Hints at market share for Windows and Linux.
In the wake of the attack of the Zotob worm, I saw a chart from Netcraft that tracked the uptime of Fortune 100 web servers over the previous 24 hours. Apparently there was no effect from Zotob among the Fortune 100, as the server that had the most downtime was running Sun Solaris, and that might have been taken down for maintenance. Zotob seemed to cause more trouble on the intranets of several media companies, as CNN reported on Zotob as if it were some cyber-Armageddon. As it turned out, CNN's LAN just got hit especially hard.
Looking over the Netcraft chart is enlightening. In addition to uptimes and other network-related statistics, the operating system of each site is also listed. The dominant OS in the Fortune 100 is not Windows, but... Solaris. Solaris runs on 42 sites, Windows Server runs on 25 of the sites, Linux on 17, another flavor of Unix runs 5 sites, and 10 sites have an "unknown" operating system. My guess is that "unknown" probably doesn't represent Windows, but some flavor of Linux. What this tells us is that Windows' representation on web servers among larger companies isn't much higher than its share of all web servers-- about 21 percent.
This stands in contrast to figures cited by Laura DiDio of the Yankee Group, who says that servers running Windows account for 65 to 70 percent of the total server market, while servers running Linux account for 15 to 20 percent of the total server market. She cites the conventional wisdom that everybody seems to believe about Linux: that its adoption takes place at the expense of proprietary Unix, not Windows. She also says:
...many organizations are installing Linux as an OS that is complementary to existing Windows servers. Nearly two-thirds of Windows environments now have Linux or some other open-source distribution present in their environments. This trend will continue ...contrary to what the headlines would have us believe, the biggest threat to Microsoft's continued dominance, at present, is not Linux. It is older versions of Windows.
DiDio is reviled among people in the open source software community for being blatantly pro-Microsoft. Her involvement in the intellectual property case brought by SCO against IBM has permanently cast her as a FUDster (promoting Fear, Uncertainty, and Doubt) in the eyes of open source advocates. I think the figures she quotes about the share of Windows servers might be a little high, and the Linux figures are a little low. I think Windows probably doesn't have any more than about 60 percent of the server market by OS, and Linux is probably now closer to the 20 percent mark, maybe a little higher. There are a lot of factors that make it hard to track server operating system share: the fact that some servers might have another operating system installed on them (namely, Linux) after delivery, if they have an OS installed at all, is one factor that should raise caution about server sales by OS. Sometimes, doing a telephone survey with companies about what their IT infrastructure is really like is dicey, too. Her comment that "If you do not know what is on your network... then you cannot truly evaluate whether Linux, Windows or Unix is right for your business" could imply some frustration in her attempts to "get the facts" in Microsoft's favor.
Having stated this caveat, if we take the premises mentioned by DiDio as correct, then what follows are some implications about the future of Windows and Linux.
- Servers running Windows account for [60] percent of the total server market, while servers running Linux account for [20+] percent of the total server market. (my estimates in brackets)
How can there be so many damn Windows servers when Windows servers only account for 21 percent of web server market share? Well, obviously, web servers aren't all the servers that are used in the world of computing. There are database servers, mail servers, file and print servers, and application servers in use. Windows has a major presence in all of these segments, and it could be that Linux has failed to displace Windows servers to any great extent in these domains. The main reasons for this, I believe, is the widespread use of Exchange as an email/groupware server, and the widespread use of SQL Server for departmental and sometimes enterprise databases. The absence of a serious groupware server is a major obstacle for wider adoption of Linux. The major database used on open source platforms, MySQL, lacks the ability to write stored procedures and triggers. (This will change when MySQL 5.0 is in final release.) In these cases, Windows server software is ahead of any similar product from the open source world. However, Postfix is recognized as a reliable and scalable email server, and Oracle and Postgresql both run on Linux and are viable alternatives to SQL Server.
- Linux adoption takes place at the expense of proprietary Unix, not Windows.
- Many organizations are installing Linux as an OS that is complementary to existing Windows servers.
- The biggest threat to Microsoft's continued dominance, at present, is not Linux. It is older versions of Windows.
Alright, if you look at these three sentences together, you notice that they are premises that are built one upon another. Since Linux adoption takes place at the expense of proprietary Unix, it follows that in enterprises that have a heterogenous operating environment, Linux is replacing existing Unix servers, and works just fine with Windows servers. Therefore, Linux is not the big threat to Microsoft, it's older versions of Windows that companies opt to run instead of upgrading to the latest and greatest. This raises the question of why organizations might be sticking with older versions of Windows on the server. I don't know of anybody who would stake their enterprise on Windows NT Server now, since it is no longer officially supported by Microsoft, and the last service pack was released five and a half years ago. Many of those who have upgraded have opted for Windows 2000 Server over Windows Server 2003. And, it's true that Windows 2000 Server would serve the needs of an organization just as well as Server 2003, unless you need to do something that only Server 2003 will support.
In general, it is in the nature of IT, like a lot of lines of business, to replace an older product with something that is similar to what you had before when it comes time to upgrade or replace software. It would be better understood and easier to support. I think the question that is unanswered, and not asked by DiDio in this analysis, is why organizations aren't moving to Windows when it comes time to replace the older Unix servers. Isn't there the possibility that organizations could also choose to migrate their servers to Linux from Windows, and that this accounts for some of the growth in Linux? Or that, as they buy additional servers, organizations are choosing Linux more and more? If you recall my analysis of the web server space (August 3), Microsoft IIS's share of total web servers has declined over the past three years as the Apache web server has grown dominant, even if the actual number of servers running Windows and IIS has grown substantially. The notion that only a quarter of the Fortune 100 are running IIS can't be much comfort to Microsoft either. So I question the notion that nobody is migrating servers from Windows to Linux; it is a logical choice for file servers and web servers. I sense some double talk here on DiDio's part to cover up the truth.
What does all of this mean? 1)I think the implications for me and for other developers are: if you want to keep working with Windows using Visual Studio and SQL Server, fine. If you want to get away from being dependent on Windows and Microsoft servers, work the Web. You will have your choice of tools, operating system, and databases. 2) As open source server products become better, there will be a greater opportunity for Linux to take market share away from Windows. The only missing pieces appear to be a MySQL that can handle stored procedures and triggers, and a groupware product that can replace Exchange for both email and calendar services. The future belongs to Linux.
Comments