So Much for the Linux Threat

I've reported in Windows IT Pro UPDATE several times over the years about Linux and its potential to unseat Windows Server as the most used enterprise OS. As a general rule, each January seems to bring a collection of "This Will Be the Year of Linux" stories, typically from analysts who've been bowled over by the Linux hype. To be fair, I've always assumed that Linux and Windows would some day run neck-and-neck in the server world, with Linux's perceived security, cost, and reliability advantages as the major reasons. Also, the past few years have been tough on Microsoft, as the company has suffered through a mind-boggling series of security snafus.

The Linux hype has just one little problem. Despite steady improvements over the past several years and the support of major IT companies such as IBM, Novell, and even Sun Microsystems, Linux seems stuck in a perpetual holding pattern, unable to eat away at Microsoft's server market share. And as the PC industry comes out of an economic recession and enterprises resume technology spending, it's interesting to note that Microsoft solutions, not open-source solutions such as Linux, are making the biggest gains.

Case in point: In its most recent quarterly earnings announcement last week, Microsoft once again beat forecasts and set an earnings record. Record earnings happen so regularly at Microsoft now that it's almost not worth mentioning. But key to the company's success, interestingly, is its Server and Tools division, which is responsible for such products as Windows Server 2003, Microsoft Exchange Server 2003, and Microsoft SQL Server 2000. The Server and Tools division made revenues of $2.8 billion in the quarter ending December 31, 2004, the same amount of money as the Information Worker division, which sells the Microsoft Office cash cow, made. Indeed, Server and Tools almost edged out Microsoft's other cash cow, the Windows Client division, which earned $3.2 billion.

Those results are amazing. Server and Tools grew more than 18 percent year-over-year, compared with flat or single-digit growth for Windows Client and Information Worker. SQL Server growth topped 25 percent. And Exchange 2003 is off to the fastest start of any Microsoft server product. These figures indicate two things: First, the IT industry is spending money again. Second, Microsoft's server products are kicking butt, and they're doing so at a point in time in which all the core products--Windows Server, Exchange, and SQL Server--are fairly mature. And because both Windows Server and SQL Server will see major updates this year--Windows 2003 Release 2 (R2) and SQL Server 2005, respectively--we might expect the upgrade treadmill to keep revenues rolling for quite a while.

"Our server business has a track record like the [New England] Patriots in the NFL playoffs," Microsoft chief financial officer (CFO) John Connors said, according to a report by Todd Bishop of the "Seattle Post Intelligencer." The comparison was carefully selected, I think. The Patriots, which have advanced to the Super Bowl in three of the last four years, are seen as a modern sports dynasty and widely respected for their leadership and team-oriented attitude. Microsoft would like to see its server products as well-respected as the Patriots are and would like to foster the notion that, although each individual server has certain strengths, they work together in such a way that the whole is more valuable to customers than the individual parts.

Some key challenges will still bedevil Microsoft as it attempts to fight back against the Linux threat, although recent history suggests the company might have finally latched onto a winning strategy. First, Microsoft must counter the perception that Linux is more secure than Windows. We're just starting to see some people come around to the notion that a largely untested solution such as Linux can be as insecure or more insecure than Windows, given improper configuration.

Second, Microsoft must prevent an upswing in support for Linux and other open-source solutions in world, regional, and local governments. In some cases, Microsoft has won government contracts by sweetening deals financially. But more often than not, fear of moving to an unknown and unproven system has kept many governments firmly in the Windows camp. And widely publicized Linux conversions--such as the one in Munich, Germany--have predictably run into problems. More important, they still represent a small portion of the overall worldwide government IT market.

Third, Microsoft should continue pushing its integration approach, which is truly a huge competitive advantage. Turnkey products such as Windows Small Business Server (SBS) 2003 and the awesome services industry built around it are unparalleled in the open-source world and will likely continue to be so for some time. Although it's interesting to make product-to-product comparisons--such as Windows 2003 versus Red Hat Enterprise Linux--few customers think in such fine-grained ways. Enterprises want solutions. And I think this is an area in which Microsoft comes out on top.

Fourth, I think Microsoft has finally won the battle over cost. Depending on whom you talk to, Linux solutions are cheaper or as expensive as Windows-based solutions. That comparison doesn't resonate very loudly with IT administrators who are already familiar with Windows and would dearly miss functionality and compatibility if they left the platform. Even if some Windows solutions are a bit more expensive than Linux-based alternatives, the benefits of Windows often outweigh what is essentially a small price differential, spread out over time.

So what do you think? Is Linux the next big thing, or will it simply snag a few key niche markets like most of Microsoft's past competitors have?

I've seen it all and done it all, Bushie

So have I, Hal. So have I...

... I can tell you that you can devise the perfect plan and spend millions of dollars in disaster mitigations plans - and something will still go wrong. In the real world, we have things called "budgets" and "cost-effectiveness" that limit our ability to totally eliminate downtime.

Hal, perhaps you're misunderstanding. There are many types of systems that simply can't afford to be offline for any reason whatsoever. Example: Air traffic control systems. If you put your budget ahead of the resulting downtime, you've already failed a fundamental and primary requirement before you even undertake to setup such a system. Consequently, what I'm saying is that cost is not the most important factor to folks that need 5x9's of availability. It's well down the list, certainly well after availability and fitness for purpose.

5x9-class sytems are designed to avoid unplanned maintenance (ie. hardware and software failures), not survive any possible environmental disaster. Planning around disaster conditions is an orthogonal level of planning -- and has nothing to do particularly with the performance of a 5x9 system. There are things that you can do to avoid environmental impacts such as providing redundant power generation systems. Most nuclear power plants are constructed with highly-reinforced concrete in order to avoid disintegration during and after an earthquake.

5x9 systems tend to be very expensive. Most organizations don't need them -- and, based on your comments, I would hazard an educated guess that you simply don't need or use one.

So what you're saying is that if you required 5 nines, you'd jump all over the first vendor to come along, rather than shop around with those requirements? Give me a break.

That's not even close to what I was saying. Of course, you evaluate and compare the offerings of 5x9 vendors. But you DON'T even bother comparing with 4x9 vendors if you NEED 5x9's.

Oh, dang, I've been in Atlanta, didn't see this thead.

Just like old times! I miss the OS wars.

I am curious, I don't actually know -- are there vendors willing to guarantee 5x9 uptime with Windows/Intel boxes?

Personally, I'm seeing a *lot* of Linux. Some Windows. It varies, but I do see more Linux than Windows in the Enterprise.

But that's just in my J2EE-centric world, and YMMV.

If I were Linus, I'd have the system automatically diff every submission against the Sun source code.

That really only solves the kernel code problems. I have a feeling that a lot of the Sun code will be more useful in userland. Having every submission to every package diffed sounds like a pain in the rear.

This is why I am such a big fan of the BSD license, I wish people would release code in that form. Too bad even the BSDs can't be built without GPL code anymore.

-paridel

I am curious, I don't actually know -- are there vendors willing to guarantee 5x9 uptime with Windows/Intel boxes?

Scratch the Windows part... are there vender's willing to guarantee 5x9 on Intel boxes?

I think when you are talking that level of reliability I think the hardware failure possibility is much more important (not than the software, than it is for say 99.9%), and if you are talking about just a normal Intel box, even if it has cool features like hot swappable power supplies, I am almost certain the answer is no.

-paridel

Hmm great link lets see what I was posting in reply to.

Bush 2000: But, if you want to run Linux in your company as a database server or in some other kind of mission-critical role that demands five 9's of availability, it's not going to have the same kind of track record or reliability as Solaris.

So again, you said Linux is not good for 5 9's why dont you prove it..

Bysh, Lets sum things up

You say solaris gives 5x9 (which they dont guarentee on their mid range) but that Linux does not.

You then try to restrict the Linux side to Intel hardware, while thats not the hardware solaris offers. While solaris does support the x86 arch they dont guarentee 5x9 on it.

I point out for a cost (and a big one) IBM will support 5x9 on the Linux intel side. The problem is you have to hand them the keys, I dont have root on our production servers. I make changes in dev/test and pass on the instructions to IBM. I point out an article about a company teamed with IBM running a 5x9's linux intel server.

I also point out that solaris on sparc can be as expensive as Linux on a mainframe (IBM has come down in cost on their Mainframes lately).

Dude youre losing this one, let it go.

Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.