In this chapter we’ll look at the ways to use Open Source software inside your company. We’ll look at these strategies from the viewpoints of the corporate user, the small-business user, and the sole proprietor. This chapter will spend more time on large firms because the use of Open Source software is more widespread and more complex in them, where Linux is a networking star, particularly among heterogeneous systems.
Open Source software finds its way into businesses by the same process that brought in earlier technologies, including the computer itself. In the technology adoption life cycle, new technologies are first the province of the innovators, highly technical people who are distinguished by their love of technological novelty and their lack of budgets. If a business case can be made for the new technology, the visionary leaders, called early adopters, will pick it up, first in a pilot project, and then in regular deployment. These visionaries have budgets, and are willing to spend substantial sums on a technological advance that will produce a competitive advantage. When enough visionaries have been successful and the technology is sufficiently spread and proven in a number of niches, it may then pass into general use. We will be taking a look at Linux from the perspective of that life cycle.
The case for the adoption of Linux (and of BSD, the other major Open Source operating system) has been low cost and high reliability; we will see how these have been the motor for the spread of Open Source through the business world. The only cost for taking advantage of Open Source software is relative to the technical proficiency of the company. The use of open standards by Open Source software is a benefit available to businesses of any size. Open standards allow interchangeability of hardware, software, and of support vendors, so that a user is able to take advantage of competition among all of these. Similarly, open standards enable software vendors to write competing applications using standard protocols open to all users. Probably the most important advantage of Open Source software is the ability of a user to customize an application for its most effective use. You can do the customizing yourself, if you have the skills, or hire whom you please to do the job. No longer does a user have to send begging mail to a software vendor pleading for modifications or enhancements. This flexibility applies at both the operating system and the applications level.
In the corporations, Linux first appeared in the hands of technology enthusiasts, innovators who run the systems of corporations and have their own networks set up at home. For those used to working in UNIX, Linux enabled them to run UNIX programs at home on their own inexpensive Intel machines. As corporations became aware of the Internet and wanted to set up e-mail services and Web servers, it was these systems people who figured out how to do it on a minimal budget. All it required was technical know-how (UNIX is found in corporations rather than in small businesses) and any Intel 386 or better machine. Corporations had many surplus Intel 386 and 486 machines, perfectly good, but made obsolete by the growing requirements of Windows software. Thus a knowledgeable systems administrator could add Internet service to the corporate system by employing an old 386 machine or two, the free Linux operating system, and free e-mail software, such as sendmail, and free Internet server software, such as Apache.
A curious development was that chief information officers (CIOs) tended to be unaware that the Open Source software was actually running in their departments. The mission had been accomplished with a minimum of fuss, expense, and probably no meetings at all. Nor did Linux call attention to itself by crashing. Sendmail, Inc. has recently commercialized the sendmail product, and their method of finding customers for customization and support contracts is to cold-call large corporations where their Internet probing has told them that sendmail is running. Even at this late date they are surprised and amused by the number of CIOs who deny that they have any Open Source software running.
The use of free software and small desktop machines are insignificant incentives to corporations of a size to run UNIX systems, and these very corporations already had the most expensive resource on the premises: UNIX personnel. Corporations are also the best positioned to benefit from Open Source software because they are Internet- and intranet-oriented, and have more direct control over their data pipelines than do smaller businesses. It is no surprise that Open Source has made its greatest progress at this level, which is the natural home of the early adopter. It is interesting, however, to see that in this relatively well-funded environment it was the low cost of Open Source that enabled the innovators themselves to become the visionaries who put the early elements of the systems in place.
It is far easier to introduce a UNIX company to Open Source software primarily because the systems personnel are in place, and because Linux and BSD run many UNIX applications. Burlington Coat Factory, with 260-odd stores, 20,000 employees, and sales of nearly $2 billion, can serve as an example of a successful switch to Linux. The company was always in the forefront of UNIX use, deploying in 1986 and standardizing on TCP/IP the next year.
The objectives of Burlington’s conversion included the consolidation of servers, concentrating on large ones to serve applications and to replace the Sun servers; the placement of thin clients onto desktops; and the establishment of a pure system, to avoid mixing Windows NT with Linux. They wanted to use their old software where possible, to write some new applications, such as for a gift registry system, and to accommodate a heterogeneous collection of desktops in Receiving. Significant obstacles to an all-Linux system were the inventory scanners and the fact that several thousand point-of-sale systems could not be replaced until sufficiently depreciated. The original plan was to run Network Computers, but summer interns hired to analyze the job persuaded management to consider using low-level PCs and Linux for the job (another example of the pollinating effect of young systems people).
Hardware had become a commodity by the time the buying decision was made, so the deciding criterion was a commitment by the vendor to support Linux. When Dell committed to shipping machines with complete Linux drivers for all the parts, they won the contract for an initial 250 machines. As for software, the thin clients run Web-based Oracle products, and use Applix for the office suite (it can read and write Excel and Word formats). The employees are happy because they have been able to keep their old applications. This helped fulfill the strategic advantage sought by Burlington: low costs with the least disruption. The chief operations officer who had to approve the spending was won over by the fact that the Linux installation represented the least change from the old system.
At this time it is far rarer to hear of a company of any size dropping Windows NT for a transition to Linux. On the other hand, there are plenty of NT corporations in which Linux has followed the path of stealthy entry into the Internet niche, with incremental servers running on Linux rather than on NT. Some of the innovators who placed these machines were Linux hobbyists at home, and some were former UNIX systems administrators who found their old skills were still useful. In the NT corporations, as with the UNIX corporations, the progress went from Internet services (mail and Web) to servers for files and printing.
Adoption of Linux for file and print servers in NT companies was helped by yet another piece of free software, Samba, which enables a Linux box to interface with an NT network. The effect is to disguise the Linux server as an NT server. The increasing ability of Linux servers to work with the file directory systems of multiple vendors has made them all the more useful in heterogeneous installations, frequently serving as the middleman in such systems. The original business proposition of Caldera in selling to corporations was that Linux was the glue that would hold everything together and make it work.
Chapter 6, "Linux vs. Windows," will deal with the question of Linux vs. NT in more detail.
In the presentations that Jim Geisman (http://www.moreshare.com) and I have made on the issues of Open Source cost and pricing, we have noted the keen interest of IT middle management in the initial cost question. For a manager with a strict headcount and equipment budget, the idea of free software licenses and smaller machines is an appealing argument. Jim and I don’t think these cost savings are the greatest that Open Source will achieve, although a manager at a large American-owned auto company recently told me that an evaluation in his division revealed that $50,000 UNIX workstations could be replaced by $10,000 Linux boxes. After looking at out-of-pocket costs, we’ll consider the benefits of higher reliability. Upper management in IT finds these benefits the most compelling financial reason to switch.
There are various ways to compare the prices of setting up networks under NT and Linux or BSD; any attempt at laying out the figures will be inexact because it is possible to buy Microsoft products at varying prices, and because filling in $0 for the Open Source side of the equation is not much of a comparison. We will see, however, that the NT packages frequently need supplemental Microsoft software to match the bundle that comes on a typical Open Source disk. Estimates for an NT network range from $15,000 for 50 users to $62,000 for 100 users. Some needed elements are priced at the Shop Microsoft Web site (http://shop.microsoft.com/) as follows:
Standard NT Server with 5 clients $809
20 for $699
Site Server (for Internet) with 5 clients $1,239
25 for $2,109
Exchange Server (for e-mail) with 5 clients $809
20 for $1,149
Thrifty IT operations businesspersons could forego the client licenses and simply use Win95 or Win98 machines (each about $120) connected to a single Workstation (about $160) for simple networking. Some corporate officers prefer Workstation for all machines because of its limited capacity to run computer games.
On the Linux or BSD side of the equation, each machine is peer networked; each is a server talking to servers. The distributions, which cost from nothing up to $80, include the software necessary for networking, Internet connectivity, and e-mail. Both the NT and Linux prices given here do not include productivity applications, which are extra. The Linux productivity packages are all cheaper than Office 2000 Professional, and some are free.
There is one commercial BSD server package, BSD OS, which has no free version. The pricing is higher than for Linux or other BSD packages: $995 for the binary-only 16-user package (source code $2,995). The 64-user binary-only version costs $1,995 (source code $3,995), and the unlimited-use package costs $2,995 (source code $4,995).
Whatever software-cost figures we project over however many users, less than 25 percent of the typical IT budget goes for software (and how much of that is for operating system licenses?). Similarly, hardware costs are less than 25 percent of the budget, and that figure is falling. On the other hand, the corporate personnel sitting at those machines carry a loaded cost of $50,000 to $75,000 on average, or anywhere from five to eight times the cost of the systems they use.
It is the stability and reliability of Linux (and BSD or any other UNIX) that gives a corporation the greatest cost savings. The frequent failure of NT systems (and of ordinary Windows systems) causes lost time and productivity. If we price this at $25 – $50 per hour per employee, and add in the cost of the systems personnel who must bring the systems up again, the cost of unreliability mounts rapidly. The larger the corporation, the more sensitive it is to questions of reliability, and the higher the value it places on high-availability systems. This has been a traditional strength of UNIX. With Linux, that reliability is available at a lower initial cost.
The stability sought by mainstream business users does not mean simply freedom from system crashes or lockups. Businesses are growing resistant to calls for upgrades; the slow penetration of Windows 98 and the light interest in Windows 2000 are an expression of this resistance. Besides this inertia, the growing publicity about continuing security holes in Windows and NT is only slowly beginning to worry businesses with external connections. The fact that these holes, although quickly repaired, are related to the interoperability of the Office package and its macros and thus give documents, spreadsheets, and e-mail — formerly thought to be inert — the power of executables, leaves conservative users wondering whether there isn’t an alternative, and whether Linux might be the answer.
Beyond hardware and software, Linux can make a strong case for cost savings in training, support, and administration because machines can be remotely configured and administered. Jon "maddog" Hall, executive director of the trade organization Linux International, reports that the University of São Paulo runs Windows, but on machines that can boot either Linux or Windows. When the Windows user reaches the point at which the machine will no longer boot Windows, an administrator can remotely boot the machine in Linux, and then use Linux to pull an undamaged copy of Windows across the wire to refresh the machine.
The real issues for an NT-based company are training and support. It may very well be that a Linux aficionado in IT has already begun the process of putting in Linux file and print servers because of their low cost, high capacity, and reliability. In that case, there is some support already on site. Even as Linux has been sliding and creeping into corporations — recently reported by International Data Corporation (IDC, http://idc.com/) as the operating system on 17 percent of new servers — there has been an outcry that the great problem with Linux is lack of official support structures. As Chapter 4, "Available Products (and More to Come)," explains, there are numerous organizations, large and small, stepping into this gap. A large corporation is more likely to have personnel who have a UNIX background, and who are also cross-platform trained. Nevertheless, the rapid increase in corporate Linux is likely to cause a shortage of personnel who can use it.
Linux supporters believe the colleges will eventually turn out enough people with Linux experience, but the real test is yet to come. Some estimates say that 25 percent of all Internet servers will be Linux-based within a few years. Linux vendors are meeting the personnel crunch in two ways: Linux software vendors are rapidly improving the operating system’s ability to be as self-installing as possible, while hardware vendors are rapidly increasing the number of boxes sold with Linux preconfigured and running. In addition, they are supplying unitary machines such as stand-alone Web servers that can be plugged into the wall and the network, and that start running as soon as they are configured by using a simple GUI that asks for IP addresses.
As mentioned previously, the desire of corporations to interconnect large numbers of employees and amounts of data across large areas makes them natural beneficiaries of network software such as UNIX and Linux. Although not yet exclusive Linux users, major corporations such as Wells Fargo, Boeing, and Cisco, as well as NASA and the U.S. Postal Service are all making use of Linux. The auto manufacturer mentioned previously is running a Latin American plant on Linux, and retail chains such as Jay Jacobs are using Linux for point-of-sale machines across hundreds of stores. Cendant Corporation runs some 4,000 servers in its chain-hotel franchises (including Ramada Inn and Day’s Inn). Gannett Media Technologies International has stated that it set up its enterprise-wide asset-management servers in 1998 by building its own middleware to its Oracle 8 database. Gannett gives the cost as $300 per server as opposed to the $1,500 it had expected to spend on another solution. Corporate uses of Linux should grow as the operating system improves its capabilities in symmetrical multiprocessing (SMP) and scalability, projects that have gone to the forefront of the Linux development effort since the negative analyst comments on these capabilities in the first half of 1998.
Linux fits nicely into corporate plans to run applications across servers to thin clients. When Sun Microsystems announced its WebTone and DataTone approach to corporate data in 1997 — the idea that all corporate data would be on remote servers accessed not just over internal networks, but over the Web, so that portable thin clients could work anywhere as if at the office — the plan appeared to be a dream suitable for a large corporation. But Sun recently purchased Star Division, makers of StarOffice, the leading office suite in Germany (after Microsoft Office), and able to run on Solaris, OS/2, and Linux, on which it is most popular in the United States. The product can be run remotely to clients, so Sun is currently giving away the suite to individual users, and intending to sell the networked version, called StarPortal.
At the small end of the scale, all the benefits of Open Source apply, and the low cost may be even more welcome. The fact is, however, that many small businesses do not have the UNIX-type expertise to install and administer these systems. As a consequence, the most likely path of entry into these companies is through the outside integrators who generally serve these companies by installing hardware, software, and providing training and other support. These integrators often serve a particular business niche or domain. Online directories of Linux integrators can be found at (http://bynari.com/BCG/director.html).
The smaller the business, the more it depends on the technical expertise of its owner. Linux hobbyists will want to apply their skills to their solo businesses, but the rest of us will need the sort of person who got us into computing in the first place: a savvy friend. There are Linux hobbyists who will help you get up to speed, and when you feel you are ready, you can bring Linux into your business operations. Local Linux groups host Linux Installfest events to help anyone who brings in a machine get up and running.
If you have worked out important systems like billing and automated correspondence on Windows, and believe this legacy prevents you from moving to an Open Source system, there is a new $299 product that makes possible a stepped transition. VMware, Inc. (http://www.vmware.inc) supplies a product that can be installed on a Linux or Windows NT/2000 machine, and will run other operating systems concurrently as virtual machines. This way you can take advantage of Linux (or BSD) stability to run the Linux applications you choose, and also toggle to a virtual Windows system that will run your Windows applications when you need them. The two systems can share files, and you can cut and paste between them. Through an OEM arrangement, VMware can supply a complete package, including both the Open Source operating systems and a choice of Windows operating system.
Once you feel assured that you will be able to hang onto your vital software that happens to run in Windows, it’s time to see what Windows software you can currently replace by adopting equivalents that run on Open Source systems. Because Linux did not attract widespread attention until 1999, there has not been time to match the wide selection of programs that have been written for Windows over the past 15 years or so. But Linux users, particularly if they are willing to look beyond programs with a graphical user interface (GUI) will still find a variety of programs they can use in the office.
Next chapter: Available Products and More to Come
Table of Contents
Copyright © 2000 by IDG Books Worldwide, Inc. Published under IDG Books Worldwide, Inc. Open Content License.