Gartner on Enterprise Linux 2007

ONE DOESN'T WONDER about where Gartner Group finds the writers to fabricate IT Industry "analysis" papers -- any street must do, given "The state of enterprise Linux" bolus Gartner barfed up recently.

Here's the lead-in: "Through 2007, the Linux market will surpass $9 billion in revenue, approaching 18 percent of total shipped revenues on an initial acquisition basis (0.8 probability)." ( ZDnet.)

First, let me say that I don't have anything personally against the guy who happened to write the article, as I've never met the gentleman, nor do I bear him any ill-will. In fact, I'm glad to see he is putting his English Lit. degree to profitable use writing well-formed sentences that parse. Beyond that, I'm talking about Gartner Group, the IT company.

But IT company is perhaps being too charitable with regard to the advisory side of the business -- the term parasite might be more apt. (Gartner also has an IT consultancy division which is... another IT consulting firm, in my book. No quibble with them, either.) However Gartner Group (the IT Analysts) distinctly reminds one of the Taylorist joke that: "An analyst is someone who borrows your watch in order to tell you what time it is." I mean, here's what they do: they ask you to fill out questionaires, take your managers' and staffs' time in lots of interviews, go away for awhile, and then sell you a fat report that tells you what you and your peers told them. What a scam! That's what I mean by characterizing Gartner and their ilk as IT parasites, lice.

But beyond that -- and I realize a judicious level of this sort of thing can be informative for those involved -- it seems that Gartner Group and similar "IT Analysis" firms have taken their business model too far. So far, in fact, that their "reports" have ceased to be based on statistics or well founded projections and have instead descended into the realm of unfounded speculations and even more suspect opinions. I'll explain.

Three things have bothered me about Gartner Group, for a decade or few: (1) Lack of Retrospection, (2) Probability Estimates, (3) Sloppiness. I'll take each of these, one at a time, and lay out what I have seen.

Lack of Retrospection Gartner never revisits past predictions to assess whether or not they got it right, were inaccurate, or just missed the boat entirely. This is not a characteristic of a successful company that's in the business of prediction (unless it gets away with this).

If you research stock market analysts, many of the decent ones show you how they did over the last year, five years, even ten years. Similarly with economists (if you press them) -- they'll show you their projection for last year versus what actually occurred with respect to an economic metric. Not so with Gartner -- they don't do retrospective analyses.

Gartner never tells you how accurate they were last year or year before or year before that, about anything, so one has no basis to judge their overall effectiveness, even while they're selling the new prophecies.

Probability Estimates Can I reasonably say there's a sure 100% probability that the Sun will rise tomorrow? No, I can't, because there is always some small chance that a solar eclipse will coincide with the exact time of sunrise, or an unseen asteroid might knock us all back to the Jurassic era, or something.... Can I really expect a 0.000% chance that the Sun will rise tomorrow? Well no, not that either, since I've got historical evidence on the order of 10^^10 that it's happened every day and is thus very likely to happen again tomorrow morning. But, can I predict whether the stock market will go up or down this year? Not as easy, is it? Even better, given what I and all my peers know today, can I predict whether IBM will sell more or fewer mainframe MIPS next year? If I and all my peers buy mainframe MIPS, and we all know our IT budget, you'd think that might be almost possible. But there's the much larger US economy, there are other countries, there are taxes, elections, wars and rumors of wars. So how precise are my probability estimates now?

Sure, there's the economists' principle of ceteris parabis, i.e., "all things being equal." And one imagines Gartner leans very heavily on this. But this ignores the crucial chipset that's anticipated but's delayed due to unexpected errata, or OS software that's delayed, or any number of other things. The IT landscape is just too lively -- some would say chaotic -- to give any metric much more than 50% chance of being met within a year, much less within a five-year timeframe.

Thus Gartner's probability estimates are virtually worthless, especially for more than a year out. Businesses plan in 2-3 years, vision five.

Sloppiness Gartner never shows their reasoning. I mean, they do all these "studies" about this and that, then they issue big fat reports that contain sheaves of unsubstantiated predictions, but they never show how they arrived at their prophecies. Their address should be at Delphi in Greece, for all one can follow the reasoning. They must be gods!

In accounting, economics, engineering, and sciences, one must be able to demonstrate how one arrived at any specific conclusion. Not so for the "gods" at Gartner. Even in their detailed subscription-only reports, it is never clear how they arrive at their oracular pronouncements complete with probabilities. One must therefore conclude that they... guess.

SO, ENTERPRISE LINUX 2007? Gartner is just guessing wildly here, because they have to say something (there's that revenue stream to feed, after all). Just a few years ago, they were proclaiming that Microsoft would "move up into the data center" (part right there, guys) and "take marketshare from Unix systems" (part right again), and "displace all the mainframe technology" (oops), plus rafts of other things that either did or didn't more-or-less happen. Linux/Apache etc. wasn't on their screen at all, even as late as two-three years ago. I don't have their archive reports, but I do recall their expectations, and they didn't expect this IT systems landscape we happen to have now. They blew it, big time. So why should we imagine that their IT predictions today are any better?

Let's reread Gartner's conclusion again just to refocus here, shall we? "Through 2007, the Linux market will surpass $9 billion in revenue, approaching 18 percent of total shipped revenues on an initial acquisition basis (0.8 probability)." Now, let's look closer.

One hardly knows where to start. What's the "Linux market" to Gartner? Is it hardware? Hardware plus software? Hardware, software, services? It seems Gartner just doesn't get it! They're so used to categorizing IT systems by vendor/technology that they don't seem to have gotten any arms around the "Linux market" at all. I mean -- they're so used to IBM selling S/390-MVS, Z/Series-Z/OS, AS/400-OS/400, RS/6000-AIX, X/Series, HP selling 9700/HP-UX, Compaq selling VMS, Alpha, Tandem, and Intel, Sun selling Sparc/Solaris, plus all of the various OEM Intel vendors selling X86/Windows -- that they can't fit cross-platform Linux into the grid. From their statement, all one can expect is that people selling systems (hardware/software/services) with a Linux label affixed somewhere might take in $9 billion (80% probability) in 2007, about five years away.

Gartner doesn't even seem to recognize that many systems destined for Linux installs won't ship with an Operating System attached to the box, but nonetheless Linux will be loaded on it. 40% of Intel systems are built by the White Box Company, and they are certainly happy to ship a server (or desktop) without an OS preloaded. How can Gartner know? In regions like Asia, how can they guess they won't be pirated Windows?

In even US companies, how can Gartner predict how many systems (servers or desktops) will be loaded with Linux, replacing obsolete OS software? What colossal arrogance lets them assume they can predict the monetary value (revenues to someone) of OS installs, configuration, middleware, databases, applications, and webservers, etc. built and run on Linux? In what Universe to they think they can predict the total Open Source services revenues five years out? Such blindness boggles the mind!

Aside from these few deep and utterly crippling problems, the article by Gartner about Enterprise Linux in 2007 contributed comforting apple pie and motherhood confirmations of IT 101 verities and some predictable yet almost stretching conceptualizations of how enterprises might use IT in a few years. Perhaps I was too hard on the author: it's possible that he not only stayed awake during his CS classes, but also took notes.

I did like one thing in this article: it projects that Microsoft's OS marketshare will increase all of 1% in the next five years. I am more inclined to foresee it decrease, but no growth seems good enough....

Gartner's superficially reasonable but deeply flawed oraculation titled "The state of enterprise Linux" is available in all its cluelessness at ZDnet here.µ