advertising information
   personal technology

 custom news
 Headline News brief
 daily almanac
 CNN networks
 CNN programs
 on-air transcripts
 news quiz

CNN Websites
 video on demand
 video archive
 audio on demand
 news email services
 free email accounts
 desktop headlines

 message boards




Opinion: Why is Microsoft worrying about Linux?

March 24, 1999
Web posted at: 9:19 a.m. EST (1419 GMT)

by Nicholas Petreley


(IDG) -- The Microsoft FUD (fear, uncertainty, and doubt) campaign against Linux officially began when Microsoft Group Product Manager Ed Muth recently made some unflattering remarks about the OS. Eric Raymond, president of the Open Source Initiative, offered his entertaining view of this move in the exclusive LinuxWorld opinion piece, "Halloween V" (see Related Stories below). To complement Eric's article, I would like to address one or two of Ed Muth's remarks in light of current events and recently released statistics.

Most of what needs to be said can be said in response to the following remark from Muth:

The there are fewer applications available for Linux, there's no long-term development roadmap, and there's a higher technical risk in using it [...] You could cut Linux some slack if it were sharply lower in cost-per-transaction than NT, but that's not the case.

As for applications and viability, what can I say in the defense of Linux that hasn't already been said? To quote one of Linux's most ardent supporters:

Applications support for Linux is also growing rapidly. Within the past few months, leading software vendors have announced that they will create Linux versions of their flagship products. For example, the leading database vendor, Oracle, recently announced that it is developing a Linux version of its market-leading Oracle 8 database. One of Oracle's chief database rivals, Informix, quickly followed suit, and Mr Soyring testified that IBM is porting its principal database product, DB2, to Linux as well. (In fact, IBM announced on November 20, 1998, that its Linux version of DB2 will be free.)

Netscape is developing Linux versions of its enterprise server software, and most of Netscape's client software is already available for Linux. Corel offers a Linux version of its popular WordPerfect suite of business productivity applications, and it is free. Star Division of Germany offers its StarOffice, a full suite of business productivity applications. StarOffice, recently priced at about $300, is now free for individual users. Sun is porting its Java Development Kit 1.2 to Linux.

Commercial software vendors such as Red Hat and Caldera now offer compatible versions of Linux at a nominal charge. Although Linux can be and routinely is downloaded for free off the Internet, Red Hat and Caldera plan to earn revenues from related software and services such as training and support. The presence of such supported versions of Linux promises to make the operating system more appealing to corporate customers.

Today, the number of developers working on improving Linux vastly exceeds the number of Microsoft developers working on Windows NT. Linux developers are currently working on "Windows-like" user interfaces "Gnome" and "KDE" to simplify its operation.

It is unlikely in any other established industry that a single person, aided only by independent volunteers, could create a product that would emerge to challenge the industry leader. Yet this is the story of Linux, and the nature of the software business.

And who is this mystery fan? The above quotes are from the January 27, 1999, transcripts of the testimony of Microsoft executive Paul Maritz in the DOJ vs. Microsoft case.

Microsoft has presented quite a puzzle. Who should we believe regarding application availability and the viability of Linux? Microsoft -- or Microsoft?

Removing the plank

But it's the phrase "there's no long-term development roadmap" that tickles the ear (in fact, it isn't unlike the sound of fingernails scraping along a blackboard, slowly). If anyone, it's the Microsoft customer who has good reason to be concerned about the "long-term development roadmap" for Windows NT.

I saw a most telling aside within a March 17, 1999, InfoWorld article, "Pentium III Xeon gets thumb's up from app vendors," which trumpets the release of the new Intel Pentium III Xeon processor. (See Related Stories, below.) Here's the passage in question:

Dual-processor Pentium III Xeon systems should also help IT managers trying to overcome Windows NT's tendency to spike to 100 percent usage when subjected to numerous simultaneous interrupts, which in turn leads to system crashes, said analysts at the Aberdeen Group, in Boston. Deploying a second processor should help alleviate some of those crashes, they said.

This article takes for granted something many Windows NT administrators have known for some time: Windows NT crashes under heavy loads.

This comes as no surprise to InfoWorld. The InfoWorld Test Center discovered the same thing when it benchmarked a four-processor Windows NT box running Microsoft SQL Server vs a single-processor IBM AS/400 running DB2. While the Windows NT box ran faster than the AS/400 under most conditions, NT crashed every time it was subjected to the heaviest loads. The AS/400 plodded along at about the same pace under those same extreme loads as it did for lighter loads.

Is this what Ed Muth is referring to when he cites "cost-per-transaction?" If so, it would seem very costly indeed that there is a fundamental design flaw in Windows NT that prevents it from handling heavy loads gracefully. The fact that anyone might find it acceptable to attempt to "solve" this problem by adding more CPU power merits outrage from Windows NT customers. Indeed this may be one of the many reasons IT is fleeing from Windows NT to Linux. But most of all, where is Microsoft's "long-term development roadmap" to fix this design error? It's nowhere to be found.

Economies of scale

Neither will you find a Microsoft clustering solution competitive with Beowulf clustering for Linux. While Microsoft struggles to take its proprietary scalable clustering solution for Windows NT beyond failover fault tolerance, IBM recently demonstrated the high scalability of Linux using easily available tools. Earlier this month at the LinuxWorld Conference & Expo, IBM clustered 17 Linux servers to create a $150,000 supercomputer that runs as fast as the $5.5 million Cray. But the coup de grace is the fact that IBM did this with freely available Beowolf clustering technology and a copy of Red Hat Linux purchased the day before the clustering demo.

The Linux portion of the above solution weighs in at $49. You can't even use Windows NT to create the same solution today. If or when Microsoft finally delivers such a solution, does anyone think it will do so for a total of $49?

In fact the only scalability demonstrations conducted by Microsoft were shown under tightly controlled conditions using an unreleased version of Windows NT. One can't help but recall the fact that Microsoft demonstrated the superior performance of Windows with Internet Explorer by creating a videotape simulation that supposedly emulated actual results. And when all Jim Allchin's men couldn't reproduce those results, the explanation he gave was that the original results were obtained under laboratory conditions -- not in the real world. From the InfoWorld article, "Microsoft shamed in courtroom" (see Related sites below):

Microsoft spokesman Mark Murray said the confusion came about after Microsoft edited the original videotape, which was filmed in the company's broadcast studios after the research had been conducted on different computers in laboratories.

  LinuxWorld's home page
  LinuxWorld free e-mail alerts
  LinuxWorld features & columns
  Get Media Grok and The Industry Standard Intelligencer delivered for free
 Reviews & in-depth info at's personal news page
  Questions about computers? Let's editors help you
  Subscribe to's free daily newsletter for IT leaders
  Search in 12 languages
 News Radio
  Computerworld Minute
  Fusion audio primers

Finally, for hard evidence we need look no farther than Microsoft's own e-mail service, Hotmail. Hotmail originally ran on Sun SPARCstations and Solaris. Microsoft quietly attempted to move the entire operation over to Windows NT, but NT couldn't stand up to the task. So Microsoft (even more quietly) moved the operation back to Solaris.

(If you just can't believe Microsoft runs Hotmail on Unix, see the recent Bull Software press release, "Bull awarded major contract from Microsoft for system management software," which states that:

The initial order is for OpenMaster to manage the hundreds of servers and network connections of Hotmail, the world's leading e-mail service, which currently provides worldwide e-mail services to more than 35 million users.

OpenMaster runs on and manages AIX and Solaris, not Windows NT.)


There's another thing Windows NT customers should consider regarding Microsoft's long-term plan for NT. It now seems as if the road ahead for Windows will be forked into at least four different routes; Windows CE, Windows 9x, Windows 2000, and 64-bit Windows. Microsoft splintered Windows considerably by releasing Windows CE as a separate operating system. Now it has acknowledged that it can't really turn Windows NT into a consumer operating system anytime soon, so it has decided not to discontinue Windows 98 development, thus splitting the future market even more.

On top of that, I've been informed by a reliable source (who has asked to remain anonymous) that there are internal reasons why Microsoft will have no choice but to fragment the development of Windows NT itself. It appears that the task of making Windows NT a 64-bit operating system is much more difficult than Microsoft had imagined.

Microsoft has even enlisted outside help from Unix programmers for the project -- programmers who are intimately familiar with the problems of moving from 32-bits to 64-bits. Add to the evidence the fact that famed OS developer Dave Cutler is heading up the 64-bit version of Windows NT, which is being developed separately from Windows 2000.

Between the outside influence from Unix programmers, Dave Cutler's reputation for being a renegade team leader, and the fact that 64-bit Windows is considered a separate project from Windows 2000, and it's looking more and more like Windows NT will be splintered into two product lines -- Windows 2000 and 64-bit Windows. If such splintering occurs -- and it seems almost certain at this point -- then the road for Win32 leads to a brick wall.

Ed Muth promised that 64-bit Windows NT will have a 32-bit subsystem to maintain compatibility. One only has to recall the 16-bit subsystem in Windows NT 3.1 to get an idea of how this will play out. Rather than build 16-bit compatibility into the kernel, Microsoft enlisted Insignia Solutions to write a Windows 3.1 emulator, which Microsoft then licensed for use with Windows NT. As a result, Windows 3.1 programs ran slower than molasses on Windows NT 3.1 -- when they ran at all. (This isn't, by the way, meant to be a complaint against Insignia Solutions products like SoftWindows. These products work as well as can be expected in their environments. The problem is that Microsoft considers the technology a replacement for a viable compatibility upgrade path in Microsoft operating systems.)

If Microsoft does make good on its promise to deliver 32-bit compatibility through a subsystem, it's likely that there will be no real upgrade path from Windows 2000 to 64-bit Windows NT. As Microsoft abandoned its 16-bit line of software a moment after Windows 95 was pushed out the door, expect Microsoft to abandon 32-bit Windows the moment 64-bit Windows NT ships.

Not so difficult as it seems

Not everything Ed Muth said was FUD, however. Muth made an entirely reasonable comment about Linux as open source: "It's a difficult business model to make work. It's one of those things that the marketplace will decide."

Indeed it looks as if the marketplace has already decided. Consider this Netcraft survey of 4,389,131 Web sites:

Developer February 1999 Percent March 1999 Percent Change
Apache 2350748 54.65 2409056 54.89 +0.24%
Microsoft 1014419 23.58 1036472 23.61 +0.03%
Netscape 264469 6.15 288511 6.57 +0.42%
O'Reilly 74354 1.73 83950 1.91 +0.18%

It appears as if every Web server is growing quickly except Microsoft Internet Information Server (IIS). But can we interpret this as slow growth for Windows NT? Not necessarily. Consider the following OS survey results from the Internet Operating System Counter:

Internet Operating System Count for January 1999

Ranking OS 9/98 count % of total 1/99 count % of total +/-%
1. Linux 223441 27.6% 287093 28.5% +0.9%
2. Windows 95/98/NT 190726 23.6% 245184 24.4% +0.8%
3. Solaris/SunOS 109579 13.5% 178350 17.7% +4.2%
4. BSD Family 170228 21.0% 150961 15.0% -6.0%
5. IRIX 43987 5.4% 52941 5.3% -6.0%
6. AIX 15132 1.9% 17362 1.7% -0.2%
7. Mac/Apple 15929 2.0% 16170 1.6% -0.4%
8. Reliant Unix/Sinix 4242 0.5% 12848 1.3% +0.8%
9. HPUX 13530 1.7% 10416 1.0% -0.7%
10. Digital UNIX 4322 0.6% 5016 0.5% -0.1%
11. SCO UNIX 3268 0.4% 3787 0.4% 0.0%
12. Novell Netware 2726 0.3% 3656 0.4% +0.1%

The above table makes it rather obvious that when Microsoft talks sales numbers regarding NT vs. Unix, it is counting only commercial versions of Unix. Despite a slowdown in growth of commercial Unix, fully 71.4% of the Internet still runs on Unix. 28.5% of that is represented by Linux, which accounts for a bigger slice of the Internet than any other operating system (including the combined total of Windows 95, Windows 98, and Windows NT).

Despite this, how does one account for the reasonably good growth rate of Windows NT? One can only speculate, but keep in mind that -- as we discovered regarding scalability -- even when NT is up to the task, it often takes several Windows NT servers to match the performance of one Unix server.

Taking a belt to NT

But the most compelling evidence comes from a survey of Windows NT customers conducted by Sunbelt Software, which claims to be "The world's first and largest distributor of Windows NT system management utilities." The survey turned up some very interesting results regarding Linux. (See Sunbelt March99 Survey Results over 1999 NT users for the complete survey results in Related Sites, below. Note that the number 1999 refers to the number of respondents, not the year of the survey).

In the next 12 months, how many Linux Servers are you planning to add?  
1-5 842
None 500
6-10 191
100+ 161
20-50 111
11-20 102
51-100 52

In the next 12 months, how many NT Servers are you planning to add?  
None 1037
1-5 625
6-10 152
11-20 65
20-50 53
100+ 24
51-100 18

Clearly even among Windows NT customers (especially among NT customers?), Linux seems to have far more momentum than NT. But the most fascinating and perhaps revealing thing about this survey is that one cannot easily explain away the results as stemming from a "lack of fairness in media coverage," regarding NT and Linux. The vast majority of those who took the survey have direct experience with both Windows NT and Linux, as one can see from the following question:

Are you currently running one or more Linux Servers?  
Yes 1337
No 405

After having used both Windows NT and Linux, which OS do these Windows NT users plan on investing in for the future? Linux wins by a wide margin, indicating substantial growth for Linux and relative stagnation for Windows NT.

Granted, open source software involves new and unusual business models. But as Ed Muth has stated, it is ultimately the marketplace that will pick the winners and losers. Given the obvious road hazards in the NT roadmap, and Microsoft's history of missing target dates and changing product directions on a moment's notice, this fact -- that the marketplace will decide -- is exactly what Microsoft should be worried about.

Nicholas Petreley is editorial director of LinuxWorld and columnist for InfoWorld.

IBM will 'wait and see' on desktop Linux
March 22, 1999
Should you join the freeware revolution?
March 19, 1999
Get started with Linux
March 18, 1999
Put your bets on Linux in the enterprise, says Caldera's Love
March 17, 1999
The secret of Linux
March 15, 1999

Halloween V
Linux and the monopoly game
Linux servers at supercomputer speeds
Pentium III Xeon gets thumb's up from app vendors
(InfoWorld Electric)
Linux Alley is crowded, but lacks apps
(InfoWorld Electric)
Microsoft shamed in courtroom
(InfoWorld Electric)
IBM will 'wait and see' on desktop Linux

Note: Pages will open in a new browser window
External sites are not endorsed by CNN Interactive.

Internet Operating System Count for January 1999
Microsoft Corp.'s Windows page
Sunbelt Software
Netcraft survey
Sunbelt's March 1999 survey

Note: Pages will open in a new browser window
External sites are not endorsed by CNN Interactive.

Enter keyword(s)   go    help

Back to the top   © 2001 Cable News Network. All Rights Reserved.
Terms under which this service is provided to you.
Read our privacy guidelines.