

111. Dunno, but it runs World of Warcraft like a charm.

if anyone cares

[Soylent Brice]

9. K&R

and
:popcorn:

[Renew Deal]

20. Microsoft doesn't do "supercomputer operating systems."

:shrug:

[Statistical]

24. Microsoft High Performance Computing 2008

http://www.microsoft.com/hpc/en/us/how-to-buy/volume-licensing.aspx

[Recursion]

26. Imagine a beowolf cluster of those things!

ob slashdot nostalgia.

[Commie Pinko Dirtbag]

78. HOT GRITS DOWN NATALIE PORTMAN'S NAKED AND PETRIFIED PANTS! Sorry, having a bad flashback here.

I'll go away now.

[Recursion]

115. +5, goatse

I should probably head back to /. more

[Taverner]

27. Shameless Ubuntu Plug here:

I have an IBM Thinkpad. Ran really slow with Windows XP on it. Even stripped down to bare basics, it took at least 5-10 minutes to boot and be useable

I wipe it and put on Ubuntu Linux - 30 seconds at its slowest.

[hunter]

46. If I had a supercomputer I would not pay to see Microsoft naked.

Nope, no thanks, I don't want to see Windows source code, no really, please, I must insist. Get out of my house! I'm calling the cops!

Linux is free, and the source code is open.

Once you pay to install software, or even worse, once you start paying to see source code, it's inevitable that you will pay again, and again, and again...

[RollWithIt]

58. When supercomputers need commercial social use friendly operating systems

Let me know.

I run Linux on 2 of my servers. I'd NEVER use Linux for regular day to home office kind of stuff.

[HillWilliam]

87. That's all I use

I run two home businesses just fine on it. Open Office works fine, las' longtime. Even my partner, who is about as computer-un-savvy as they get does just fine on Linux. 'S'what you get used to, I guess.

Before I threw Windows out of the house, I spent more time running around fixing everyone's machines than I did getting my own real work done. Now real work gets done, everything works, everybody's happy, and that makes me happy. (If papi ain't happy, ain't nobody happy LOL)

[cherokeeprogressive]

80. Yawn...

Windows user here, from version 3.1 on up. None of my needs have failed to be met. Ever. Been using proper backup techniques from way back, and still haven't lost as much as a photo or mp3 to system crashes or anything that anti-Microsoft folks like to crow about.

I have a five year old HP, an eight year old Dell, and an 8 year old Fujitsu laptop, in addition to the desktop I built myself about 6 years ago. All are working fine, no virus problems, no hiccups.

Take it somewhere else. Since I kinda doubt I'll ever sit in front of a "supercomputer", this story is industrial strength moot.

[KansasVoter]

85. LOL....linux. Who gives a shit. Not an end user OS. Unix is popular also.

[Rex]

88. People are kind of ignorant about the history of Unix and it shows.

The only people that should be surpised are people that don't know what Unix is.

[nichomachus]

91. And 75 percent of critical business applications run on IBM mainframes

[Romulox]

92. I use Windows 7 for work, because it's ultimately the cheapest. I use OSX at home, because it's

best.

Linux? No thanks, though I'm interested in trying Google Android on my netbook at some point!

[Cronus Protagonist]

110. They should try it with OSX

I'd like another belly laugh.

[L0oniX]

113. Nice magnet post for pro MS shills. I'll keep my BSD box.

:hi:

[Commie Pinko Dirtbag]

114. Poster discretion is advised.

Saying "meh, I like Windows and will keep on using it" does not a shill make.

Hysteria and out-of-proportion confrontational behavior, OTOH, does.

But you probably knew that already. :hi:

[L0oniX]

117. What I don't get is why everyone is so pro Winsux when if the same standard was...

applied to a car it would fall under the lemon law. MS XP has over 90 security updates which makes the latest Toyota problem kinda small. Does anyone need to mention MS's predatory past of trying to stop any competition? It's no wonder MS dominates the office and home user OS market. Do we really want just one OS with no competition? That would be equivalent of having a Walmart only market. I'd have more respect for MS had Bill not ripped off some of his beginning ideas from IBM's Dosshell.exe and Deskview. I will give him some credit for the 8 bit keyboard encoding although I believe that was part of a group effort. IMO he is the original pirate although he did pay for Dos 5 with Dosshell.exe. At the time there also was Norton Commander which I always used and later used the Norton Desktop and network drive. I just have to wonder what we could have had had MS not stomped on any competition. Thank god we at least have Firefox which thankfully runs on KDE desktop. Maybe this all will explain my attitude to those who seemingly don't get it.

[chrisa]

121. Linux sucks if you want to run games.

But Windows sucks at everything, so Linux can't be that bad. :)