Slashdot Mirror
Windows Servers Beat Linux Servers
RobbeR49 writes "Windows Server 2003 was recently compared against Linux and Unix variants in a survey by the Yankee Group, with Windows having a higher annual uptime than Linux. Unix was the big winner, however, beating both Windows and Linux in annual uptime. From the article: 'Red Hat Enterprise Linux, and Linux distributions from "niche" open source vendors, are offline more and longer than either Windows or Unix competitors, the survey said. The reason: the scarcity of Linux and open source documentation.' Yankee Group is claiming no bias in the survey as they were not sponsored by any particular OS vendor."
That means that Red Hat Linux has to have at least 1,461 hours of annual downtime, which is 60 days. (This is so that it would then have no more 5,844 hours of annual uptime, in order to allow 20% more of that to fit into one year at 365.25 days.)
I don't think so.
I hate writers who don't understand math.
I have run both windows servers and linux servers over the last 10 years and my experience is higher uptime with linux servers. Windows machines deal poorly with memory leaking apps and need rebooting for every service pack or required update. I only need to restart specific processes with linux when there is a justified upgrade.
Another article claiming my OS is better than yours, another article with virtually no information, and the information therein is off-the-scale incomprehensible and inconsistent.
Here's a casual observation: the article says, "
" Later in the article, this: " Let's just say a Linux server has 24 hours of downtime a year (higher than the "survey" says). That leaves 364 days of uptime in a year, 365 days in a leap year.Implied in the article then, a Windows 2003 server would have to be "up" approximately 20% more to satisfy the "claim". Now, I am not a calendar "expert", but I'm having a difficult time believing that Windows 2003 server is up an average of 364 * 1.2, or 436.8 days a year. If it is, I'm buying.
Also from the article: "..., But standard Red Hat Enterprise Linux, and Linux distributions from "niche" open source vendors, are offline more and longer than either Windows or Unix competitors, the survey said. The reason: the scarcity of Linux and open source documentation...."
First, this is a survey, it hardly points to data that support this survey, in my book a no-no when trying to prove a point. Secondly, assuming there's truthiness in this, my inference from the previous paragraph is, "Red Hat would be a little easier to set up and use if it had better documentation..."
Is it 20% more uptime? Or is it 20% less downtime? There's a very, very big difference there -- two months of downtime is pretty severe, and if you have that, you have some serious problems. From the reverse perspective, three nines of uptime allows for nearly nine hours of downtime per year. If that downtime is reduced by 20%, that's nice, but not really noticeable for most users.
You can never go home again... but I guess you can shop there.
http://www.businessweek.com/the_thread/techbeat/ar chives/2005/04/the_truth_about_1.htmln ux/story/0,10801,82070,00.html ... "The courts are going to ultimately have to prove this, but based on what I'm seeing ... I think there is a basis that SCO has a credible case," DiDio said. "This is not a nuisance case."
s /videos/didio_video.wvx
http://www.computerworld.com/softwaretopics/os/li
Laura DiDio, an analyst at The Yankee Group in Boston, said she was shown two or three samples of the allegedly copied Linux code, and it appeared to her that the sections were a "copy and paste" match of the SCO Unix code that she was shown in comparison.
DiDio and the other analysts were able to view the code only under a nondisclosure agreement,
Watch the "expert" Laura Didio on video from a credible source:
http://www.microsoft.com/windowsserversystem/fact
Enjoy her!
*lol*
How come I never get any of these "impartial surveys"? I have racks and racks of RHEL Linux servers that I only reboot when:
a. a machine suffers a hardware failure (fairly rare) or
b. there's a kernel update that impacts security
In the case of (b), I apply the updated rpms and reboot which normally results in a downtime of approximately 60 seconds for that server. This might happen a few times a year (single digits).
For our small number of Windows 2003 server boxes, it seems that each "windows update" cycle recommends a restart. We'll call that a once a month reboot when Microsoft gets around to releasing their monthly cleanup. Total server downtime is maybe 2-3 minutes (windows takes a bit longer to reboot on the identical hardware used with our Linux machines).
So while I *could* say that our windows servers are down XYZ percent more than our Linux servers, in terms of actual downtime, both platforms are about the same, with Linux seemingly holding a small edge in my experience.
Cheers,
That would be about 304 days, as 20% of 304 is 60.8 (304+60.8=364.8). The 20% must be taken as 20% of the RedHat uptime, not the Windows.
But yeah, that's way too low for RedHat.
What documentation issue?
There are boatloads of documentation available. Ever hear of The Linux Documentation Project? Plus, most distributions offer lots of very good documentation. Why there was a Slashdot story just two days ago about the excellent Ubuntu documentation. There are no fewer than 600 books available about Red Hat distros available for sale on Amazon. Not to mention that Red Hat Enterprise Linux itself includes lots of lots of documentation and most of it is available on the Web gratis. Plus the hundreds of open source apps that include very good documentation with their package. Have you actually read the documentation and free books available on the Samba website? It's darned good!
Any perceived documentation issue is Laura DiDiot's head.
My blog
According to Netcraft, they have a whopping 4 days since last reboot: http://uptime.netcraft.com/up/graph?site=www.yanke egroup.com/
They also go with the bulletproof reliability of MS IIs
Power tends to corrupt, and absolute power corrupts absolutely.
Your math is wrong. 20% more downtime means 1.2 times as much downtime as the Windows box, not 20% of the year.
So if the Windows box is down for 10 hours per year, the Linux box is down for 12 according to the study.
I can easily clone a production server and walk it through the upgrade process
apt-get dist-upgrade
It is truly awesome. You can test and re-test the entire process every time they release a bug fix for any of the packages you'll be using. (Yeah, you can do it with gentoo, also.)
(troll) Silence. Your sensibilities offend the Slashdot drones. (/troll)
I'm a Windows admin. It's what I know, and the only OS I have significant experience with. At my last job, the server with the most uptime was a RHEL3 box that only got rebooted when the ERP database performed its semi-annual crash ritual. Compare that to the four W2k3 boxes that were down about five or six days a year on average for various OS maintenance issues (in Microsoft's defense, we were *doing* a lot more with the Win servers, the Linux server only had one function)
Linux is a hard OS to administer without training. It's not something you can just dive into, and a lot of admins get it shoved on them because upper management decides on a software package that requires it. The result? Downtime because the admin is unfamiliar with Linux and doesn't know where to find the answers. So in that sense, this report is spot-on.
I do question the validity of the data, though. It seems like they picked a sample set that would yeild the results they wanted. A better survey would be to review servers with similar functions, regardless of whether users have both installed. It's no secret that Windows admins have a harder time with Linux and I agree something needs to be done to help them (us) take the plunge with confidence...but this study isn't going to have any impact on anything and was just a waste of someone's money. If they're looking to throw cash away, they should be throwing it at me, not studies.
120 characters for a sig? That's bloody useless.
This is the postfix gateway, external web server, dns server, etc for our little (1000 employee) company:
root[loki:/]# w
10:57am up 1030 day(s), 21:27, 1 user, load average: 0.05, 0.02, 0.04
This happens to be a Solaris 9 system. It has never crashed. Actually, over the past 5 years we have had 1 software related bug take down one of our solaris systems (multipathing bug in the FC drivers when used with active/passive disk arrays). This is based on an environment with 40+ solaris based servers (running a wide variety of services, this is not a '40 identical servers shop')
The best our windows boxes can manage is 6 months (and that is if we skip a few of the security patches).
I can guarantee that during the past 3 years, every single one of our windows systems (60+ servers) has had an issue that is core OS software related (not counting the security related ones). Kernel memory leaks are the most popular (file server reboot every 115 days or it will freeze up). Security worms are another fun one, but kinda rare today compared to the good old days.
I administer both Linux and Windows servers as well. Windows servers (2003 here, specifically, but the same applies to other versions as well) actually work ok and are probably as stable as Linux as long as you don't change anything meaningful on them. Adding users, changing settings, etc is all ok, but don't you dare install anything on a working Windows server without a full, bootable drive copy or a SAN snapshot. That's where Windows servers lose their reliability in my book.
Blanketly saying "Windows is more/less reliable than Linux!" is flat out wrong (or at the very least, misguided) anyway. What were these machines doing? Were they sitting there just passing packets and not reconfigured once, or are they being constantly tweaked and redeployed? How many people were using them?
Uptime is also usually measured in percentages in the business world. I'm willing to bet the author of this FUD saw "99% uptime for Linux, 99.2% uptime for Windows... That's 20% more!"
I have to reboot Windows2K3 jsut about everytime an update is avaliable from Microsoft. I started using the system only a few months ago.
I have not had a reboot of the Linux system we use here in well over a year, (448 days to be exact) even though I have updrad applications and applied many patches.
--fatboy
Boy, the maths in this post seem to be getting screwed up pretty bad, but I'll put in my 2 cents to see if that sheds any more light on things.
Let's use hours. There are 8760 hours in a typical year. (365 x 24)
Let's say your windows server is down for 30 hours in a particular year. That means it has an uptime of 8730/8760 or 99.66%. Your Linux server has 20% more downtime. That's 36 hours per year. (30 x 1.2) and therefore 99.59% uptime. Is anyone really going to notice a 6 hour per year or 0.07% difference in uptime? (remember, we're not talking specific outages here, just a mathematical statistic - not like "Yeah, if that 6 hours was during our peak time")
Maybe I got that all wrong, but that's how I read the statistic.
-Arthur
Cave ne ante ullas catapultas ambules
With Debian, grab deborphan and debfoster and you can weed out un-needed packages quickly and easily.
"deborphan" compares the dependencies of each package so you can see packages that are installed that nothing else needs. Delete the ones that you don't need.
"debfoster" shows what all the dependencies are for a particular app. For example, Apache can have all kinds of packages it is dependent upon. If you want to get rid of that app, you can also quickly purge all the packages that were installed as dependencies for that app.
Once you've got the machine stripped down to the basics, just check all the files in the non-home/non-data/non-log directories to make sure that they each belong to a package. Or that you know why you put them there.
It runs sweet.
It runs clean.
It runs exactly what you want.
Nothing more/nothing less.
Which makes patching the box soooooooooo much easier. And it means that you have fewer potential security holes because you're running fewer apps.
You probably should not read the DiDio-bashing going on over at Slashdot today, but I do see what I believe is an error in the presentation of the data in the press release http://www.yankeegroup.com/public/news_releases/ne ws_release_detail.jsp?ID=PressReleases/news.server reliabilitysurvey.DiDio.htm.
The specific statement, "with nearly 20% more annual uptime" is I believe factually not supported by your numbers. Do you mean that Windows has 20% LESS DOWNTIME than RHEL?
"on average, individual corporate Linux, Windows and Unix servers experience three to five failures per server per year, resulting in 10.0 to 19.5 hours of annual downtime for each server."
If RHEL had 19.5 hours of downtime, and WIndows had 15 hours of downtime, this would be 20% less downtime. 5 hours less downtime per year is actually real data and would be useful to the press release.
On the other hand, 20% more annual uptime would actually result in RHEL being down nearly 61 DAYS per year assuming Windows is up 100.000%.Note: 60.8333 days = 365 - (365/1.2)
----------- The report may be correct. The press release is most certainly in error.
"Windows 2003 Server, in fact, led the popular Red Hat Enterprise Linux with nearly 20 percent more annual uptime."
My turnips listen for the soft cry of your love
In the case of redhat, you can use the standard mail systems shipped with the OS... Infact, you should never install things manually because then you won't be able to update them using the system package management system.
The same problem can occur with windows, people could be running any one of many mail servers on it, and they won't all be centrally updated.
I have encountered the same problems you describe with multiple systems, a consultant sets up the machine and then leaves, it happens with windows too, but less often, and it's much harder to fix when they've made all kinds of weird registry tweaks, usually the fix is to reinstall, leaving the same problems for someone else in the future.
There really is no excuse for leaving multiple copies of sendmail installed, some from source and some from rpm... But quite often it's necessary to do manual tweaks to any system to make it behave in the way you want... There's also no excuse for not installing your packages through whatever package management system exists, so you can keep track of them and update them more easily.
http://spamdecoy.net - free throwaway anonymous email - avoid spam!