The Linux community disputes Microsoft’s TCO claims.
How do small businesses save money on technology? That is the theme of this year’s small-business issue in a nutshell. In a time of tightened licensing and escalating software costs (relative to total costs), one strategy to consider is open-source software. But while Linux and its cadre of tools have very low initial costs, the numbers get muddled when you consider talent acquisition and training, downtime, and other hidden costs.
Framingham, Mass.-based IDC recently released a study comparing total cost of ownership (TCO) of Linux to that of Windows 2000. Considering that Microsoft sponsored the study, you might not be surprised by the results, though it did send shock waves through the Linux community. As reported by ZDNet, In four out of five server loads (file, print, etc.), Windows 2000 has a lower TCO, projecting current costs out over the next five years. According to the study, Web servers constitute the only server load in which Linux is cheaper.
Recently, I talked separately with two prominent members of the Linux community about how Linux can save small enterprises money. Mark de Visser is vice president of marketing for Red Hat, the leading Linux distribution. And Darl McBride is president and CEO of the SCO Group (formerly known as Caldera) and a member of UnitedLinux. The study came out just after my talk with McBride and just prior to my talk with de Visser. Because the study informed our talk, I’ll lead off with de Visser.
Mark de Visser
Q: How do you assess the IDC study?
A: The outcome of any study is dependent on the input parameters. And there are a number of questionable parameters in this study. First of all, the study does not differentiate between Linux distributions. Because the study compares Windows against a sampling of modern Linux distributions, downtime and administration costs become the most important factors in assessing the costs of the system.
Most Linux distributions come in two parts. There is an enterprise-class distribution and a consumer distribution. Enterprise-class distributions are specifically designed to drive down total costs. Consumer-level distributions are designed to show off the latest and greatest Linux software. Because they include a lot of third-party tools that can lead to system conflicts, consumer-level Linux distributions are more prone to downtime. Enterprise-class distributions are based on the most stable systems available, with integrated tool sets and minimal conflicts.
If you have one Windows system, in this case Windows 2000, you should compare it to a single enterprise-class Linux distribution, for example, Red Hat Linux Advanced Server. Linux Advanced Server is a managed service that updates more than 1 million enterprises on software and security patches routinely. Our customers have not had downtime (and we’re retrofitting between 5,000 and 10,000 systems a day for the service). If you do this, you actually save money over Windows 2000 on downtime and administration costs, and the TCO results come out in favor of Linux.
Second, the study projects the current costs out over the next five years. Those costs will probably change. For one thing, I would be surprised if Microsoft didn’t raise prices in five years. Of course, we know what’s happened this year: They’ve both raised prices and tightened licensing restrictions. Is this study a tacit agreement on Microsoft’s part not to raise prices for a five-year period? It would be interesting to see what Microsoft says about that. If you have 100 percent market share, the only way to achieve revenue growth is through increased prices.
Finally, costs are a matter not only of the process of reducing administration but also the skills necessary to reduce administration. We have been working on the skills part of the cost equation for a long time, and our Red Hat Certified Engineer (RHCE) program is a good indicator of all this work. But we realized a while ago that we needed a mechanism to get administrators up to speed on basic skills more quickly, so we developed and recently released the Red Hat Certified Technician (RHCT) exam. We are the only Linux distribution that has developed this two-level certification system. As more people become certified in Linux server administration, the costs associated with skills will tend to even out over the next five years.
Q: I’ve heard a lot about unbreakable Linux from Oracle and others. This would seem to fly in the face of Microsoft having less downtime. What do they mean by unbreakable Linux?
A: They’re talking about clustering. The low cost of Intel-based hardware enables users to replicate a server several times into a cluster of redundant servers. So one server going down has no effect on the uptime of the entire enterprise. If you use Linux in a managed service, the service will alert you if one server in a cluster goes down so you can get it back up and running while others take over its load.
We have a customer who installed a managed cluster in February , and they’ve had zero downtime while doing multiple millions of transactions since they installed it.
Q: How does this type of clustering compare with Beowulf clusters used in supercomputing?
A: Beowulf clusters are used for parallel processing to achieve very high computational performance rates. That’s different than the availability type of clustering in so-called unbreakable Linux, which doesn’t enhance performance; it enhances reliability.
Q: Doesn’t Windows 2000 enable availability clustering as well?
A: I do believe that Windows 2000 has that capability. It’s just not talked about much.
Q: You mentioned the need for Microsoft to grow revenues. This is a constant pressure for a public company. Being a public company, how does Red Hat respond to Wall Street’s pressure for revenue growth?
A: We are very fortunate to be in an industry that is growing. Linux is growing just on its own momentum. Our growth comes as a natural consequence of that. Once the server platform growth slows down, there are other platforms we can grow into, including the desktop and embedded markets. So there’s plenty of room for growth in the foreseeable future.
The Mark de Visser personnel file
Mark de Visser, vice president of marketing, joined Red Hat in May 2001, after serving as vice president of marketing at PlanetPortal, an Internet infrastructure company. He worked more than 12 years at Borland International, first in Europe and later in the United States, where he helped launch products such as dBASE for Windows, Delphi, C++Builder, and JBuilder. As vice president of marketing communications, he achieved broad recognition for Borland as an enterprise software company.
Q: How does the name change from Caldera and the UnitedLinux platform improve the SCO Group’s chances for success?
A: It differentiates us from enterprise-level distributions such as Red Hat. We’re focused on small- and medium-sized business sites. That does not mean we only sell to small- or medium-sized companies. Sometimes an SMB site is located inside a larger company, especially when small business units are given autonomy within a large enterprise.
Linux is very well suited for a smaller organization because it scales incrementally without large initial cash outlays. SMBs can tailor their enterprises to better fit their needs with Linux. You can choose only those features you need rather than buying an expensive, bloated package and only using some of it.
Q: What other attributes of SCO Linux make it the distribution of choice for SMBs?
A: There are a lot of reasons SMBs should look seriously at SCO. I think the primary reason is reliability. We measure reliability or mean downtimes in terms of hours per year. Other competitive products measure their mean downtimes in terms of days per year. We also have a strong retail vertical, which gives SMB retailers an edge in available tools.
Q: How does SCO Linux compare with other distributions within UnitedLinux?
A: UnitedLinux will be basically the same as other distributions in the group. But we have a dealer network of nearly 16,000 resellers who can support the product on a local level. SCO has always been the UNIX for the Intel architecture. Also, with the advent of SCO Linux, we can leverage our existing 8,000-member developer network and all the tools they have created for SCO Unix. This allows SCO customers to tap into a toolkit for their current systems to make these applications run under Linux.
Q: Do they simply move those UNIX applications over to Linux with new shell calls or will they need to recompile the applications for Linux?
A: Most of the time, they will be taking the application libraries and making them available on Linux. There might be some recompiling involved, but it will be much simpler than starting from scratch.
Q: With the developer network in place, how will SCO add value to its offerings with new applications in the near future?
A: Some of the tools that already exist and are simply being updated include an Office Server, which gives companies Microsoft Exchange functionality at a fraction of the cost. Companies can plug a whole group of Outlook clients into this server without much hassle. Another offering is called SCO Biz, which is an e-commerce platform for retailers.
Q: If SCO is spending a lot of money developing all these add-on applications, how can it justify making them available as open source for anyone to plug them into other Linux distributions?
A: Some apps will be open source and some will be proprietary. The key is the application platform that can accommodate either type of application coming from our developer network. Even our open source applications will be more easily managed using the SCO application platform.
Q: How does SCO monetize its open source applications?
A: As the Linux train continues to gather momentum, more and more server sales will come from folks who need the applications to run under them.
Q: When I talk to IBM, Microsoft, or other large software companies, there seems to be a shift in focus toward delivering solutions rather than just delivering products. That means not only providing software but helping companies integrate everything through personalized service. I guess that’s implicit in the name Web services. How does your reseller network figure in your own ability to service your clients?
A: You’ve hit on a key advantage for SCO. When you can get personalized service from one of our 16,000 resellers, it increases the value of the distribution immensely. We encourage our customers to team up with the resellers–especially in the SMB market, which is 82 percent of our installed base–where the resellers essentially are their IT staffs.
Q: Where do you see SCO Linux at the end of 2003?
A: We see 2003 as a potential breakout year for Linux in the SMB marketplace. We now have both UNIX for Intel and Linux complete with an application developer network and reseller network specifically focused on this marketplace. So we think we’re in a pretty good position.
The Darl McBride personnel file
Darl McBride, the SCO Group president and CEO, has 19 years of executive management and leadership experience. Before joining SCO, McBride was the president of Franklin Covey’s online planning business. Prior to that, McBride has been the CEO of PointServe, a workforce optimization software company; and the founder, chairman, and CEO of SBI and Company, a professional services company. McBride has also been the senior vice president of IKON Office Solutions where he managed 4,000 employees and the buildup of a $500 million systems integration unit. From 1988 to 1996, he worked at networking leader Novell where he concluded his tenure at Novell as vice president and general manager of Novell’s Embedded Systems Division.