Linux Making Inroads, But Not At Windows' Expense

zaphod123 writes "According to this article, the stories about Amazon (and others) switching to Linux have been misrepresented. The Linux install has replaced a proprietary Unix system, not a Microsoft Windows product. This is still "A Good Thing" for Linux, but not the downfall of Microsoft that some have foreseen."

Comments

Try it on grandma.

[OsamaBinLogin]

>When asked whether the company would ever consider replacing its
>Windows machines with Linux, Busch said absolutely not, noting
>the lack of "robust office packages" on that platform.

I often think that this excuse really is more like "we can't get naive users to use it without being crippled". Linux distros need to test their software on non-Unix people more. Humans. Typical office people who, if you ask them if they have a Mac or a Windows box, say, "Yeah, I think so".

>And Busch threw another wrench into any mass Linux migration by
>noting that the overall cost of Linux and Windows 2000 is almost
>identical after you factor in support and maintenance.

in other words, after you get done with the hassles of Linux, and the hassles of Win2k, the hassles of Linux are a little bit more. time=money, so the cost of that extra hassle is the same as the cost of Windows & its apps.

So much for free-as-in-beer.

This hassle is invisible to the Linux developers cuz they know how to fix or work around glitches when they arise. So it seems "easy to use" for them.

Try it on grandma. then report back.

Re:Try it on grandma.

[rw2]

The cost of admining a Linux network is going to end up being the same as admining a Windows network. There is basically no difference.

Well the handful of clients I've seen (or switched myself) switch to Linux certainly don't bear that out. Windows installs tend to degrade over time. In no small part because they are much more likely to be run wide open in order to allow people to get their jobs done. Once they can install their own software the registery gets polluted and the machine stops working. What next? Field trip to the workstation because the remote admin on Windows is less common and less capable than Linux.

So there you have two reasons why the cost of ownership on Windows is higher. And I haven't even started talking about resurecting infected machines, making site visits only to tell the user that there is nothing that can be done because the issue is in Microsofts ticket system but they haven't done anything with it or any of the other closed source problems.

Yes, I know that solutions exist, but this was a cost discussion and the solutions cost money. With linux they are an intrinsic property of the OS.

Re:Try it on grandma.

[einhverfr]

Hmmm....

It is true that the majority of Linux wins are at the expense of UNIX, which explains the support of Sun and IBM (if you can't beat them, join 'em). However, there is a little more to this that meets the eye.

Proprietary UNIX is great in some niche markets, but it does not compete cost-effectively with Linux or Windows. If the migration costs were the issue, why would people be moving both to Windows 2000 and Linux from UNIX, not just Linux? This trend is also evident when one looks at the Netcraft numbers (and actually reads their comments).

The real issue, though, is that adding Linux servers IN PLACE OF Windows 2000 servers for certain tasks may also be happening. If people are already switching from UNIX to NT-based OS's, then Linux's wins are definitely at Windows expense, in denied market share rather than in lost market share (Windows never had the market share to begin with). All of this is on the server.

Also, the data in the article was out of date (about 2 yerars old). Linux currently has about 2% desktop market share in the corporate environment accordign to the IDC as of last Feb. (I assume that most of these are technical workstations). But again, this may deny the 64-bit XP some market share as time goes on.

Now for the ease-of-use question:
in other words, after you get done with the hassles of Linux, and the hassles of Win2k, the hassles of Linux are a little bit more. time=money, so the cost of that extra hassle is the same as the cost of Windows & its apps.

On a corporate level, yes. On an individual level, not so sure. I have watched people who are not computer gurus struggle endlessly with the insanity of WIndows. So it is not really newbie friendly either. In fact the only newbie friendly OS is arguably Mac OS!

However, I have found that newbies that get started with Linux learn much more rapidly about their computer because it is more transparent. A good example of this is my parents, who went from being lost on Windows 95 to being lost on Red Hat 6.1. Funny thing, if I set up the desktop with their use in mind, they had fewer problems than they did with Windows 95. They started using their computer more, and now (even though they no longer use Windows) are able to help all their friends use Windows. So I think that Windows is "user friendly" because that is what people have struggled with and put a lot of effort into learning. Not that it is innately so. Expect Linux to take more of the desktop in the next few years...

Re:Try it on grandma.

[SCHecklerX]

I often think that this excuse really is more like "we can't get naive users to use it without being crippled". Linux distros need to test their software on non-Unix people more. Humans. Typical office people who, if you ask them if they have a Mac or a Windows box, say, "Yeah, I think so".

Funny how these people used to be quite proficient with things like word perfect for dos and quattro pro for dos. If they were forced to go back to that, they would use it. It is their job, after all. They are capable, but everyone gives them the pointy-clicky thing that they can use, but never really understand. It used to be that people actually understood the tools they used to get their work done, since they actually *HAD* to read manuals on using that tool. Nowadays they are just amusing toys that actually get in the way more than help.

The hassles you speak of are imaginary. If people knew their tools, no matter what those tools may be, there is no hassle. The problem these days is you have people using tools that they certainly do not understand, and there is no incentive for them to do so.

From the thank-you-capt-obvious department....

[stiggystiggy]

How is this news? Netcraft said this way back in their June Report:

"Linux is the second most commonly used operating system. Linux has been consistently gaining share since this survey started, but interestingly not significantly to Windows detriment. Operating systems which have lost share have been Solaris and other proprietary operating systems, and to a small degree BSD."

Also, why didn't the article spell out the loser? It's Sun, Sun, Sun, Sun, SUN! Sun is getting clobbered on all fronts. Their hardware is nice, but not so nice that is it needed for 90% of the applications on which it is used.

Solaris, their operating system, has few advantages over Linux, nowadays. Frankly, without adding the GNU tools, Solaris is virtually unusable! (And, who's gonna pay $10k for their compiler when GCC does the job?)

Sun is about to hit a brick wall. Unless they change direction dramatically, Linux is going to gobble them up, just as SGI consumed Cray. Cray was meaningful for a long time, until the capabilities of "Minis" (as Supercomputer folk like to refer to UNIX machines) silently approched the power of super computers at a fraction of the cost.

The same is happening with Linux-Sun. For a small fraction of the cost, Linux on commodity hardware (Intel) is approching the power of Sun's products. It's inevitable, without some sigificant change.

Re:From the thank-you-capt-obvious department....

[nion]

For a small fraction of the cost, Linux on commodity hardware (Intel) is approching the power of Sun's products. It's inevitable, without some sigificant change.

I disagree, and it's not because I work at Sun. Commodity hardware is not nearly on a par regarding uptime and reliability as Commercial hardware. People don't buy Sun because it's cheap. People buy Sun because it WORKS. If you think I'm biased, replace Sun with IBM or SGI or Compaq or any other corporate entity that builds server hardware. You don't base your $$$ infrastructure on a $2k LinTel machine.

Sure, you can build a rather good system with commodity hardware. The PHB's MAY allow the techies to install Linux around the network. But when it comes to making a mission-critical application, they're not going to allow them to run down to PC Joe's, pick up a $2k box, install a $30 OS and believe it will run 24/7 without failure.

Re:From the thank-you-capt-obvious department....

[Ami Ganguli]

Ok, it's definately true that you don't buy a cheapo clone and use it for a mission critical server. But on real hardware (high-end Intel, RS/6000) Linux is every bit as reliable as commercial Unix. The only thing that's missing is "hot-swap anything" features that are only available on really high-end hardware.

It's true that Linux can't go up against Sun in every market yet, but I think the original poster is correct in saying that Sun needs to do something before they lose their edge. McNealy said recently that Linux was no threat since anything new developed for Linux could be incorporated into Solaris, but that's stupid. If you're selling a higher-price product you can't compete by matching the lower priced product, you have to be better.

Re:From the thank-you-capt-obvious department....

[Arethan]

Small applications are the only areas that Linux will ever beat Solaris on Sun hardware. I'd like to see a Linux cluster beat SunFire15k or even an older E10k in a performance vs. total cost of ownership chart.

Let's see, so we hire 10 people at 80k a piece to manage our 500 machine cluster, which will need to be replaced every 4 years at a minimum, just so that we can utilize a free OS?

Or, do we shell out some bucks up front, and get fault tolerant hardware, running an OS in a 5th generation VM environment, that will only require 2 people to manage, and will not need to be replaced for at least 10-15 years. (Upgrades not being considered replacements)

I'll stick with Sun, thanks. I'd much rather deal with a single machine, using extremely fault-tolerant tech than having to deal with 500 commodity pc's that are going to go through the usual 4 year replacement cycle.

Linux and Sun both have their place. Linux is a nice server, and a moderate desktop OS for the tech-savvy (at least I use it as a desktop). It's good for ftp servers, web servers, even small to mid database servers. Sun, on the other hand, is great for extremely high availability situations, where the 0.001% of down time in a 99.999% uptime plan could cost the company a few million in revenue.

Linux is saturating the low end market. Good! The low end market could use some low-cost & stable server software that runs on inexpensive hardware. But Sun caters more to the high end market where uptime is critical and data-sets are unbelievably large.

And no, Intel is no where near doing what Sun can already do. Go shoot your precious linux server with a .44, and see if it's still up. I'd guess the answer is no. Doing the same with a properly configured SunFire 15k would result in a high replacement cost, but an up and running system nonetheless. Processing power? Single cpu vs single cpu is getting closer. But for fault tolerance, full hotswap upgrades (as in lets pop a few more cpus into this machine...while it's running), high end SMP (way more than 8-way, try 72-way), and high end memory size (as in my server as 200GB of ram, what do you have under the hood?), Intel isn't even close. Sun, OTOH, has been doing it for a few years now.

So...how was linux going to kill Sun again?

Fight the right battle

[TommyAquinas]

Because of its robustness, modularity and stability, Linux is highly able to replace Solaris, HP-UX and AIX type licensed OS's in the enterprise. The people who buy these systems buy them to get the best technical solution to their problems and consider cost of ownership, which is high in any OS choice given the task, secondarily.

Trying to get Linux to beat Windows on the desktop is fighting yesterday's battle. Want to kill Microsoft? Sap it's growth, which is in server OS's and embedded systems (XBox, Pocket PC, etc.)

The amount of energy spent by the development community in trying to be the next Microsoft is astounding, but very few vocal developers seem to even focus on what Microsoft is trying to become.

To borrow a phrase from the Old West, "Cut 'em off at the pass" and focus on making an OS that runs devices better than Windows ever will, an OS that runs DB2 and Oracle better than any other and an OS that can be extended and integrated with server side applications at compile time with more ease.

If you take away Microsoft's revenue growth, you take away their stock price. Take away their stock price and you take away their monopoly.

Re:however

[spudnic]

This is the point. They went shopping for a more cost effective:stable solution and they came up with Linux.

They could have possibly opted for Windows. So we can say that we are stealing potential sales from Microsoft and slowing it's widespread acceptance as a server OS.

But is that the whole story? Would they have even had to make a decesion like this if there hadn't been a $free alternative? Could the switch to Linux be argued if it cost the same as Solaris? What if Linux and Solaris where expensive, but Windows was free? What would the decision have been then?

Well, it doesn't matter because Linux is $free, Windows isn't, and they obviously had enough trust in it to move many systems over to it before the Christmas rush. That's really saying something.

Winformant?

[rw2]

I think I trust winformant to tell me about Linux about as much as I trust slashdot to tell me about Windows... :-)

Much as I like Linux,

[einhverfr]

I have to admit-- Linux for community documentation, support, and features.

But-- FreeBSD is STABLE (check longest uptimes at Netcraft when you get a chance). If I could go for 4 years without rebooting with Linux... They have even dethroned Irix when it comes to stability.

So yes, they are a very practical alternative to Linux. It is really that Solaris and HP-UX are not so practical or cost effective in the small ISP market.

I actually now believe that Linux will form a shield which will allow BSD to grow into certain niche markets, such as high-availability web servers (currently MS and Sun).

But what would the Windows growth curve have been

[EisPick]

... if Linux didn't exist? I think these analyses ignore the loss of momentum that Linux has caused Windows.

Five years ago, as NT was replacing Netware in most enterprises, many predicted that Unix systems would be the next to fall under the Windows steamroller. However, in cases where simplicity and the availability of commodity hardware are more important than raw performance and scalability, people are turning to Linux to replace Unix systems, not Windows.

So while Linux may not have made major inroads in replacing existing Windows servers, it has prevented Microsoft's hegemony on the desktop to spread to the server side, and has given Unix (generically) a new lease on life.

I think that's a pretty major story.

Very very wrong: here's why

[devphil]

Not only do I admin and program on Solaris boxes, I'm also a GCC library maintainer. There're my qualifications.

Solaris, their operating system, has few advantages over Linux, nowadays. Frankly, without adding the GNU tools, Solaris is virtually unusable!

"Frankly," you're utterly wrong. Not only is Solaris just fine and dandy, it has features for programmers which aren't anywhere near to showing up on Linux. For example:

• The proc tools, for manipulating a running process. Try using pldd(1) to find out exactly which libraries have been mapped in, or pstack(1) to print the call stack. Yes, that's right, print the function call stack of a running process from the command line, without a debugger, and it works flawlessly. There're about a dozen of these tools, man -s 1 proc for more.
• Kernel watchpoints on memory. Tell the kernel you want to know when a process tries to read or write from an arbitrary block of memory. That process when then be frozen (or killed, as per your instructions), allowing you to find out exactly what piece of code is trashing memory.
• A boatload of malloc(3) implementations. Want a version of malloc that uses brk? It's there. Want it to use mmap instead? It's there. Want a version of malloc that never reuses memory? Or a malloc which sets kernel watchpoints on its own bookkeeping structures, for when your code is hosing the malloc internal memory? It's all there.

Linux has none of these.

(And, who's gonna pay $10k for their compiler when GCC does the job?)

Severely uninformed statement, my friend. GCC doesn't generate SPARC code nearly as well as Sun's compiler. (Ask the GCC developers.) It's good but it's not there yet.

GCC cannot even generate a 64-bit binary yet. (Very close, but still some bugs.)

There are plenty of reasons to buy a SPARC, and to use Solaris, and to use Sun's software. It's all about the right tool for the right job, and Linux quite often isn't it. (I write this sitting on a Linux box.) Quit'cher karma whoring. :-)