Windows, Linux gain in server sales, but what about 'free'?
The 64,000 matchstick question...
Microsoft is moving into a "dominant role" in the server market, says IDC, "even as Linux grows." Which is an intriguing headline to the research outfit's take on the 2002 operating system market, but who's winning?
The data IDC puts forward perhaps speaks of a more even match than "dominant role" suggests. The company notes that Microsoft's share of new server licence shipments grew from 50.5 per cent in 2001 to 55.1 per cent in 2002, while paid for (aha...) Linux server licences took 23.1 per cent of the market. The paid for market itself grew 9.6 per cent.
You can make several observations from this little pile of data. First, we're talking about new licence shipments here, and as companies do not in general replace their server systems at particularly high-velocity (hence NetWare's continued, commendable, but perhaps slightly perplexing existence), the process has a slow, drip drip effect on installed base. By maintaining, even increasing a little its 50 per cent share of new shipments Microsoft may be slowly turning the game its way, but it remains a long way from ruling the server world.
Next we should consider that we're dealing with paid for shipments here. You might reckon that 23.1 per cent of this for Linux is pretty good, when the product is basically free, when skilled Linux techies have a pathological aversion to paying for stuff, and when only a handful of tiny outfits (and, of course, their rather larger friends) has even begun to address the commercial Linux market seriously and competently (sometimes). Put that against a background of customers rightsizing server systems, dropping in commodity Linux boxes where appropriate, and you would expect Intel server systems to outperform overall IT spend, and Intel server operating systems to do well because of that.
So you start with an expectation of growth for Microsoft, even a necessity, given that standstill in a growing market actually equals reverse. These kinds of numbers should therefore make Microsoft's server strategists wonder first, if the company's growth rate is enough, and second, if Linux's paid for growth is too much.
At the moment (since you ask) we at The Register will confine ourselves to ask the questions rather than putting forward an answer, but given that commercial Linux's act improved during the period, that it continues to do so, and that it is doing so particularly in more upscale projects, we think we'd be inclined to worry if we were Microsoft server strategists.
The next question is the one you can't easily answer - how valid is it to assess the server market on the basis of paid for sales only, or alternatively, how important are the untold, and virtually untallyable, numbers of free open source installations? You can see the problem research firms have here, because you can't even track numbers of downloads effectively, never mind figure out what it was the downloaders did with the software after they downloaded it. So if you you've got to track something, paid-for only is a possible candidate.
But it's not entirely satisfactory, even simply comparing unit sales is not entirely satisfactory, because you're neither measuring the whole market nor the actual impact of specific sales. Think, for example, Unix downsizing - large Unix installations will tend to look to Linux here rather than to Microsoft, and the importance of switches in this area will be light years away from whether a small business drops in a Red Hat server or a Windows one. Meanwhile not that many major businesses run the entire server show on Windows yet, so maybe you can still see Microsoft as the challenger in the Unix/Linux continuation of Windows versus Linux, and not necessarily the challenger with the advantage.
As regards the importance of free installations, there are numerous answers, depending, but impact is at least theoretically measurable, if you think of importance as something you notice. Which does seem reasonable. OK, so a techie downloads Red Hat Advanced Server, installs it across a 20,000 client system. Clearly this is important, and given that the techie has to eat, the system has to be supported and therefore other people have to eat as well, even with 'free' there's money going out, and it's measurable either as money or as an absence of spend (i.e. flagging sales upscale not being entirely balanced by paid for growth elsewhere, or Microsoft's, or indeed Red Hat's antipiracy squad wondering how come the business exists, seeing it doesn't buy software). Measuring support contracts via Red Hat, SUSE or IBM Global Services doesn't work entirely work either because only a small proportion of this spend is likely to go on these more traditionally set up approaches to service at the moment.
But it seems pretty clear that free does have an impact in some cases, and that if we're to find out how much (before, that is, we wake up one morning, the impact is all around us and we say, 'ah, that's what was happening, then...') we need to stop being overly concerned about unit sales and start at least sampling eatspend.
But on the paid-for numbers as they stand, we'd call it a draw, with Microsoft at least not being rolled back, and the dynamics still to play out over the next few years. And money? Says IDC's Al Gillen: "Microsoft continues to defy the overall market trends, and has again pulled the market upward from both a unit shipment perspective and from a revenue perspective. Linux was the only other bright spot in 2002, with that operating environment posting both revenue and shipment volume gains."
The company says that "overall operating systems and subsystems market revenues grew by a surprising 4.3 per cent to $18.6 billion, led by a 12.4 per cent increase in revenues aboard the Windows platform." The numbers here are for the whole OS market, rather than just servers, but you can see who's defying market trends - ask yourself, why?
Microsoft's new licensing regime had an obvious effect on the company's revenues, at the very least (if we confine ourselves to putting it politely) inducing companies to spend now, in order to reduce their costs on Microsoft systems later. And clearly here there is going to be an associated effect on unit sales. IDC's numbers, incidentally, refer only to licence shipments rather than actual deployments, and in some cases the Microsoft numbers will include entirely paper transactions, where customers are tallied up as, say, Windows Server 2003 'sales', but are still going to be running Win2k for the next several years.
Clearly, you must allow for the effect of the licensing switch in the 2002 numbers, and consider how much the "market trends" were driven by increased licence shipments and revenues being squeezed out of existing Microsoft customers, in Microsoft's direction. Again, a hard one to count, but it's conceivable it could account for the difference between 50.5 and 55.1 per cent of new licences. So still a draw, at best, we reckon, and check again this time next year to see if a more obvious trend can be identified. ®