CNET también está disponible en español.

Ir a español

Don't show this again

Tech Industry

Commentary: The way of Linux

In 2003, the open-source operating system will go mainstream as cost-sensitive IT executives discover the compelling economics of Linux on Intel.

    Commentary: The way of Linux
    By Forrester Research
    Special to CNET
    January 24, 2003, 3:00PM PT

    By Ted Schadler, Principal Analyst

    In 2003, Linux and open source will go mainstream as cost-sensitive IT executives discover the compelling economics of Linux on Intel processors. Smart CIOs will set aside the religious wars to see the truth: Open source is good enough for many data center tasks.

    Despite its recent rise to fame, open-source software isn't new--after all, it's anchored the Internet with BIND and Sendmail for decades and runs more Web servers than any proprietary alternative. What's different now? Open-source software--and the hardware from Intel and Advanced Micro Devices that most users run it on--has passed the good-enough point. Put another way, today's open-source rewards outweigh the risks.

    But sorting through the open-source propaganda and avoiding open-source choices that don't make sense--like desktop operating systems--isn't easy. To avoid gaffes, start your strategy-planning process with answers to these questions: What's different about open-source software? Which open-source products should firms use? What's the right open-source deployment strategy? Which vendors can help with open source?

    Open-source software differs from commercial software in important ways that stem from its development processes and licensing terms. CIOs should understand three key dimensions:

    First, development. The scientific method yields great platform software. The open-source development process is analogous to the scientific method--with published results, peer review and so on. It's particularly well suited to commodity infrastructure software because the incentives for open-source contribution--accolades of a large community and the personal satisfaction of seeing your code broadly adopted--are mostly found in platform software rather than niche application software.

    Related story

    Linux has arrived as a mainstream
    operating system.

    Second, licensing. Open-source software licenses are usually free. But beware: Other costs--like documentation, support and commercial add-ons--will still swell a company's IT budget. Revenue for Red Hat, to look at just one example, is from subscriptions (code base maintenance) and service rather than licenses. One note: The GNU General Public License (GPL) terms--under which direct enhancements to Linux must be made open source--are sometimes seen as a fatal flaw in the open-source story. Forrester disagrees.

    Third, support. Open-source software has four kinds of vendors competing for support: distributors like SuSE for Linux and Covalent Technologies for Apache; hardware vendors like Dell Computer, Hewlett-Packard and IBM; software vendors like SAP and Oracle; and traditional systems integrators and outsourcers. These vendors compete fiercely on quality and price, yielding an efficient market for buyers.

    When and how to choose reports more than 50,000 open-source projects. But let's be clear: Only a handful are ready for data center deployments. In general, companies should do the following:

    • Accept open source in embedded systems. Linux is popping up in many storage subsystems and network appliances; every device needs an operating system, and Linux is free and reliable. Companies can comfortably accept Linux inside a DataPower XML switch or in a Panasonic DVD-RAM drive. Why? Because it's the hardware vendor's job to support the software--not yours.

    • Consider open source for commodity platform software. Open source's sweet spot is in platform software: the Linux operating system, Apache Web server and PHP scripting language. The bottom line? If you're a Unix shop today, you'll probably be a Linux shop tomorrow. Why? Your skills, procedures and even applications will port easily--and you'll save on the Web server, operating system and hardware.'s well-publicized move from Solaris to Linux on Pentium-based HP servers helped cut Amazon's technology capital budget by 25 percent.

    • Avoid open source on the desktop and for most applications and tools. With the success of Linux and Apache, it seems like every vendor and developer has initiated an open-source product. But for the next three years, CIOs will be smart to avoid the bulk of them because most will fail one of Forrester's three enterprise open-source filters.

    Devising a strategy
    Companies can deploy and benefit from open-source software without jumping into the open-source community. In fact, companies that don't have a deep technology ?build? culture should treat open source like commercial software: Hands off the code. How should beginners deploy open-source software?

    • Fund an open-source competency center and staff it with skeptics. CIOs making a commitment to open source should also commit to a team that can demystify licensing issues, manage code rollouts and check a project's sanity level. Staffing the center with skeptics--not gurus--will keep corporate technology policy far away from the open-source socialist fringe.

    • Treat open-source deployments more rigorously than commercial ones. With open source, the cost barrier to upgrades is gone, right? Wrong--most of the cost of upgrades is in labor, not software licenses. Indeed, the hidden costs of upgrades, such as incompatibilities with other applications and hardware patches, may be even more severe in an open-source environment. Smart IT executives will learn by engaging a partner like HP or Red Hat for Linux deployment support.

    • Don't modify open-source code unless you want to join the community. Only companies that want to see their utilities, tools and extensions maintained by someone else should modify or extend the code--especially with code like Linux that's protected by the GPL. See the Open Source Initiative for more information about open-source licenses.

    The role of the CIO
    Companies need the same help with open-source software that they need with their mission-critical applications and systems. For example, no company should make a bet on a product like Apache or JBoss without lining up vendors to help. Ask the vendor with the most skin in the game--typically your hardware vendor, but sometimes an application vendor--to support your open-source deployments. For Linux, CIOs should:

    • Standardize on a Linux distribution: Red Hat or UnitedLinux. Linux distributions (the combination of the open-source kernel and add-ons and vendor-specific features) are not identical. First choice: Red Hat, which has more market share than all the other Linux distributors combined. Smart alternatives include SCO and Germany-based SuSE, both of which are part of the UnitedLinux consortium.

    • Look to infrastructure giants for broad support. Revenue for Linux distributors' comes from service and support, and companies should engage these specialists when needed. In the long run, however, gear vendors like HP and IBM are your best bets for Linux support, while Java application server vendors like BEA Systems and Oracle are best for Apache. Already, HP claims 5,000 Linux service professionals, while IBM crows about its 4,000 Linux customers.

    The effect on the software industry
    What will be the five-year impact of open-source development on the software industry? One thing's for sure: Open-source software will gobble up more application and middleware workloads.

    • A maturing Java-on-Linux stack will depress software server pricing. JBoss and Apache Tomcat are already closing in on commercial Java 2 Enterprise Edition (J2EE) application servers and "servlet" engines. In five years, low-cost open-source middleware will compress the prices of commercial application servers from BEA, IBM, Oracle and Sun--and Microsoft, too.

    • IBM will shed nondifferentiating middleware. Today, IBM wields Linux brilliantly to retain its Unix, iSeries and mainframe customers in the face of the Microsoft juggernaut. By 2005, IBM will embrace enterprise-ready open-source middleware as its own to tap into developer interest in open source while retaining the value of its WebSphere brand.

    • Microsoft will finally accept, then master, open-source development. Microsoft today treats Linux as a competitor to be slain. But customers don't see it that way: They just want a simple migration path from Unix/RISC to Intel. Microsoft will ultimately realize that in the battle for customer wallets, it must support a Windows-Linux cohabitation strategy. One result? Redmond will shock the tech world by open sourcing the lowest layers of its common language runtime, allowing Windows applications to run on any box, from mainframe to IP phone.

    • Larry Ellison will establish the "Ellison Prize" for software innovation. Today, open-source software is better at creating stable versions of known technologies than introducing entirely new kinds of software. What's missing is not the process--rather, it's the motivation to innovate. In contrast, scientists can look to prizes like the Nobel as a motivator for innovation. So today's software barons should create their own $1 million prize, starting with an award for Kernighan, Ritchie and Thompson, the Bell Labs crew that gave us C and the Unix kernel.

    • Open-source distributors will survive as service companies. Red Hat's IPO and subsequent stock run-up suggested that a new business model for open-source distribution would be fabulously profitable. Hogwash. Linux distributors are service and software distribution suppliers, a profitable business more like Electronic Data Systems or SAIC--not like Microsoft.

    Finally, in addition its own research on Linux, Forrester recommends two open-source books. Eric Raymond has written the definitive analysis of the open-source development process and culture in "The Cathedral & The Bazaar," and "The Business and Economics of Linux and Open Source," by HP's Martin Fink, explains issues of licensing and deployments.

    © 2003, Forrester Research, Inc. All rights reserved. Information is based on best available resources. Opinions reflect judgment at the time and are subject to change.