Linux vs. Windows 2000:
A Level Playing Field in 2000-2001
We believe that Windows 2000 and Linux will be equally matched by
the end of 2003. In part, it will be based upon expected changes resulting from the
Internet. At the same time, the open computing philosophy results in multiple suppliers
for a key computing component: the operating system. We believe that the decision will be
based upon each company’s analysis of the following criteria: compatibility, support,
applications, performance, scalability and ports, reliability.
Windows vs. Linux is not a simple battle between operating systems. The supremacy of an
operating system occurs at a critical juncture in computing. Information appliances will
emerge. The Internet will continue to expand. The need for standards and universal
interconnectivity is necessary. Can an infrastructure based upon proprietary solutions
lead to a solution?
In effect, a new computing generation will emerge over the next five years: the virtual
generation. It will be based upon an open, workflow application architecture. In the past,
when each new generation emerged, a change of leadership occurred. First the transition
from the mainframe vendors to ISVs. Later, from the ISVs to PC vendors. If history repeats
itself, there will be new vendors or sources of technology resulting from this transition
to a more open environment. Key to this transition will be the movement from proprietary
infrastructures to open ones.
In part, this transition will occur because of the Internet’s success. It has
shown that a computing community rather than one company can produce a standard
environment. There is no "owner" of the Internet, only of its content. Within an
open infrastructure, no one would own the operating system or communication protocols;
companies would own the application.
While standards could address and provide a solution to the issues identified above.
They have succeeded within the communications environment. However, it is unlikely they
will succeed within the operating system environment. Attempts at defining a standard
operating system command language failed in the 1980’s. Attempts at standardizing
Unix have failed. Why? Because each company represented on the standards body wants to
protect its product. A de facto, proprietary standard could emerge, but the
spectrum of hardware platforms makes that infeasible.
Currently, there are three potential players for operating system/environment
dominance: Java, Linux and Microsoft Windows (NT, 98, CE). The Java Virtual Machine can
run on multiple OS platforms. We will, focus, therefore on Microsoft Windows 2000 and
Linux, the two contenders through the early part of the next decade.
Linux is the contender with a fraction of the operating systems market. In the past two
years, it has attracted considerable attention. Linux is based upon the concept of open
source software – source code that is available to all that are interested.
Consequently, it has no owner other than the community that created it. Unlike proprietary
operating systems, there is no us vs. them – only us. Everyone
can be involved in the growth, design, implementation and testing of Linux. This community
spirit pervades the Linux and open software world. While detailed statistics are
unavailable, the estimated 1000 active developers in top business and academic locations
are ready to add a hand. As new demands are placed on the system, it evolves. These
changes in Linux reflect its users needs rather than a market analysis. Most of the effort
in the development of the latest version, V2.2, went into a major rewrite of the kernel to
accommodate multi-processor hardware platforms.
As commercial users begin to adopt the operating system, we expect to find areas that
are important to them. The new, needed functional capability will be added to the system
either from the users themselves of from distributors targeting this market.
If Linux succeeds, it will, partially be indebted to Microsoft. Microsoft expects to
release Windows 2000 in late 1999 for both the server and the desktop. A customer can
implement only a server-based system. Without the client, many of the operating
system’s new features will be lost. Updating both the server and client is a massive
change in the corporate computing environment, greater than the introduction of Windows
3.0 or Windows NT. Early polls by IDC indicate that 30% of IT organizations will do
nothing: they will keep Windows NT 4.0. However, many organizations, requiring a better
operating system will reevaluate their operating system options.
Overcoming the momentum of Windows NT/2000 is a formidable task. If Linux is to
continue its growth, current Windows NT users must first the current and future costs of
the transition to Windows 2000. Then, they may reevaluate their operating system options.
Initially this reevaluation will occur for specific applications, such as the Internet.
Then, with demonstrated success in one application domain, the new should gain wider
acceptance.
We believe that the decision will be based upon each company’s analysis of the
following criteria:
- Compatibility
- Support
- Applications
- Scalability and Ports
- Reliability
- Performance
The latter two factors can only be assessed after the final releases of the operating
systems are available. The remaining criteria will provide an indication of the
positioning of Windows 2000 vs. Linux 2,2.
Compatibility
There will be a software compatibility gap in Windows 2000 over Windows NT4.0.
Microsoft has publicly stated that Windows 2000 will focus on reliability, not
compatibility with applications running on Windows NT 4.0. These incompatibilities will
occur in both software and hardware.
The only indication we have of software compatibility is the Beta 2 release. A survey
of Beta 2 users estimate that 40% to 50% of Windows NT 4.0 programs would run under
Windows 2000. Companies developing their own NT 4.0 applications might have a more
difficult time, unless they were careful about following programming standards.
Unlike Windows NT, with a history of incompatibilities beginning with Windows NT 3.0,
Linux has kept backward compatibility as a design requirement. Applications running on
older versions do not become obsolete with a new release of the system. Even with a major
update as occurred in the release of version 2.2. Major changes to the kernel were
implemented. Yet, older applications continued to run.
The remedies for the compatibility issue are simple in concept, but may prove
expensive. For applications that are licensed, it will be necessary to obtain one that is
compatible with Windows 2000. For applications developed by companies, there is only one
choice: modify or reimplement the application.
Comparability issues also apply to hardware. Windows 2000 will only run on some
computers. This may not be a serious problem with servers, where fewer, more standard
servers, is the norm. Because many of the new features of Windows 2000, server edition
require Windows 2000 on the desktop as well. If the installed base of client PCs is
incompatible, it must be replaced. The magnitude of this is observed in the same IDC study
of operating systems, where 48% of the respondents planned to upgrade from computers
running Windows 3.1 or older.
The cost of migration depends upon the ubiquity of the applications, the cost of new
hardware and training. With an additional 18% of the IDC study group upgrading from
Windows 95, we foresee major migration expenses within many companies planning to adopt a
Windows 2000 solution. With inter-version compatibility as well as compatibility with
Windows clients, a Linux server solution would produce a favorable price comparison.
With lower powered development tools, the initial cost of application reimplementation
is higher under Linux. However, with a history of backwards compatibility issues as far
back as Windows NT 3.0, and potential compatibility issued with an IA64 version of Windows
2000, the prudent choice may be to move to Linux.
Faced with lower hardware costs, higher software costs and equal training costs, for
both operating systems, we believe companies will reevaluate their operating system
options. The difference will, in part, be based upon the strategy of the user
organization. Different costs will be occur whether they developed their own applications
or if acquired them. Given this slim line of demarcation, we believe that, from a
compatibility perspective, the edge goes to Linux.
Support
Support can be examined from two perspectives: resolving problems in the installation
and use of the operating system and correcting errors. Microsoft, with its proprietary OS
solution provides limited, free support in addressing installation issues for the desktop
product in the first 90 days. For server issues, there is no free period. The cost is $195
an issue, while desktop problems only cost $95 per issue.
For corporations deciding to use Linux, more formal support processes are necessary.
The major distributors of Linux provide active support. Unlike Windows NT, where there is
only Microsoft, each distributor provides support in a timely fashion. In fact, this
support may be a differentiating factor between distributors.
Distributor support is only the tip of the iceberg. With the increase in the number of
Linux users, a wide range of Use groups have sprung up. There are almost 300 online groups
discussing Linux. Because many existing Linux users are technically oriented, they offer
useful support of many problems. In few of the groups with which I have examined, most
focus on adding to the operating system or assisting other members of the Linux community.
This support network is so powerful that it received the Infoworld support organization
of the year award in 1997.
Another support option, unavailable with Windows NT/2000 is available. With the source
code available, a company could fix a bug in-house. By distributing the correction, the
company would increase the stability of Linux. As more science departments use Linux in
their operating systems classes increases, the number of Linux systems programmers will
become common. In the long run, this may become the way errors are corrected.
With the many avenues of support, the clear winner is Linux. The competition between
distributors should keep the support costs below that required for Windows NT/2000.
Applications
Without question, Windows NT/2000 owes its current and future dominance to the
thousands of applications designed to work with the operating system. Because of the
competition, these packages are cost effective and those that are successful are of very
high quality. This considerable investment in application software by Microsoft’s
customers has ensured the dominance of Windows. This may change. With potential questions
about the compatibility between Windows NT 4.0, Windows 2000 and the future version for
the Intel IA64 may reduce this advantage.
While a number of applications exist on the Linux platform, they are not well known.
The best known are server-based programs, not desktop applications. This has been
primarily due to the command language based user interface. Now, two major GUI interfaces
have emerged: KDE and Gnome. As each interface gains popularity, more desktop applications
will emerge. While only one will survive, it is too early to identify the winner.
The number of applications running on Linux is growing. Support from IBM, major data
base vendors, HP and others making available addition tools and software to perform many
of the applications required of a server today. Computer Associates, Hewlett Packard,
Intel, Informix, IBM, Oracle, SAP and Sybase have all made investments in porting their
product lines to Linux. These products are supportive of the place Linux is expected to
play at the server level.
On the desktop, only Corel, which plans to become a distributor of the Linux operating
system, has begun to release its office products. Its initial offering, WordPerfect 8 for
Linux, was released in 1998. By the end of 1999, a full office suite should be available.
With this release, the major, desktop, commercial infrastructure software will exist.
While porting Windows based applications may not be necessary in the future. (The WINE
project should make it possible to move Win32 code quickly and easily to Linux), tools are
necessary to develop Linux native applications from current, popular Windows applications.
Inprise has announced that it is porting JBuilder to Linux. If its previous strategies are
continued, we expect to see a similar announcement for C++Builder and Delphi within six to
nine months. This will be critical. First, it will provide Linux with a well-known visual
development environment. Secondly, it will make it easier to port other applications
developed in this environment to Linux.
Microsoft has demonstrated with Windows NT 3.0 that an easy to use development
environment is critical for the acceptance of a new operating system. By programming to a
class library, rather than APIs, a more portable development environment exists.
Microsoft, with its Visual C++ development environment provided the developer community an
easy way to upgrade Windows 3.1 applications to Windows NT 3.0. Once the release of
development environments is achieved, we expect to see the rapid increase of new products
for the desktop and server appearing on Linux.
Scalability and Ports
We usually define scalability in terms of moving to a larger, more powerful hardware
platform: scaling up. Over the next decade, however, scaling down will
become necessary. As computers become smaller, it has become necessary to move the
software environment to smaller and smaller footprints.
With a range of platforms as different as embedded systems (ELKS) on the bottom end and
mainframes on the top, Linux ports appear to address this issue. Many of these ports are
still in progress. A number of ports have been made, primarily to workstations. This
provides the user with a wide spectrum of platforms from which to choose. The Open Source
philosophy is responsible for the variety of supported or projected hardware platforms.
The decision to port to a system is not constrained by financial requirements
The introduction of Intel's IA64 (Merced) will not be x86 compatible. This will be the
first time Intel has produced a processor that is not compatible with previous ones. From
the operating system perspective, it will require porting the operating system and
compilers to the new architecture. Microsoft and VA Research (work supported by Intel) are
working on a version for this processor. Porting Linux to other platforms is an active
activity performed in many Linux programs with no loss of compatibility. Microsoft has
only ported Windows NT/2000 to the Compaq (DEC) Alpha processor. With many companies
wanting to use this processor, a delay in either the VA Research or Microsoft could
provide a significant opportunity.
Organization
As a single company, Microsoft centralizes all contacts between the user community and
the developers. As such, it provides a concrete organization with which users can
identify. As one developer once said: "Microsoft throws its multi-billion dollar
weight behind NT."
Linux, is the complete opposite. Yes, each of the major distributors is a vocal
supporter of the operating system. Yes, there are a number of Linux user groups. Yes,
there are Web sites that keep all informed of the current state of the myriad activities
surrounding Linux. Nevertheless, a central vendor Linux organization: no. To increase
credibility, some type of organization will be necessary.
Further, each Linux distributor specializes in the version for one hardware platform.
Currently, the major distributors have addressed the largest potential market: Intel x86
environment. If supported versions, compatible with the Sun and Alpha platforms, existed,
it would significantly improve the position of Linux.
Action Items
As is shown in the summary table below, Linux and Microsoft NT/2000 will start to play
on a level field by the end of 1999. If the Linux environment continues to grow, it will
have achieved a position whereby it could achieve a majority penetration by late 2002.
| |
MS Windows NT/2000 |
Linux |
| Compatibility |
2.5 |
3 |
| Support |
2 |
4 |
| Applications |
4 |
1 |
| Scalability and Ports |
2 |
3 |
| Reliability |
- |
- |
| Performance |
- |
- |
Total |
10.5 |
11 |
Table 1
Ranking of Factors
Vendors – Opportunities of this magnitude occur about every ten years. If Linux
achieves a significant market share over the next five years, a massive market for
applications will emerge. Develop plans for the port of your programs to Linux. Vendors
having products that are infrastructure based and unsupported under Linux should begin the
implementation of a Linux version.
Users - Begin the process of tracking and evaluating Linux. We recommend that you use
one of the distributor-based products rather than load a basic system. Each distribution
will have a GUI, applications and an easy installation package. The evaluation should
encompass both server and desktop situations. If the performance and stability of the
operating system meet your requirements, determine a conversion strategy. . Determine the
essential applications.
You should be able to obtain a reasonable idea of the Windows 2000 functional
capability by evaluating Beta 3. The final product will allow you to compare it with
Linux.
Both products will require retraining administrators. To maximize the payback of Linux,
having Linux savvy personnel available will pay dividends in the future. For both
products, we recommend that training begin as early as possible.
Entire contents (C) 1999 by Integrated Business Information Systems Ltd..
(IBIS, www.ibisl.com) All rights reserved. Reproduction of this publication in any form
without prior written permission is forbidden. The information contained herein has been
obtained from sources believed to be reliable. IBIS disclaims all warranties as to the
accuracy, completeness or adequacy of such information. IBIS shall have no liability for
errors, omissions or inadequacies in the information contained herein or for
interpretations thereof. The reader assumes sole responsibility for the selection of these
materials to achieve its intended results. The opinions expressed herein are subject to
change without notice. |