IT Pundits Espouse Linux Benefits Including and Beyond TCO

by Timothy Prickett Morgan

For several years now, there has been a lot of mud slinging between the advocates of the Linux platforms and the advocates of the Windows platform. There is a lot of talk about total cost of ownership (TCO) differences between the platforms, and big companies with a lot at stake on both sides of the slinging (and some on both sides, ironically) expend a lot of effort and spend a lot of dough talking up Linux or Windows. The firm establishment of Linux as a viable and vibrant alternative to Unix and Windows has done nothing to change this.

Perhaps the most impressive thing about Microsoft compared to any other IT supplier in history is that it absolutely understands that you have to spend a lot of money on marketing to get money from specific markets. The late 1990s and early 2000s are unique in that there is a widespread mass marketing machine--which includes a diverse and well-read IT trade press as well as general TV, radio, and Internet advertising media--that IT players can pour their money into to make their marketing pitches. Since Microsoft has the biggest pockets in the IT world--the company has tens of billions of dollars of profits a year to throw at any problem it wants to address, and no company in the history of the world excepting perhaps the Roman Empire of the early BCs has ever come close to Microsoft in terms of profits--it has myriad marketing campaigns to compare and contrast its Windows platforms and their applications to alternatives. Microsoft has not disclosed how much it has spent on the "Get The Facts" campaign against Linux, which has been running since January 2004, but it is probably not a small number. It takes money to get IT consultancies to generate white papers and case studies, and then it takes a lot more money to plaster ads all over the press and the Internet to get people to look at this material.

The open source Linux community does not have such deep pockets as Microsoft has, even if you added up the pockets of every IT company that contributes to the Linux cause. The Linux community does not have one central organization that can coordinate and focus messages to specific markets, either. As you might expect, the Linux community is reliant on the enlightened self-interest of third parties to pay for white papers and case studies to be developed to make the case for Linux. IBM is the latest such vendor to sponsor research reports looking at the benefits of Linux.

In a conference call announcing two white papers that discussed Linux TCO and other issues beyond TCO compared to Windows and Unix platforms, Scott Handy, vice president of worldwide Linux at IBM, said these two reports were not a reaction to Microsoft's Get The Facts campaign, but rather a desire to quantify some other benefits of Linux as well as to update an existing and popular TCO study from 2002 by consultancy Robert Francis Group. The idea that this is not a reaction to Microsoft's protracted anti-Linux campaign is silly, but IBM is credible (as is Hewlett-Packard) in that, as Handy points out, it has a balanced view concerning Unix, Linux, and Windows. IBM will sell whatever the customer wants, basically. But IBM only spends money when it has to, and it would not have paid RFG or Pund-IT to do reports unless it felt compelled to offer countermeasures against Microsoft's own sponsorship of reports and case studies. I will leave it to the ethicists to debate the wholesomeness of vendor-sponsored reports, even if they are given freedom to speak their minds as Handy suggested IBM had done. All we can do is report on what each side in the TCO wars is saying and interject whatever good sense we can.

The RFG study, which is titled "TCO for Application Servers: Comparing Linux with Windows and Solaris," has more quantitative results than the Pund-IT study, which is called "The Unanticipated Second Stage Benefits of Linux." According to Chad Robinson, one of the two analysts who put together the RFG report, the way that people are examining and comparing Linux to alternative platforms in the enterprise is changing. "Several years ago, TCO was the metric, but now, it is only one of many metrics enterprises are looking at," he explained in a conference call hosted by IBM. He said that while TCO was something that was important to potential Linux shops, the idea of skills transfer from Unix to Linux platforms and the ability to get an operating system from multiple sources (which is not possible with Windows and only somewhat possible if you consider Unixes more alike than they really are) were key factors in the buying decisions among enterprises. He said while Microsoft has closed the TCO gap with Windows Server 2003 and Sun Microsystems had made great strides with its Solaris 10 platform, the issue of TCO is more complex. "With Linux, customers can choose the server platform, the distributor, and the support providers--and do so independently if that is to their advantage." This is possible only because Linux runs on more platforms than any other single operating system ever has in the history of computing--and the number would be larger if there had not been so much processor and server consolidation among server makers in the past decade. "Linux is cheaper still, and that is important," says Robinson. "But Linux delivers both monetary and non-monetary benefits." Again, the key non-monetary benefits are that it can run on everything from a cell phone to a mainframe and that Unix skills can transfer pretty easily to Linux.

It would not be TCO mud slinging without some numbers, and RFG was happy to oblige. RFG calculates that an application server running on a Linux platform costs $40,149 over a three-year time frame. Compared to $67,559 for a Windows box running the same application and $86,478 for a Solaris-based application server. These numbers are based on interviewing 20 medium and large enterprises (those with more than 250 employees) and doing a comparison of Java-based application servers running on Linux, Windows, and Solaris. RFG copped out on the Solaris comparison by pricing out Sparc-based iron, and justified this by saying Sun's base of Opteron servers was too small. Pricing for the server iron was not what these customers paid, but rather the price of the closest matching system in the SPECjbb benchmark database. In the case of Wintel and Lintel iron, this price was the average price of configurations from the Web sites of Dell, Hewlett-Packard, and IBM. Sparc iron was priced from Sun and directly off its Web site. Then the J2EE workload is added to the box running on top of Linux, Windows, or Solaris. The results were tweaked based on what RFG found during customer interviews; for example, Linux customers tended to have higher server utilization and run more apps on their servers than Windows boxes, so they tend to spend less per unit of work for a server. After normalizing the three platforms against the SPECjbb workload, RFG then calculated how much iron customers would have to buy to support a particular J2EE-based application server, and then added up the hardware, software, and maintenance (including human system administrators) costs for these platforms. The result is the numbers above, and you can bet Sun and Microsoft will be complaining about them and Red Hat and Novell will be tucking them into their salespeople's briefcases.

Interestingly, RFG found that for a given amount of work, Lintel iron is cheaper and Lintel software is cheaper, but Lintel support for the J2EE environment in particular is about 60 percent more than that for Windows and Solaris. This ended up being a small portion of the overall cost of each platform, however, and RFG said it's mostly the result of Windows and Solaris shops having dedicated Java administrators and system administrators, each with honed skills in their areas, while Linux users tended to have the Linux admin also take care of the Java platform, too, and often it took them more time and money to do so. This is, RFG says, a skills issue, not an inherent difference in the way Java is supported on Linux.

Charles King, the researcher over at Pund-IT, only talked to three customers for his case studies, but the idea was not to gin up some numbers but to get a lot of detailed information from customers in different industries and geographies. (One could argue about how relevant such an approach is to the server market at large, of course.) King echoed the sentiments of RFG by saying on the conference call that the initial focus on Linux was merely on lower-cost hardware, better TCO, and very strong point solutions--Web serving, file and print serving, and so forth. "But now people realize that Linux can help simplify IT efforts and also help companies to better use and retain their IT staff," he says. The customers King talked to were focused on consolidating their workloads, and Linux was one of the key enablers of this--particularly because even so-called legacy platforms can support Linux. (You can see why IBM, which is pushing Linux on its zSeries mainframes, iSeries minis, and pSeries Unix boxes, likes to pay for such research.)

King looked at the Linux deployments at Boscov's Department stores, a Pennsylvania retailer, Alliance Unichem, a European pharmaceutical wholesaler, and Zachid Tractor and Heavy Machinery, a Saudi Arabian industrial equipment maker. IBM picked all three of these companies on behalf of Pund-IT and set up the interviews, and it is no surprise they are poster children for IBM's Linux strategy. Alliance built a Linux-based data warehouse in 2002 and has since moved its entire Boscov's consolidated Windows servers onto Linux partitions on its mainframes; and Zahid is in the midst of a Windows-to-Linux migration, too.

While all three of these companies have very little in common from a business perspective, they all saw some things in common as they deployed more Linux and became more familiar with it. The so-called first stage benefits for Linux deployments had to do with having lower initial acquisition and software licensing costs as well as the ability to run on cheap iron. The second stage benefits Pund-IT found that Linux is a platform that can be deployed on many different kinds of servers and is often used as a server consolidation vehicle (often consolidating Unix and Windows workloads), and it is a very popular platform for newbie programmers and administrators (as Unix was 15 years ago). By being popular and cross-platform, Linux is helping customers recruit and retain IT staff. Another second stage benefit these companies saw was that they could standardize various software stacks and administrative tools more easily. The popularity of Linux has also made it relatively easy, says the Pund-IT report, for companies to convince even Windows shops that a jump to Linux is not a dead-end career move. So IT staffs are OK in making such a move.

RESOURCES

TCO for Application Servers: Comparing Linux with Windows and Solaris (Robert Francis Group)

The Unanticipated Second Stage Benefits of Linux (Pund-IT)