Open Source: Assessing the TCO Picture
Open source technology sounds like it’s free but it isn’t; like its commercial equivalent, it has to be implemented and supported. However, the total cost of ownership (TCO) has proven very positive for a number of Canadian organizations.
And with the big IT players stepping up to the plate, adoption no longer involves taking a wild swing.
Open source has created a paradigm shift in the way software is provided. In simple terms, open source software is code that is put in the public domain with a series of provisos: any party is allowed to use and distribute the software at no cost; but equally important, any party is free to improve and revise the software, with the stipulation that they make their source code for doing so publicly available. In other words, if you take advantage of contributions from the open source community in developing a product, you share your creation with the community as well. This collaborative approach has led to a rapid process of improvement for open source products, and a strong support community.
While this technology is new to many people, it’s been around for a while. Nigel Fortlage, vice-president of information technology at George H. Young International (GHY), a Winnipeg-based customs broker, first started experimenting with Linux eight years ago.
“A young fellow that we had working for us had Linux on his home machine,” recalls Fortlage. “He introduced me to Linux at that time.” GHY was then at the very early stages of building their IT infrastructure. They had no LAN, no e-mail, no Internet. Fortlage was concerned about the high costs of building these infrastructures from scratch. So they became early open source adopters almost by accident. “There were no restrictions to what we could look at, and we made what turned out to be a cutting edge decision,” says Fortlage.
Historically, Linux has often come into organizations through the back door. John Lathrop, president of Calgary-based Linux Leap and author of Linux in Small Business: a Practical Users’ Guide, describes a common scenario: “Typically they slide it into a company’s operation, in background areas such as file and print servers. Executives might not even know it’s there. When they see how it’s working, how cheap it is, and how little maintenance it requires, they start pushing quietly for a general adoption.” It’s no accident that Lathrop is based in Calgary; Linux is particularly popular in Western Canada, having been widely adopted by the energy sector.
Things have changed since the early adopter days. “Linux now has such high visibility,” says Lathrop. Indeed, the accidental experience of early adopters is being replaced by a mainstream push to open source. We’re no longer talking about obscure file and print servers either. Organizations of all sizes are doing serious cost and feasibility evaluations before taking the leap, and are concluding, for a variety of reasons, that the open source option is built on a solid base.
Linux and beyond
Much of the open source discussion is focused on Linux. As an operating system that can run on anything from a cheap PC to a mainframe, Linux is the virtual foundation upon which open source rests. It is touted by many experts as not only comparable, but vastly superior to commercial alternatives. Closely related to UNIX, it has the security and stability associated with UNIX operating systems sold by IBM, HP, and Sun. This makes Linux appropriate in the mission critical space – that is, in applications that require bulletproof reliability. It also means that support personnel with a UNIX background can easily make the transition to Linux. But the story gets better. Linux is also more efficient, and able to run on a smaller “footprint” than its UNIX counterparts, resulting in more modest hardware requirements.
Of course, you can’t run an enterprise on just the Linux platform; you have to have applications that run on Linux. This is why big vendor buy-in has been so critical. With big players such as SAP, Peoplesoft, and Oracle making their enterprise applications available on Linux, new options continue to open up. Combine this with the availability of hardware platforms and support from vendors such as IBM, HP, Sun, Dell, and you have a very solid option. So solid, in fact, that some pundits now predict that UNIX itself will be completely replaced by Linux over the next few years
Open source is not just about Linux; there is significant growth in the application space as well. Open source equivalents are available for a wide range of common commercial applications, and can be used as “drop in” replacements. Apache, for example, is a solid rival to commercial Web servers, preferred by many solely based on its technical merits. Open Office is a free office suite that includes word processor, spreadsheets, slide presentation, and draw capabilities. Mozilla is a popular open source browser. The list goes on, and it is growing.
Vendors in the open source community make a living not by selling licenses, but by selling support. Organizations can download Linux free in its native form, and receive support on an ad hoc basis, or download a supported version from companies like Red Hat or Novell. In the case of Red Hat, support includes a fairly sophisticated provisioning of automatic program updates and patches – a huge timesaver, according to Lathrop. The other part of the support offering is telephone support, which is modeled on what vendors provide for commercial software.
Fortlage moved from the native to the supported model when he got involved in enterprise applications. “I’d come from this world where we downloaded the open source product, got the pieces we needed, and tested it out,” he explains. “And other than our time, there was virtually no cost involved in being able to do that. But when we decided to take Linux to the enterprise, and take open source to the enterprise, that world had to change.”
The enterprise level
Given what’s at stake at the enterprise level, Fortlage and others look beyond the initial costs. Total cost of ownership, or TCO, is a loosely defined metric for what it will cost to operate a technology over a period of time, typically 3 to 5 years. A classic example is the cheaper PC that, because it is difficult to support, actually costs the company more over time than a model with a higher initial price. Because open source software is initially free, TCO is the only way we can compare it with the alternatives.
Vendors often use TCO as a marketing tool, and this can create confusion about what TCO is all about. According to Chris Pratt, IBM’s manager of e-services strategic initiatives, vendor TCO reports have to be read with a critical eye.
“The first questions you have to ask are what are the components of TCO as defined in the study? Who paid for the study? Where are they coining from?” he says. These studies can’t be generically applied; you have to determine which parts are relevant and which aren’t. For example, a study could talk about reduced staffing costs in an area where you have only one person that can’t be eliminated. “Read the studies carefully”, advises Pratt, “and compare them against what happens in your own business.”
A software platform like Linux interacts a great deal with other IT components, so TCO gets complicated. Through its various technical attributes, it affects TCO through its respective influence on hardware costs, support and training costs, and licensing costs of other software products. The following categories are often considered:
* Security: Linux operators report excellent resilience to viruses, hackers and other threats. The use of Linux by organizations such as Boeing and U.S. Department of Defense are testament to this. This reduces security staffing requirements, and the dependence on security infrastructures.
* Stability: Stable platforms mean less administration time spent on tweaking and re-booting systems, and that can translate to a lower headcount in large organizations. In a smaller company, it can free up staff to implement other cost-saving measures.
* Scalability: The ability of a platform to support growth into a larger infrastructure prevents expensive growing pains. This made a big difference for GHY when whey moved to a consolidated server platform, itself a very TCO-friendly project.
* Performance and overhead: A popular benefit of Linux is that it runs on a small footprint – that is, it makes modest demands on the hardware it runs on, reducing hardware costs and support costs.
Cutting edge in Calgary
Factors like these were key when the City of Calgary decided to migrate their proprietary UNIX platforms over to Linux. Dan Ryan, manager of desktop and infrastructure at the City, had to move quickly on the TCO issue. A little less than two years ago, he was handed a 10% across-the-board budget reduction. While looking for areas to cut, Ryan found that the city was spending $650,000 per year on maintenance contracts for 140 HP and Sun UNIX servers. After exploring various options, such as taking the servers off maintenance entirely (considered too risky), Ryan’s staff told him about Linux, and he wanted to hear more.
Like many organizations, the city was using its UNIX platforms to run mission-critical applications, so the due diligence was considerable. Their reliance on large vendors, and particularly their use of the Oracle platform, meant vendor buy-in \was essential. “For us this was the turning point: having companies like IBM and HP investing in it,” says Ryan. “But having Oracle certified on Red Hat was the absolute turning point for us. That told us that this thing was ready for prime time.”
The vendor that wound up conducting the proof of concept for Linux was, interestingly enough, HP, whose own proprietary UNIX stands to be widely replaced by Linux. The tests proved that this was a solid, stable platform. “These are as bulletproof as any UNTX environment,” says Ryan.
If there were any surprises, it was that performance expectations were not only met, but dramatically exceeded. Batch processing proved to be 200% to 500% faster than the UNIX equivalent. The city is now able to run applications on two central processing unit (CPU) machines that previously ran on five CPU machines. This means lower hardware costs, and lower hardware support costs. But the big savings are in software license fees. Because Oracle, like many other software vendors, licenses their product on a per-CPU basis, the city was able to translate the better per-CPU performance into significant savings, and a corresponding improvement in TCO.
Many organizations have also reduced costs by “downsizing” their applications from proprietary UNIX servers to a Microsoft platform running on Intel-based hardware. Because Intel-based machines are an open hardware platform (in that many operating systems can run on them), hardware is more competitively priced. Ryan considered this option, having a strong Microsoft background. But the changeover and retraining of staff would have been prohibitive. Conversely, the Linux skill sets were basically already there because of the staff’s UNIX expertise.
Other advantages
The need to implement large-scale budget cuts is by no means the only compelling event that brings Linux to the table. Winnipeg- based Loewen, a manufacturer of Douglas Fir doors and windows for the luxury market, developed a light Web presence a few years ago through a hosted provider. When their presence grew and they needed to implement a full-fledged e-commerce site, they decided to host it themselves.
Implementing this on a Linux platform was a natural; the Web server space is a big stronghold for Linux due to its security and stability features. In fact, high profile Web sites such as Google and Yahoo run on Linux. Dave Lehman, e-business manager for Loewen, noted that he doesn’t get hit by worms and viruses the way some of his colleagues do.
The final and largest frontier for open source is the desktop. Pundits are cautious about making predictions, but organizations are starting to run pilots, particularly in government sectors such as education. DHY and Loewen are starting to do pilots as well. Lehman points out that the vendors are making it easier. Novell, for instance, has announced that its own products will soon have a Linux option. IBM has also moved applications, such as Lotus Notes, to the Linux platform. If Microsoft were to make their applications available on Linux, this would undoubtedly tip the balance in many organizations.
Open source application environments provide an alternative at the desktop that organizations are also starting to consider. A leading example is Star Office, a supported version of Open Office distributed by Sun Microsystems. Lathrop believes that in a small office, such a transition wouldn’t be difficult. In larger organizations, however, there may be a number of dependencies in place that would make it more difficult, especially if there is a support organization. Moving an application from one platform doesn’t affect the productivity of individuals, but changing their desktop environment could mean making significant changes to their work environment. Change management costs for this kind of transition would be much more significant.
These caveats aside, the open source desktop option can stand up well to a TCO analysis. Chris Pratt at IBM points out that open source software architecture is based on a “build up” rather than a “tear down” model. This means that you put only the functionality you need on a server. Commercial software packages, on the other hand, tend to include bundles of features and applications, all of which have to be either removed or supported. According to Pratt, removing components isn’t always easy, and that is where the trouble begins. Available programs tend to get used, and if this use isn’t managed, you get security problems, and disruption due to poorly maintained products in the environment. Moving to open source could have the effect, therefore, of setting up a more secure and well- managed environment, removing a number of hidden costs inherent in the old model.
Another interesting aspect of open source, which suggests lots of opportunity, is that the openness of the products makes them easy to adapt to a company’s specific needs. DHY, for example, needed the ability to create PDF files at the desktop, but the cost of putting a PDF-generating client on each user machine was prohibitive. So using two open source products, they developed an innovative solution with only a half hour of programming work. Ghostwriter, an open source PDF generation tool, was combined with Samba, a file and print tool to create a PDF server. To create a PDF file, the user creates a file to be converted to PDF, and sends the print file to the server. Ghostwriter then converts the file, which is sent back to the originator as an e-mail attachment.
Open source is a natural for this kind of creativity. That factor alone could change the way organizations use software.
Jacob Stoller is principal of StollerStrategies, a Toronto-based consultancy focused on technology issues.
Copyright Society of Management Accountants of Canada Aug/Sep 2004