Slashdot Mirror
Gartner Claims Less Linux Than IDC
Quite a few readers submitted news about a recent Gartner study which implies a far-lower-than-usually-supposed percentage of servers running GNU/Linux. Reader mfarver, for instance, writes: "Only 8.6 percent of servers shipped in 3rd quarter 2000 were running Linux, claims a recent Gartner Dataquest report. A previous study published by IDC estimated linux held about 24% of the server market share. Unsurprisingly the Gartner study was partially commissioned by Microsoft." Roblimo has penned an interesting piece up at NewsForge about why those numbers might smell a little funny. Hint: how many machines have you bought running any Free operating system from the get-go?
What's the point? The point is that this community was already a vibrant, rapidly expanding one well before anyone decided it was the Next Big Thing. Long before you could read about it in the New York Times. Long before Slashdot became what is is now. Long, long before would-be nerds pondered the latest market figures for Linux-based systems from the Gartner Group and IDG.
Let the PHBs be. The community will survive and prosper with or without them. If everyone who only uses Linux or Free Software altogether because of advertising, press releases, support of large corporations, or mainstream media attention stopped using it right now, what would we lose? Well, a few fine folks would lose their jobs. At least one or two more "Linux vendors" would probably close up shop - but that's likely to happen anyway as part of a much broader downturn. Our market, if we cared to measure it, might shrink by 30% or 40% or even 70%. The software would not vanish from FTP sites - at least, not from all of them. The sources would not go away or suddenly become closed. And if I had to wager, I'd suggest many of the people no longer paid to hack this stuff would still spend time on it. In short, things would - and will - go on as they always have. The bubble was just that. You'll see it years from now as a spike in the plot of market share with respect to time. But don't worry the spikes or the valleys, watch the long-term trend. That trend has been toward a very high-quality set of products enjoyed by more people for a long time.
Forget Gartner, forget IDG, and forget the silly people who think they matter. Step back and look at what you're running today, and enjoy the fact that, all the negative press notwithstanding, you've got Free licenses that never expire for some damn fine software. Relax, and go outside for a breath of fresh air, lest the flames ruin your good mood.
It is true that the usefulness of Linux diminishes at the very high end. But in the realm we're talking about, nobody even thinks about NT. In fact, NT's usefulness is minimal (even by comparison with some NT "base") in midrange servers and nonexistent at the high end. The hardware NT (and x86 Linux) support tops out in the lower midrange of servers. A maximal 8-CPU peecee is a minnow in a school of sharks. At least Linux gives you an option to move beyond the limits of the peecee - an option NT once offered but no longer does.
No, toward the high end you see things like Irix and AIX and MVS and OS/400. Definitely not NT. But that doesn't mean the numbers aren't meaningless; they most certainly are.
I don't know about you, but I would find it very difficult if not impossible to pick up meaningful experience in a world in which there is little relation between action and consequence. Two people can do the same thing on two identical Microsoft systems and get completely different results. More interestingly, you can do the same thing twice on the same box and get different results. Most maddening of all, you can do something, then reverse it, and end up in a different spot than the one at which you started. This kind of environment is not conducive to the type of learning that really good admins go through. There's actually a nicely similar situation discussed in Carl Sagan's Cosmos - he argues that if there were no rules of physics, science would be impossible because we could never learn by experimentation, and no experiment could ever be duplicated.
This entire idea, in fact, also skirts very closely the Monkey Rule. Specifically, a monkey can never be a good systems administrator. Sure, he can handle the common cases and push the right buttons if he's well enough trained, but the moment something out of the ordinary happens it's all over. At the risk of sounding disrespectful, I'd suggest that your 16-year-old sister most likely suffers from the exact same problem. So, no, I wouldn't likely pay her my salary for doing my job, and I'd hope no manager would either. It's not just that you need more admins to manage a dodgy infrastructure. It's back to Brooks: 5 people are not always better than one. Especially if those 5 people don't or can't contribute anything to the solution. Are there managers who've never read Brooks? Surely. They don't last long.
I'm a leaf on the wind. Watch how I soar.
Three points.
As has already been stated, Gartner asked end-users what they installed on their computers after they bought them. Not what was pre-installed on their computer. Implication: either Linux doesn't have the marketshare zealots want to belive, or accurately assessing server marketshare is difficult. You decide.
Garter is a research agency that has its value locked in its reputation. If it produces poor quality research on a regular basis, people will stop subscribing or purchasing its reports. Short term gains (ie fabricating truths in reports for specific companies) will lead to long term losses (ie loss of reputation leading to loss of credibility and revenue). Are you arrogant enough to believe they haven't thought of this? The reality is, they analysed the information available to them and made the best, least biased forecast as they could.
Garter is not in the business of assessing obscure technical facts. They provide a strategic business perspective on technology. Tech-heads are not their market. They don't care about the operational aspects of the technology. The people who run business (which, most commonly, are not tech-heads and have different skills) are their market. They care about the strategic implications of the technology and longer-term market trends.
You think you could do a better job and produce higher quality research? Don't whinge - go out and do it. If you produce research of such a high calibre, you'll drive Gartner out of business in months. I'm constantly amazed at how many people on Slashdot are willing to spout off on things they know nothing about. I don't tell people how to fly a plane - I realise I don't know how. But, the village idiots on slashdot are always willing to provide legal advice, assume everyone in management is a PHB, that companies never ever know what they're doing, that everything is part of a conspiracy and that anyone who doesn't know how to write a sound compression script using bash is an idiot. On the other hand, some people here have some seriously big clues. You my friend, are not one.
In summary, read the bloody report and get some perspective before spouting off.
This is your life, and it's ending one minute at a time.
At least according to Gartner:
Respondents were screened to ensure they were knowledgeable about server purchases over the quarter, and they were asked what percentage of their server purchases were Linux servers, he said.
"We went to end users, rather than looking at just sales numbers, and asked them what servers they had bought over the past three months and what operating system they had installed on it over the same period," Hewitt said. "There was no question about whether Linux was preinstalled or not, we simply asked about new shipments and this is what we found."
Just Microsoft is involved does not necessarily mean the numbers are bad. It is hard to say.
That's because only extreme viewpoints gain karma. I usually post on both sides of the fence for anything MS/Linux, and then watch them *both* get modded up.
Got Rhinos?
I've administered both NT and Unix platforms (including the pricey proprietary ones). Unix wins on TCO, in my expereience, and it's no contest if it's x86 hardware unix.
Why? Business requires a certain minimum level of functionality from their servers, some level such that the temptation to go back to paper and pencils isn't a factor, which varies from business to business and from usage to usage. In order to reliably meet a given level of functionality (performance or stability), I have found that pretty much without fail NT required more machines and more powerful machines (== more expensive machines) to meet that limit than Unix did. More machines means more admins. More machines and more admins mean more cost.
This is not to say that NT goes down every five minutes. But in order to keep one NT server up even approximately as long as a Unix machine, I find that I must restrict it to doing just one service at a time (i.e. just mail, or just file serving, or just web, or just DB, etc.); whereas on a Unix machine I can frequently roll several services onto one machine with no significant drop in performance or reliability.
I am not an Open Source zealot. I am a pragmatist, and I only evaluate the tools I use based on how well they meet my or my client's needs. Microsoft simply does not provide good tools to run mission-critical services on, however "cost effective" they may be.
--
News for geeks in Austin: www.geekaustin.org
News for Geeks in Austin, TX
Not that this fact is particularly relevant, because perhaps the hosting locations that use Windows don't make the top 50 uptime slots. Seriously, though, what I'm pointing out is that there are a number of ways to skin this cat: IDG and Gartner have two different assessments, and I have a third. It wouldn't surprise me if we are all correct.
"The best we can hope for concerning the people at large is that they be properly armed." - Alexander Hamilton
How many UNIX system admins are there in the world? What is their pay range?
How many Windows system admins are there in the world? What is their pay range?
Lets face it, my 16 year old sister can learn to be a windows admin from one of those tech schools that'll make you certified in 3 months. And you can pay 5 of those newbie admins for the price of one unix admin.
Its not that I enjoy these numbers. Honestly, I think if you want to run a windows server, you'll need the extra admins to take care of the additional crashes and glitches, so it isn't cost efficient. But John Q. Manager doesn't know that. Such is life.
Good quote, too many chars. Seriously, the slashdot 120 char limit sucks!
"While I accept my results may not include some desktop and workstations configured as Linux servers, I simply do not believe that Linux is shipping on 25 percent or more of all new servers and I just cannot believe the IDC figures," Hewitt said. "They are simply pushing the envelope and overstating what the operating system is actually doing."
If the majority of customers got their Linux via download or some other way, "then the market is in even worse shape than my survey shows," he said. "How many support contracts are vendors going to get from those customers? I have already told Red Hat that they are stretching the numbers they put out to the marketplace, which resulted in a very lively debate."
These paragraphs seems to imply that he was looking for what OS the machines were shipping with. And accepting that if the 25% of all servers are running Linux (i.e. the IDC study is correct), then those customers aren't buying support contracts.
The wording of the article is confused. What you quoted makes it sound as if they were looking at what OS servers bought in the past three months are running. What I quoted makes it sound like they wanted to know OS the server shipped with.
Can anyone make sense of this?
Education is a better safeguard of liberty than a standing army.
Edward Everett (1794 - 1865)
Does anybody have numbers for things that really count? Things like how many of the real heavy traffic sites are run on Win2k vs. Linux vs. *BSD vs. HP vs. IBM? For many of the big companies stability counts more than free. Win2k is maybe not doing good here, but does _free_ Linux do good here (vs. e.g. HP or IBM) or are commercial Linux implementations through IBM (and that can be AIX or Linux or OS/390 for the real insane)? It shouldn't concern the community if more servers are running Linux or Windows or Amiga OS but it should more count what people are doing with it. And I would not bitch too much about Win2K, there _are_ big sites that are running on it. And you _can_ make it secure. And you _can_ make it performant. Not necessarily if the admin is a moron, but ...
Are corporations buying more Windows servers than Linux servers? Yes. Are corporations buying more Windows licenses than using free software? Yes. Are corporations buying more MS support contracts than Linux support? Yes. Do the majority of corporations operate efficiently? Hell no. Are corporate decision-makers aware of the benefits of open source? Typically not.
Studies like this do nothing to prove or disprove the value of opensource. What I would like to see are comparisons of similar-sized companies that use either Windows or opensource... how do their server-farms compare? Who's more stable, more secure? Who's budget is lower? How much does each company spend on support, hardware, etc? How about some real side-by-side comparisons of real-life scenarios, rather than a guess of how many servers are out there?
Another thing - all of my servers were bought without Linux (some with no OS, some came with windows). I download distros and keep them on an FTP on my LAN. I install from that with the bootnet image. Even if Gartner asked the purchaser what the machines were intended to run, they would not have known and said windows since we are main an MS shop, except for my systems.
Methods they have used so far:
We should compile all this stuff and put it in a central location so that we can refer to it at a later time. I bet it wouldn't take long for all the inconsistencies in their arguments to fall apart, and it would make for great debate fodder.
--------
Bleah! Heh heh heh... BLEAH BLEAH!!! Ha ha ha ha...
Here are some relevent links:
An Opinion from OsOpinion
The Gartner Report Itself
--Volrath50
I'm not the one you bashed, but I was wondering where do you get this information?
As has already been stated, Gartner asked end-users what they installed on their computers after they bought them.
As one of a Gartner customers, we failed to get hold of the detail information on how they did the survey, they (inofficially) said it's commercial secret. All we know is the market segment and sample size. So, are you sure they really did?
Garter is a research agency that has its value locked in its reputation. If it produces poor quality research on a regular basis, people will stop subscribing or purchasing its reports.
Good point, but in fact customers choose Gartner because they want to get what they want to hear, and Gartner chooses who to list in their report. Once my friend's company complaint Gartner on their unfair comparison to their product. Their response was like "You were lucky we ever listed you" attitude. Gartner offered to come over to evaluate the situation, turn out gave them an opportunity to make business. My friend has then grown up thereafter any never take their study seriously.
Are you arrogant enough to believe they haven't thought of this?
Are you naive enough to believe they haven't thought of doing study in favour of big corps. for long term benefit?
Garter is not in the business of assessing obscure technical facts. They provide a strategic business perspective on technology. Tech-heads are not their market. They don't care about the operational aspects of the technology. The people who run business (which, most commonly, are not tech-heads and have different skills) are their market. They care about the strategic implications of the technology and longer-term market trends.
Hmm, isn't that exactly the major problem here?.....no wonder why they rated Rambus having brilliant future....
But, the village idiots on slashdot are always willing to provide legal advice, assume everyone in management is a PHB, that companies never ever know what they're doing, that everything is part of a conspiracy and that anyone who doesn't know how to write a sound compression script using bash is an idiot.
village idiots on slashdot? Compare to people like you who are so naive to believe that reports from big and rich companies must always be trusted, we are real idiots.
In summary, read the bloody report and get some perspective before spouting off.
I read those bloody reports quite frequently for my job. Have you ever wonder how do they come up with those probability factors that predict future trends on something, and how accurate do they turn out to be? I'd be much grateful if you could talk Gartner's to release the details, formula and source data of their research. (No Sir, it's company's secret!)
In my previous job, my company did a lot of contract work for the State of West Virginia, which uses the "Gartner Group's Tier" standards for all their purchases.
Many things on it were illogical, such as the blatant RAMBUS memory cheerleading, etc, despite benchmarks that proved that PC-133 SDRAM was faster on Pentium 3 systems than RAMBUS (and was cheaper).
Basically Gartner, like any company of it's type, says what it's paid to say. Why do they do this? Because they wanted to get paid. Remember when Ziff-Davis would only talk about Liuux when parroting Microsoft's "Haloween Documents"? That changed overnight when Linux companies formed and started advertising in ZD magazines.
The same will happen with Gartner when a Linux company buys their seal of approval.
As the article stated, this "study" used the flawed model of judging server OS's by what OS shipped with the system, rather than actually surveying what was actually USED on that server sold.
This is like measuring what radio station is #1 in the market by what station the radio was set to when it was sold. Or how many people buy Ashland Gasoline because that was the fuel in the car when it was sold...
Unfortunately, Gartner's flaws are not well known by the people who use them as the HOLY WRIT (usually the non-techical people in corporate and government bureaucracy).
=== The price of freedom is eternal vigilance
Linux/GNU/FREEBSD is doing more to help "reduce/reuse/recycle"
Old age and treachery almost always overcome youth and skill.
No doubt, within minutes, the board will be awash in criticisms of Slashdot's anti-microsoft stance and defences of that stance.
I've noticed a trend with regard to this process. No matter how passionately the stoical marketeers defend Microsoft, nor how predictably Linux' yes-men define the particular news story as the turning point in the eternal battle between the forces of freedom and the forces of evil, the truth will lie somewhere in the middle. What's depressing is that those who take the middle ground are morbidly few. Who would have thought the tech sector would create such starry-eyed romantics (as many online activists seem to be)?
Well, 100% of the machines (1) in my apartment came with no operating system at all. So in this part of the country I guess the dominant OS would be AMIBIOS 95.