Author: JT Smith
Is there a chance that Microsoft, proprietary Unix vendors and the Open Source community might stop swearing at each other and actually begin providing
potential customers with usable information? Judging from the news of the last
week or two, it hardly seems possible. However, both Microsoft and the Linux
community are beginning to use cost and performance arguments that can
help their potential customers make intelligent choices. It might even
help repair the tattered reputations of a few vendors.
The scare tactics and disinformation campaigns of Microsoft, Sun,
and even some Open Source advocates have become boring and tiresome as well
as shrill and incredible. Sun is still recovering from its crude disinformation campaign targeting Linux users. John
Stenbit, the bluff, burly chief information officer for the U.S.
Department of Defense had to
personally swat down a whispering campaign by Microsoft that
claimed Stenbit’s flirtation with Open Source posed a threat to
national security. Stenbit, the former chief of TRW’s systems integration
business, has been advocating a
highly decentralized network-centric systems architecture [link is PDF]
that doesn’t sound much like Microsoft’s vision of the future, and he’s
considered a Unix guy in defense technology circles. Open Source advocates who
compare Microsoft to drug
pushers and corrupters
of foreign governments don’t do the credibility of Linux community
any favors, either. Is it no wonder that the many large customers are
so turned off? A senior IT manager I spoke to this week put it best when
he justified the freeze in his systems budget, saying, “We’re waiting for
our vendors to grow up.”
A couple of years ago, I suggested that the only way to re-energize
the market was to improve the debate about the merits of Linux vs. Windows.
It was past time for some honest comparisons of the type that
matters most to the average company. Vendors needed to explain the difference
between Linux and Microsoft systems in the only manner that means
anything to the average business manager: cost. Evangelists for Linux (or that
matter, Unix or Windows) should have to compare the cost — the total
cost — of owning a computing system based on Linux compared to the total
cost of owning an enterprise architecture based on a proprietary Unix or
Microsoft platform.
These “total cost of ownership” (TCO) calculations are a little more
complicated than comparing the cost of a package of a couple of Red Hat
CDs to the cost of thousands of Microsoft licenses. Even at
Microsoft prices, the cost of software is only a small part of
the total cost of managing a large IT environment. In recent years,
Gartner, KPMG, and Forrester have all issued reports that agree that hardware
and software combined represent less than 20% of the total cost of owning a
system.
Far more important are the costs of running the local
help desk, performing system audits and other administrative functions, and writing and supporting custom code. At most larger companies, these and
similar costs are far more significant than the price of the enterprise
software platform. So are the costs associated with improving or
degrading employee efficiency. A California-based financial services company I
worked with last year calculated it could justify the cost of a wholesale
upgrade of its enterprise architecture based on the fact the the proposed
system could shave five seconds off the time it took most of their 2,000 employees
to log on to their complex of system resources.
Reliable TCO assessments can be hard to do. Many firms attempting a
TCO analysis quickly give up. There claim there are simply too many
factors to account for. Many are difficult to calculate objectively. Most
evaluators lack the required expertise in accurate cost accounting and performance
benchmarking Even so, most companies can, if they try, assess
the financial impact of :
- Controlling the number of desktop and server images
- Reducing logon failures
- Reducing support costs
- Minimizing end-user training costs
- Lowering set-up costs
Enhancing flexibility by allowing users to “work from anywhere”
while accessing the system resources needed to do their jobs
Other factors are harder to quantify. For example, what is the
degree of risk associated with reducing unauthorized access to system
resources by outsiders? And what is the financial impact?
After a slow start, there’s a growing trend by vendors and
purchasers to perform TCO analyses when justifying competing IT platforms. Firms
like Compaq have introduced TCO as
an element of their sales and market education programs. Apple
enthusiasts are
promoting TCO analyses to justify retaining Macs in the workplace.
A variety of organizations have tried to use TCO analyses justify
the deployment of Linux throughout the enterprise.
Many of
these Open Source “analyses” are fairly primitive. Some do little more than
wonder aloud, “what possible combination of costs could Windows servers offer
a lower TCO than Linux or OpenBSD servers?” without an honest look
at those costs. Others have done their homework. Last year, Red Hat
sponsored a
devastating IDC study that quantified the TCO benefits of
running Linux compared to Unix. The study confirmed the findings of a
1999 Gartner Group study that suggested companies deploying a Linux
platform would enjoy a 20% overall cost advantage compared to firms running
Unix. Proprietary Unix vendors have been on the defensive ever since.
What about Microsoft? A study for the U.S. Department of Defense
conducted by MITRE [PDF] gave the lead to Linux over Microsoft NT in the back office because Linux was easier to manage, had more robust security features,
and supported remote monitoring and management more effectively. MITRE
found that each of these features resulted in measurable cost savings and
risk reduction. A
study for LinuxWorld.com was less credible, because the author cooked
the books a little. For example, he assumed that Microsoft-based systems
would require all parties to upgrade upgrade their equipment and software
every few years, but that, for some reason, Linux users somehow would
add nothing to their desktops and system managers would add
nothing to their back-end systems during the same period.
In any case, Microsoft’s product managers claim that much of the
reliability and support costs cited in these studies were specific to the
limitations in NT. Beginning last September, they began to use TCO justifications
to promote Windows 2000 as “An Operating System Even a CFO Would Love.” Microsoft has put together a “Rapid Economic Justification” SWAT team to promote sales based on quick returns on company investments in IT hardware and software. In
addition, the company has begun to publish a series of white papers and
customer profiles showing, the company claims, that Windows 2000 and its successors are generating dramatic improvements in reliability and significantly lowered
administrative and support costs.
Are these claims for real?
Compared to NT, they almost certainly are. Even in mixed NT/Unix
environments, it is not hard to believe the numbers in an upcoming customer study
that claims a company could reduce its internal domains, cut the number of
servers and reduce IT support staffing by 20% by moving off NT to a
different operating platform. Other Microsoft-sponsored TCO studies, such
as one focused on Allegis, merit a slightly more skeptical look.
According to that study, Allegis claims it could generate new product faster
running on Windows 2000 than on Unix. Perhaps.
In any case, such papers, and similar efforts by Open Source and
proprietary vendors represent a marked improvement over earlier, cruder efforts to draw cost comparisons between Microsoft and its
competition. Moreover, they advance the competitive discussion. Competing systems
vendors from the Microsoft, Linux and Unix communities need to move away from
from rancorous exchanges of disinformation and begin to focus on issues that
will advance the interests of their customers in a safe, secure,
inexpensive operating environment. If they compete hard enough, their products
might improve. And that wouldn’t be a bad thing, either.
