The Business Case for Linux

. Essay Far f ro m si m p l y p ro d u ci n g f reew are cl o n es o f ex i st i n g t ech n o l o g i es, Li n u x an d t h e o p en - so u r ce w o r l d n o w t u r n o u t so m e o f t h e b est so f t w are at an y p r i ce. Here i s t h e v i ew o f so m eo n e w h o’s m ad e a p ro f essi o n o f m ak i n g Li n u x w o r k f o r b u si n ess. The Bus i nes s Cas e f or Li nux Evan Leibovitch, Starnix n a world where no business can afford lengthy computer downtimes, selecting the appropriate operating system is critical. Most companies choose either Microsoft Window s or one of several commercial Unix offerings. Into this arena comes a third option, the Linux operating system, until recently considered of interest mainly to the academic communit y and hobbyists. Linux offers some of the best features of both MS Window s and Unix and introduces several benefits of its own. As a result, Linux is increasingly regarded as a viable contender in commercial installations. More than any other OS, Linux offers businesses both exceptional technology and a model that encourages the best possible mix of communit y and commercial involvement. My company, Starnix, was formed as a direct consequence of an extensive analysis of the new Linux marketplace, its commercial opportunities, and the obstacles it faces. Starnix’s choice of Linux as its strategic operating environment was not the result of the zealous enthusiasm that has occasionally characterized parts of the Linux communit y. Rather, it grew out of the partners’combined experience in Linux, Unix, and MS Window s environments— about 35 years’worth. The Linux models of communit y development and freely available source code seem incompatible with conventional assumptions about the way soft ware is created and distributed, but these obstacles are minor compared to the phenomenal potential that Linux— and open-source soft ware in general— offers the computing world. I 40 I EEE So f t w a r e Ja n u a r y / Fe b r u a r y 1 9 9 9 0 7 4 0 - 7 4 5 9 / 9 9 / $ 1 0 .0 0 © 1 9 9 9 . A REBIRTH OF UNIX During the Unix wars, which peaked in the early 1990s, vendors danced a fine line bet ween support ing Unix st andards and t rying to snare users into single-vendor dependence by adding proprietary “enhancements.” Interoperabilit y often took a back seat to politics and salesmanship. It took the threat of Microsoft Window s NT to get Unix vendors to agree on a common graphical interface, and even that was accomplished with great reluctance. At a September 1995 Unix trade show in New York, Novell announced it was about to dump the rich Unix legacy—previously dumped on Novell by AT&T—on the Santa Cruz Operation. SCO obtained the System V code in return for some stock and a small amount of cash. That’s all it came to be worth; a casualt y of wars fought in its own name. Unix was beginning to look like an unwanted orphan. Moments after that announcement, I visited the m odest Red Hat and Caldera booths at the show ; other attendees followed. Even given its many rough edges in 1995, Linux showed prom ise as a badly needed overhaul of Unix. Indeed, Linux was a rebirth of Unix, both technically and socially. Unix concepts that worked and had proven their value were kept, while much of its excess baggage was left behind— or so we hope. CURRENT BUSINESS USES OF L INUX In our experience, the existing business use of Linux mainly takes t wo forms: ♦ Linux can be found in developm ent shops, Internet service providers, and ot her businesses w here com put ing services and product s are t he main source of revenue. Here, developers and others who are Linux-friendly may already be in decisionmaking positions. ♦ Linux can also be found in businesses where Linux-friendly staff bring it in for its perceived technical merits, often without the explicit direction of IT management. Either experimental systems are set up that simply evolve into production systems, or management doesn’t specify an operating system for particular project s— perhaps assum ing that a more mainstream choice would be made. Demonstrating that Linux is technologically on a par with other mainstream OSs, such as Unix or NT, has never been difficult. First used by ISPs because of their razor-thin margins, Linux’s low cost made it a popular choice. As Linux started proving its reliabilit y and flexibilit y, it crept into installations where local technology staff deployed it, often in secret, as an alternative to mainstream systems. However, in general Linux has not found its way into corporate IT and small business systems. Selling managers of companies not mainly technically oriented on the benefits of Linux— “through the front door,” so to speak— is m uch different than convincing development shops, ISPs, and other technology-based businesses. Demonstrating that Linux is technologically on a par with other mainstream OSs, such as Unix or NT, has never been difficult. TECHNICAL STRENGTHS OF L INUX Some of Linux’s technological advantages, once they are spelled out, make it an easy sell. Scalability Linux can run (albeit at a limited level) on older 386 and 486 systems, making it attractive to organizations that have spare computing power or otherw ise unused system s. For sim ilar reasons, Linux will find an eager home in the third world and other sit uat ions (such as som e schools) w here lack of funds demands the use of slower or cast-off systems that cannot run Unix or NT well. At the other extrem e lie system s based on the Beowulf clustering technology, which closely ties together multiple Linux systems of any size. Using this abilit y, Linux practitioners have taken their systems to supercomputer levels of performance. The Avalon implementation, for instance, turns 140 Linux systems into the 113th fastest computer in the world.1 Flexibility There are few com put ing t asks a Linux system cannot be made to do. The key to this is free availabilit y of the operating system’s source code: anyone with the right expertise, or willingness to hire that expertise, can make any desired modification or enhancem ent. When Caldera t ried to port t he Novell NetWare server to Linux, it had to modify the kernel to enable the traditional Unix streams facilit y. Now it merely ships the patch with every copy of NetWare for Linux. However, Linux flexibilit y is about more than Ja n u a r y / Fe b r u a r y 1 9 9 9 I EEE So f t w a r e 41 . source code. It maintains the Unix ideal of using chains of small,nimble tools rather than huge,monolithic programs. By keeping the graphic subsystem out of the kernel, for example, it allow s dedicated servers to operate without any fancy resource-stealing video. In addition, it maintains powerful Unix remote-administration, batch, and scripting facilities. Linux users consistently report uptimes measured in months and years rather than days or weeks. Reliability By putting the video subsystem in the kernel, Window s NT leaves its system performance and stabilit y dependent on the speed and qualit y of thirdpart y video drivers. By using a modular kernel, which encourages the loading of modules only as needed, Linux minimizes potential problems. While most of the comparisons of crash resistance bet ween Linux and other OSs are anecdotal, Linux users consistently report uptimes measured in months and years rather than days or weeks. Even by Microsoft’s ow n research, “Linux has been deployed in mission critical, commercial environments and boasts an excellent pool of public testimonials.” 2 SOCIAL BARRIERS TO ACCEPTANCE Without claims of scalabilit y, flexibilit y, and reliabilit y, Linux could not get a foot in the front door. But technical attributes are far from enough. Linux faces its most significant hurdles in the nontechnical aspects of soft ware deployment: marketing, support, and general providing of “comfort” to IT managers. Marketing Challenges Until recently, Linux had a visibilit y problem. With few large soft ware vendors behind it, Linux publicit y pales next to that of Microsoft and traditional Unix vendors. Despite that gap, however, Linux has achieved a high level of public awareness. Linus Torvalds m ade the cover of Forbes m agazine in August 1998, and almost every major new s outlet has devoted space or time to the task of introducing Linux and the more general concept of free soft ware. One fortunate circumstance that has benefited Linux is the aggressive efforts of conventional vendors to tackle each other. The media—and the ven- 42 I EEE So f t w a r e Ja n u a r y / Fe b r u a r y 1 9 9 9 dors—have painted an active battle bet ween RISCbased Unix systems (such as Sun’s Sparc systems running Solaris) and Pentium-family systems running Windows NT.Linux,most commonly running on Intel systems,has been able to use this to its advantage by coming up the middle, demonstrating the best of both worlds. It combines the reliabilit y and adaptabilit y of Unix with the commodit y pricing and large variet y of suppliers offered by the Intel world. Also helping Linux is the number and caliber of soft ware com panies porting their applications to Linux— am ong them Netscape, Corel, Novell, and IBM. Oracle, Sybase, Inform ix, CA-Ingres, and Soft ware AG have either ported their database products to Linux or committed to do so. Only Progress Soft ware has yet to com m it, and even t hey say they’re looking hard at Linux. To be sure, the novelt y of Linux— a communit y developm ent project capable of holding it s ow n with the products of billion-dollar companies— also appeals to the media, which seems eager to paint a David-versus-Goliath scenario. In the wake of the US governm ent act ions against Microsoft, m any people and companies seeking an alternative to MS Window s are finding it in Linux. The Importance of Support At the top of the list of obstacles to mainstream IT use of Linux is its communit y’s insufficient provision of support that is not purely technical in nature. While the Linux world has shown it can track down bugs, t his is not as broad as t he um brella handholding offered by Unix vendors and Microsoft. Established companies have armies of salespersons and analysts skilled at translating corporate needs into computer solutions. In contrast, those driving innovation in Linux have generally been soft ware programmers, who concentrate on the tools and facilities that they and other programmers need. Some corners of the Linux world are directly addressing this. Companies such as Caldera have packaged a commercial version of Linux and have positioned themselves to market it through a channel of value-added resellers and integrators (such as Starnix). Such net works and resellers offering personal service will be critical to Linux’s success. Com m ercial OS vendors offer one t hing t hat Linux does not yet provide: a comprehensive and respected scheme of training and certification. In the form at ive years of Linux, cert ificat ion was not an issue because program m ers’ cont ribut ions could easily be judged by peer review— you were as good . as your last piece of code. This model does not work well in corporate IS. To that end, groups such as the Canadian Linux Users’ Exchange are working with Linux vendors and professional trainers to develop an effective testing and certification scheme. Training firms, eager to capitalize on the growing popularit y of Linux, are joining the dozens of book titles on the subject to offer the education necessary for serious corporate work using Linux. If and when a certification effort happens, it will offer a standard level of expertise that employers and contractors can expect, much as program s such as those for the Microsoft Certified Engineer and the Certified Net ware Engineer provide for other OSs. The Ease-of-Use Issue The one social barrier to Linux’s success that can be addressed through technical solutions is in ease of use compared to other OSs. Its programmer-oriented initial developm ent has assum ed a certain level of fam iliarit y, and this attitude has deterred many potential corporate users. Current attempts to simplify administration tasks, such as linuxconf 3and COAS,4 show promise but are still immature. More general ease-of-use issues have been resolved as the Linux communit y has settled on t wo GUI systems, KDE5 and Gnome.6 Open-Source Culture The open-development model that drives Linux is in some ways both its greatest benefit and its primary weakness. This distributed model started by Torvalds has proved to be flexible, even though it maintains a single direction in a diverse communit y. The m odel has been effect ive from day one, and Linux has avoided much of the fragmentation that has beset the BSD freeware effort (and, in fact, the Unix field in general). Yet som e of t he sam e elem ent s t hat m ake t he Linux model attractive to technology fans are downright scary in som e corners of IT m anagem ent. Because there is no “Linux Inc.,” there is no central authorit y to blame— that is, nobody to be held liable, legally or otherwise. According to a paper that describes commercial levels of Linux acceptance,7 one of the most revealing indications of the influence of fear, uncert aint y, and doubt (FUD) is t he common belief that Linux has no vendor support. Such allegations are false. The Linux model combines com m unit y-driven support w it h t he availabilit y of commercial-grade (and commercial-cost) backup, and has routinely been found superior to t he convent ional help-desk m odel em ployed by most commercial vendors. The effectiveness of the Linux model was recognized when it was awarded InfoWorld ’s 1997 Product of the Year award for best technical support.8 (Ironically, t he editors at InfoWorld half-com plained about not know ing to whom they could present the award!) However, t here is cert ainly potent ial for problems, and some fear fragmentation in the communit y. As a volunteer effort, Linux developers trade in respect and ego rather than dollars. The possibilit y of ego clashes in a world that depends so heavily on volunteer work always exists, and threatens the kind of splintering that has plagued the Unix com m unit y. Another perception of dissension comes from within other areas of the non-Linux freeware communit y. As Linux gains popularit y, many are stepping out of the woodwork to either claim a piece of its honor or complain at being left out.9,10 Much of t he Linux crowd has also refused to m arch lockstep to t he drum of t he Free Soft ware Foundation. Contrary to FSF doctrine, many Linux dist ributors have encouraged t he addit ion of com m ercial and alm ost -free soft ware solutions (such as the BRU backup system and t he xv shareware graphics viewer) w hen t hese solutions are seen to be superior to freeware offerings. St ill, Linux has a long way to go to provide all the am enit ies offered by com m ercial vendors. The Linux com m unit y is addressing m any of t hese issues; t hey m ust succeed if Linux is to b ecom e a t rue alternat ive to conventional system s. Commercial OS vendors offer one thing that Linux does not yet provide: comprehensive and respected training and certification. THE NEED FOR OPEN DISSENT While som e w ithin the Linux com m unit y have sought to m uzzle dissent, others have suggested t hat t he Linux com m unit y needs a Gates- or McNealy-like parent figure.11 But changing the rules by closing opinions, hiding discussions behind closed mailing lists, or imposing solutions without consensus would be not hing but dest ruct ive. To submit to those who complain about the public airing of differences is to change the principles that Ja n u a r y / Fe b r u a r y 1 9 9 9 I EEE So f t w a r e 43 . have allowed Linux to challenge the commercial systems, so well and so fast. Just as dangerous would be to insist upon the path dictated by the FSF that collaboration with conventional soft ware vendors is to be avoided at all costs. Listening to the needs of the end user and corporate communit y is vital. Every organizat ion has it s internal squabbles about direction. Most companies have the luxury of keeping internal dissent internal, and let t ing t he public know only the sweetness and light that result. The Linux communit y’s openness— the same openness that allowed it to achieve the technical success it has to date— eschew s nondisclosure agreements or other forms of limiting freedom. It also means that the dirt y laundry hangs outside, in day-glo colors, for all to see. On the positive side, the open process has led to t he form at ion of groups such as Linux Internat ional 12 and init iat ives such as t he Linux St andard Base project,13 to set t le differences bet ween various proponents’approaches in a (usually) nonconfrontational st yle. In a manner intended to emulate Torvalds’ st yle in leading the kernel development process, these cooperative projects seek to encourage innovat ion w hile recognizing t he fine line bet ween diversit y and confusion. Those disturbed by open discussion, dissent, and even open disagreement can just tune it out. Just as one can belong to a church without needing to endure rancorous board meetings, one can simply purchase one of the commercial Linux distributions, let the developers argue all they want, and listen in only as desired (if at all). The product’s qualit y w ill not have changed one bit. All the qualities that m ake Linux a superior choice for so many computing solutions still exist, and they are progressing faster than conventional systems. M icrosoft, in its now-famous “Halloween document,” 2 stated that Linux suffered from “taillight-following.” That is, according to Microsoft, the most important features and facilities of Linux are simply open-source reactions to what others have already done. Such an assessment is simply wrong. Linux did not merely reinvent Unix; it reinvented the way operating systems are made. Netscape, IBM, Sun, and others have already started open-sourcing some of their technologies; others are considering sim ilar m oves. And far from sim ply producing freeware clones of existing technologies, Linux and the opensource world now produce som e of the best soft- 44 I EEE So f t w a r e Ja n u a r y / Fe b r u a r y 1 9 9 9 ware at any price. In stark contrast to much of the concern expressed about it, Linux also offers some of the best support; in fact, sometimes the best support is soft ware that simply works reliably. Still, in many ways, Linux is immature—especially in its abilit y to cope with the nontechnical needs of the business communit y. As it matures, it will find itself accepted into an increasing number of corporate environm ents, running databases and kiosks and just about any business computing task possible. This evolution of Linux— and of the entire computing world— is well underway. ❖ REFERENCES 1. Top 500 Supercomputer Sites, 5 Nov. 1998, http:/ / www. top500.org/ top500.list.html. 2. E. Raymond, “Halloween I (1.10), Open Source Software: A (New?) Development Methodology,” http:/ / www.opensource. org/ halloween1.html. 3. Linuxconf, http:/ / www.solucorp.qc.ca/ linuxconf/ . 4. Caldera Open Administration System (COAS), http:/ / www. coas.org/ . 5. K Desktop Environment (KDE), http:/ / www.kde.org/ . 6. GNU Network Object Model Environment (GNOME), http:/ / www.gnome.org/ . 7. E. Leibovitch, “The Four Phases of Linux Acceptance: An Approach to Linux Advocacy,” July 1998, http:/ / www.xunil. com/ xunil/ j4phases.html. 8. E. Foster, “Best Technical Support Award,” Infoworld , http:/ / www.infoworld.com/ cgi- bin/ displayTC.pl?/ 97poy.supp. htm. 9. A. Leonard, “The Saint of Free Software,” Salon , http://www. salonmagazine.com/21st/feature/1998/08/cov_31feature.html. 10. Perform ance Com puting , http:/ / www.performancecomputing. com/ features/ 9810of1.shtml. 11. N. Petreley, “Down to the Wire,” Infoworld , http:/ / www.infoworld.com/ cgi-bin/ displayNew.pl?/ petrel/ 980824np.htm. 12. Linux International, http:/ / www.li.org. 13. Linux Standard Base, http:/ / www.linuxbase.org/ . About the Author Evan Leibovitch is a partner in Starnix, a Linux-centric products and services firm based in Brampton, Ontario, Canada. He also writes a monthly column on Linux business issues for Computerworld Canada magazine. He is heavily involved in Linux user groups and was a cofounder of the Canadian Linux Users’ Exchange (http://www.linux.ca). He is also involved in a community project to implement professional certification for Linux. He holds degrees in philosophy from York University and in journalism from Ryerson Polytechnic University. Readers may contact Leibovitch at evan@starnix.com.