Print - New Study Suggest Linux Has Foundational Reliability Problems

At IT Forum in Barcelona on Tuesday, Microsoft announced the results of a study it commissioned which concludes that a foundational design problem in Linux prevents that system from being as reliable as Windows in real-world scenarios. Stung by criticisms of past studies, Microsoft commissioned the highly regarded Security Innovations (SI) for this particular study, which focused on e-commerce Web applications. However, Microsoft and SI maintain that the problems with Linux would no doubt manifest themselves in virtually any scenario. And now, Microsoft is reaching out to Linux makers such as Novell and Red Hat in order to commission future studies comparing Windows and Linux.

"This isn't about 'can' or 'can't,'" Ryan Gavin, the director of platform strategy at Microsoft told me in a briefing yesterday. "There are a million different ways of doing things on Linux, but unfortunately half a million of those are wrong. Customers are starting to hit wall in Linux because of dependency issues. It turns out the componentization model there has some detriments with regards to complexity, manageability, and time to market. Windows has a key foundational advantage over Linux."

What the study discovered was that Linux is essentially a house of cards because of massive dependency problems. In the SI study, sets of experienced Linux and Windows administrators were asked to manage Linux and Windows Server machines, respectively, over a simulated one-year time period. During that time period, the machines--which were running eCommerce Web applications--were upgraded in realistic ways, as if to meet changing needs and requirements. The Linux machines utilized Novell SuSE Linux 8, and were upgraded to SuSE 9, while the Windows Server machines migrated from Windows 2000 to Windows Server 2003. Additionally, new features were added to the eCommerce applications over time, and both systems were upgraded with whatever patches and security fixes were released during that time period.

"The Windows systems were dramatically more reliable," Gavin told me. It took the Linux administrators six times longer to administer solutions when compared to the Windows admins. Additionally, the patch rate on Linux was almost five times higher than that of Windows. During the tested time period, there were 187 patches installed for SuSE, compared to just 39 for Windows. And the Linux patches took twice as long to apply, with 14 critical breakages, where dependency failures caused necessary applications to stop working. Windows suffered no such stoppages.

The issue with Linux is that commercial Linux vendors such as Red Hat and Novell typically only support the file versions they ship in their systems. If an administrator arbitrarily updates a component version in order to gain new functionality, that system won't be supported by the OS maker, effectively placing the customer in the OS business, according to Gavin.

Despite Microsoft's best attempts at ensuring that this study was competently and independently designed, Linux backers will no doubt find exception with it. I'll be examining this study further in my Windows IT Pro UPDATE commentary next week.

Wow, now that would be great for the garden. Ran a webapp on windows server 2000, average uptime was measured in hours. Migrated to linux, here is my current uptime measurement now. ~$ uptime 10:55:34 up 104 days, 21:59, 2 users, load average: 0.24, 0.20, 0.19 Key difference is the OS. Updating is simple since it running on debian we run two commands, apt-get update followed by apt-get upgrade. zero downtime since deployment which was 104 days ago. total updates done to date unknown. but since i don't need to reboot who the hell cares? With windows the mantra is update, reboot, test, update again, reboot and keep going till no critical updates available. Notice a reboot is required to update windows, but not linux. Guess a reboot doesn't qualify as downtime huh? Like i said, great for the garden. Sad that the only fruit this "IT study" would yield would be for its ability to be confused with industrial strength fertilizer. Gee, i though Monsanto had that market locked up tight and had zero competition.

Even if you didn't reboot your debian box, how many times did you have to interrupt your webapp to update it and/or your web server (Apache???) to update them?

Publish the whole study- put somebodys reputation on the line here. As Far as I can tell there is absolutely nothing of value in this except to identify the foundational value of the writer and the "Study" as shaky. Pretty telling that MS spend so much money on studies that all seem to find flaws in any other option other than MS. I consider it just another weakness MS KNOWS it has- "studies" or no. Paul- you've actually written some decent stuff- you should be ashamed of spreading this fertilizer.

Paul, you're fine. Don't worry about the peanut gallery (myself included). What I find interesting is that people here are unwilling to even consider the fact that the study was accurate. That is a possibility you know. Some people do good things for the wrong reasons. So let's assume that there's some possibility that this study is accurate. On a related note, the OS wars are just plain sad. All the existing operating system are just plain bad. We should be demanding great solutions from all operating systems, not just claiming the other one is inferior. They all are.

This is just another FUD campaign started by Microsoft. It's full of lies and inaccuracies. The fact is that Linux's security, reliability and stability is much higher than any Windows. With Windows you find out every day a new security flow came out, and you have to apply tons of patches, with Linux you don't have 1/10 of the Windows security flows. This study is just pure Microsoft FUD.

I see we've traded Apple fanatics, who are afraid to post unanonymously, for Linux fanatics who have no fear of looking fanatic-like in public.

What is interesting to note is the article is about Linux vs. Windows, yet the study is Novel Linux vs. Microsoft. For those of you that don't know Novel only became involved during Suse Linux 9, and one cam make an educated guess that many of the breakages that occured during this transition where the result of that change-over. That being said there is no excuse for any breakages in a server OS. And yes restarting a service and rebooting both count for downtime. While the article may raise some eyebrows in the MS world - and a few I told ya so - the comparison of one Linux distribution vs Windows is misleading. The assumption that all Linux distributions face the same "fundemental flaw" is easy to make but difficult to prove. When a study is done to put Redhat vs. Novel Suse vs. Debian vs. MS Windows then, and only then, can you compare apples to apples.

"What I find interesting is that people here are unwilling to even consider the fact that the study was accurate." It's hard to do that when the study is commissioned by Microsoft. Studies NOT commissioned by Microsoft typically give opposite conclusions.

"What I find interesting is that people here are unwilling to even consider the fact that the study was accurate" The problem is that the findings are quite off practical experience. I have installed SuSE 8.x, 9.x and now 10.0 on several computers. The installation itself is way easier than any Windows system. The configuration can be tricky, but once the system runs it runs rock solid. Our FTP/Internal HTTP Server has an uptime since April 2005 updates automatically and needs only to be restarted after a Kernel update. I.o.w. we practically never touch it. I do not understand how Linux Admins needed 6 times more work to administer a server.

>>"With Windows you find out every day a new security flow came out, and you have to apply tons of patches, with Linux you don't have 1/10 of the Windows security flows" Have to strongly disagree with that. There was a time when this was true, but these days you get two or three days of Windows vulnerability announcements a month. The notion that there still are daily Windows vulnerabilities is now an urban myth. On the other hand, whilst not part of the Limux core granted, you get near daily announcements of vulnerabilities in components and libraries shipped with your Red Hat/Suse distro's such as PHP, OpenSSH, GDK etc, components which often need to be installed in order to do anything useful on a Linux box. The security researchers are relishing finding buffer overflows in these new targets and its been this way for many months.

Microsoft will NEVER publish a 'study' that favours anything but MS products - not while Ballmer lives and breathes. You want to be a former Microsoft employee fast? I don't believe anything Microsoft has to say. How often have they been on the losing side in court battles with the government over unfair practices - pricing and other? I believe that is called lying, no matter how you paint it. Search Google for Steve Ballmer ... interesting behaviors from one who is a legend in his own mind. His social habits are basically nothing better than those of a thug - a very rich thug, at least for the time being. Unfortunately most people don't care about any of this. They just know enough about computers to power them on and double-click on the email or chat thingy. The idea of having anything but MS-Windows whatever on their computers freaks them out. Resistance to change is a very human characteristic. I'm not speaking of home users only here - most business people react in the same way. Its very difficult to get people to try something new, especially since it has probably taken them donkey's years to accumulate what knowledge they do have. Linux represents a very real alternative to MS-Windows - and thus a major threat. THere are Linux packages that provide equivalent functionality as those available for MS-Windows that run faster on less hardware. X-Windows gives one a wide choice of desktop managers - some are MS-Windows work-alikes that can help people feel comfortable right away. The server capabilities that Linux provides with SMP and clustering are robust and cost-effictive. The setup and administration tools are X-Window GUI-based - as is the installation of the operating system. Microsoft realizes the potential that Linux has. That is why they take every opportunity to try to discredit it. Pay people enough and they will report anything you ask.

To steve_dupuis "Pay people enough and they will report anything you ask." Hm hm. Or maybe the report just reaches conclusions you don't like.