Linux and Unix Uptime vs Microsoft on Netcraft

After reading this article be sure to see my follow-up comment about it.

Could it be that Microsoft, Microsoft shills and/or Microsoft fanboys are “gaming” the uptime table at Netcraft? Unless I am misunderstanding something, frankly, I think they are. I was going to show a friend of mine the uptime table at Netcraft to display the ability of Unix and Unix-like operating systems to be stable and reliable. Yet I get there and I see this:

Netcraft Microsoft Bullcrap

For several years I have watched and used the uptime statistics at Netcraft and for the majority of that time the top ranked systems have been Unix, usually BSD and company, and Linux. I have not been to Netcraft to see the statistics in several months. So I was slightly irritated to go there today and see that either Microsoft, Microsoft shills, Microsoft fanboys or a combination of all the above have obviously skewed the chart. How do I know this? This bit of information off that same page near the bottom tells it all:

How to Game the Statistics

Just get enough dedicated people to request sites over and over that run your favored system and you too can have your own favorable Netcraft uptime chart. I think Netcraft needs to rethink how they generate the longest uptimes chart. It is obviously being abused by people who favor Microsoft to falsely show Microsoft operating systems dominate the uptime statistics. Anyone in the IT industry with half a clue, ethics and a sense of honesty knows and admits Microsoft could not dominate on uptimes. There are too many patches that require a reboot of a Microsoft server for that to be true.

Apparently some Microsoft administrators do not patch their systems. I sure would not want my web site hosted on their unpatched Microsoft servers. Of course, these could be server farms where systems do get patched behind a load balancer, but then we are still talking about false statistics. No small business I know that has a Microsoft server in it can keep that server running for over 1,000 days without a reboot. Not if they want to keep it secure with up to date patches. But I do know of small businesses with Unix servers that easily go for more than a year without a reboot if there is no power outage that outlasts their backup power system(s).

What do we in the Unix and Linux community do about this? Maybe write Netcraft and suggest they change how they handle generating that chart. Otherwise I suggest we do … nothing. We should not start our own gaming war to change the statistics. Why? Wallowing in the crap filled mud with pigs only gets one dirty, and the pigs like it. So, just do as I am doing and state the facts to your friends, acquaintances and business associates. But do not use Netcraft to back the facts up at this point.

This article has had this many unique views:

hit counter
hit counter

Notice: All comments here are approved by a moderator before they will show up. Depending on the time of day this can take several hours. Please be patient and only post comments once. Thank you.


Published by

Gene A.

Gene is a "Unix Guy", network technologist, system trouble-shooter and IT generalist with over 20 years experience in the SOHO and SMB markets. He is familiar with and conversant in eComStation (a.k.a. OS/2), DOS (PC, MS and Free), Unix, Linux and those GUI based systems from Microsoft. Gene is also a follower of Jesus (forgiven, not perfect), and this does inform his world view.

20 thoughts on “Linux and Unix Uptime vs Microsoft on Netcraft”

  1. Hmm, it seems I am late to the party on this one:

    Did Microsoft Buy Netcraft?
    August 24th, 2007 by Nicholas Petreley

    Lies, Damn Lies, and? Netcraft/Microsoft

    I know I have not been to the Netcraft uptime pages in several months. Probably late in 2008 was my last visit. But the last time I visited I know the chart still had a majority of Unix systems on it. I do recall there were more Microsoft systems than I expected at the time. I still suspect manipulation by outsiders over deliberate favoritism by Netcraft at this point.

  2. I found out about this some time ago and can’t remember the exact details but the reason has something to do with the short time span. All the *nix computers are in months and years leaving few to “compete” in the 7 day time spans and the times that get set on “reset”. I’m sorry I just don’t remember.

  3. Well, now that I looked at the “Boycott Novell” link posted here, I probably had old information. Never, ever trust Microsoft.

  4. Just like the garbage PCLinuxOS pumped into DistroWatch’s “open to abuse” rankings. Any site can display stats. It’s up to the viewer to try to guess if the stats have any value.

  5. This chart doesn’t make much sense and probably doesn’t mean what you think it means. Note the huge block of servers with the exact same number, 1281 days. Also, Windows Server 2008 came out ~530 days ago so it’s unlikely to run for 1281 straight days anywhere. These numbers surely have a meaning, but it is not straight days of running without reboot.

  6. I had not checked in a while, but I remember much smaller numbers here for Windows. As someone else pointed out, these can’t be days.

    Also, I remember looking at a Windows site that was near the top some months back. What I got was something like a one sentence webpage. Wow, can the server handle sending out a simple static page on every query?

    These netcraft stats meant something when they applied to random sites not trying to game the system.

    ISO, Netcraft, .. it seems Microsoft does a fine job of gaming formerly respectable organizations. They have been exploiting loopholes forever: Dominating third party vendors on their Windows. Exploiting monopolies. Exploiting “open” standards to spread their patents. Exploiting the Bill and Melinda Gates Foundation to give Bill’s private investments advantages over competitors. Exploiting tax loopholes (including with the B&M G Foundation) to avoid paying their fair share of state and federal taxes and also get billions in subsidies through the Stock Options Income Tax Benefit . Using patent troll companies owned by Myhrvold and Gates to attack/extort others without fear of retaliation. ….

  7. Carme (comment #7) and hip (comment #8). Good point about Windows server 2008. However, most of this chart is explained in the Netcraft FAQ. Specifically weird things like WS 2008 having over 1,000 days of uptime might be due to some of the oddities explained in the FAQ:

    Even so, I think Netcraft needs to do some serious reworking of their statistics gathering if they want to be taken seriously again.

  8. Linus did this, not Microsoft. It’s a side effect of the switch from a 100MHz clock to a 1000MHz clock; with the faster clock, the uptime markers Netcraft uses roll over after 45 days. If you read their FAQ, Netcraft says point-blank that they “cannot give reliable uptime figures for many Linux and FreeBSD servers.”

    No one’s likely to suggest we switch back to 100MHz clocks; the faster clocks make a lot of sense. Netcraft ought to re-design the way they collect statistics; until they do, the report doesn’t show what it claims to show.

  9. Netcraft statistics can be manipulated by any company choosing to ‘target’ a specific test or set of tests.
    When Netcraft started compiling statistics, it was not thinking that any company would use unsavory business practices to skew the numbers. When MS found that many *nix users would point to Netcraft as proof of a more reliable OS, MS targeted Netcraft. First was the GoDaddy fix. All the ‘parked’ domains were switched from Apache to IIs. This changed the stats significantly, all of a sudden there were millions of IIs hosted domains. MS announced the switch to the world before it took place. I believe MS ‘worked’ with GoDaddy to make this switch.
    Now we have these amazing Windows uptimes on Netcraft. We all know an unpatched Windows box connected to the Internet will become a spam zombie. If these numbers are to be believed then these Windows servers are providing the majority of spam we get to enjoy everyday in our inbox.
    MS has a history of dishonesty. Unfortunately, there is no law forcing a business to be honest or to have a soul.

  10. I don’t know if it’s anyone “gaming” the system, but certainly the statistics aren’t accurate in many ways. Linux can’t show uptimes longer than (depending on timer frequency) 450 days or 45 days, it just rolls over to 0 again even though the server may still be up.

    The thing is, the “uptime” count is really more of a hack anyways. It takes advantage of the timer value in the tcp header. This isn’t what it’s supposed to be used for, but it just so happens to also be useful for that. Linus and other kernel hackers don’t seem to care that this value is used by netcraft to show uptimes.

    Flat out, Netcraft is broken, and there is no reliable fix. The reason so many systems have the same uptime count is that netcraft counts per hostname, not per machine. That could be 10 hosts on one machine or 1 host on 10 machines, netcraft sees them as 10 hosts, each unique.

    The same is true, by the way, of netcrafts hostname survey. It counts hosts, not servers. Many people take the numbers to mean 60% of the web servers run Apache, but that’s not true. 60% of the hostnames are hosted on servers that run Apache. That could mean that 80% of them are IIS server, and 20% of them are Apache servers, but with more hostnames on the apache servers.

    So in reality, Netcrafts numbers just don’t mean much and never have. So it makes no sense to get upset about them, and accuse “trolls” or “fanbois” of something.

    By the way, the Server 2008 statistics are probably because they’re behind an ISA Server “Internet accelerator” that hasn’t been rebooted in 3 years. This is a Windows box that is just basically a proxy server. It will still report as Windows 2008 and a Windows OS because the host headers are forwarded on from the main server. This is why you often times see a Windows Server host header with a Linux OS because there’s a Linux proxy server in front of it.

  11. @ Ridgeland (#5)
    Do you actually still believe that old lie? Tex and the game made it very clear on the PCLinuxOS forums that “gaming” distrowatch would not be tolerated. Ladislav said there was no evidence of gaming.

    A couple of years ago, PCLinuxOS happened to be the hottest distro (meaning, more people were looking at its page, than any other single distro’s – nothing else) for a couple of months. Apparently this upset the Ubuntu fan base, because they couldn’t believe that a handful of part-time developers could make something that could rival Ubuntu, so they started a smear campaign.

    These days, the PCLOS community hopes it doesn’t regain the top spot on DW, because the tiny bit of prestige gained is not worth the hate.

  12. Hi All,

    I have looked into this further. The FAQ at Netcraft is quite revealing about their flawed statistics. Specifically see this FAQ. It appears that the real problem may just be that Netcraft is unwilling to change its uptime statistics gathering method to account for faster system timers. They deliberately remove such systems from their statistics. This includes the majority of modern *BSD and Linux kernel based servers.

    Of course the uptime list containing only Microsoft based systems is still quite suspicious. It is unlikely to have only Microsoft systems listed due to the deliberate deletion of modern Unix and Linux systems from the list. I am quite confident there are still older Unix and Linux based servers “out there” with long uptimes. That the uptimes list is generated arbitrarily based on this fact, “For performance reasons, we limit this monitoring process to the most frequently requested sites.” is the real issue here.


Comments are closed.