billwg wrote:
> Roy Schestowitz wrote:
> > |
> > | Everybody starts somewhere, however, and just as the cost and performance
> > | benefits of open-source operating systems such as Linux gradually
> > | changed people's perceptions, open-source routing could well follow
> > | along the same path.
> > `----
> >
> There is no real difference in commercial deployment of OSS software
> versus deployment of proprietary software in terms of support and
> promotion costs.
That's not entirely true. In fact, royalties are only about 1% of the
total cost of ownership.
There are fundamental philosophical differences between the way
Linux/UNIX/OSS does things and the way Microsoft Windows does things.
Windows has the advantage in terms of initial "day one" deployment.
MIcrosoft Software Installer makes it very easy for even the most
unqualified operator to install even the most sophisticated software.
A guy who can barely read english can install SQL server for an
American company from India, using Remote Access while the "Real"
admins are sleeping. Configuration is all based on graphical user
interfaces, which means that remote access requires a high bandwidth
connection, but it can be done. The Windows administrator believes
that everything should be done through the graphical user interface and
that only the application installation should be scripted, and that is
up to the software publisher. Since most Windows servers only run one
application, this isn't too much of a problem. Most Windows servers
don't run more than one server application because of the risk of
current or future possible "DLL Hell" where DLL versions needed by one
program are incompatible with DLL versions needed by another program.
The advantage of this approach is that you can have administrators with
very little specialized training who can work very cheaply, and they
can install what needs to be installed. More servers are needed, as
well as back-up servers to provide failover (because configuring a new
system takes so long), but since most TCO costs are based on "per
server" costs, Windows usually wins in terms of "low cost
installations". By keeping the TCO window short, to less than 2 years,
Windows can look very attractitve.
Linux administrators on the other hand, tend to like to script
everything, including the installation, typically, they will edit a few
configuration files and start the installations, often in the
background. This is because they want to be able to very quickly
rebuild systems in a minimal amount of time. Linux servers also tend
to run more applications on each server, Linux clustering technologies
make it possible to use multiple active servers which provides both
performance and redundancy. As a result, far fewer servers are needed,
and because each server has more software being installed and
configured, each Linux server has a higher "Install and Configure"
cost.
Routine maintenance is another matter. Windows administrators use the
graphical user interface to handle even the most routine weekly tasks,
including log rotation, restart of servers to clean up memory leaks,
and the weekly or biweekly reboot of each server. All of these
functions need to be done by hand. In addition, patch applications,
security patches, and other updates need to be reviewed before they are
applied, and usually need to be tested on "test" servers before they
can be safely applied to the production servers, especially if the
servers are running 3rd party software. It's ' pretty safe bet that if
the server is only running IIS or SQL Server, that the patches are
pretty safe, but if you are using any non-Microsoft products, it's very
wise to test first. As a result, the weekly costs of running these
servers is very high, especially when you have to perform these support
tasks on a substantial number of servers using these graphical user
interfaces.
Linux administrators, however, also script routine maintenance. They
use shell scripts and cron jobs to handle regular rotation of the
scripts. They use inetd and fork to make sure that there are always
servers ready to do work, but that processes don't run long enough for
memory leaks to steal all of the system memory. They use rc.d scripts
to restart services, often in a few milliseconds, without substantially
disrupting the service to customers. Because so much of this is
scripted, the administrators are able to support development, test, and
support efforts. Fewer Linux administrators are needed for each shift
for the same number of servers. As a result, the weekly costs per
server are lower. Remember also that we needed fewer servers because
each server could do more work.
> The difference lies in the use of unpaid volunteers
> to do actual code development and, in a more limited sense, QA testing.
Note the statement above. Linux administrators do a great deal of
scripting and support automation. If they have done their job well,
they often have bandwidth to support these additional efforts. Very
often, they focus their efforts on projects and activities which will
forward the business of their employer. Since their employer is in the
business of provding good insurance services, medical treatment,
shipping service, or other primary business, and not in the business of
selling software, the exchange of software between administrators is
looked on as a way to make everybody's job much easier.
This doesn't mean that there aren't companies who make profit selling
software for Linux. Quite the opposite, IBM, Oracle, SAP, Seibel,
Sybase, Computer Associates, and hundreds of other companies, large and
small, are now making substantial amounts of revenue from custotmers
who want their commercial software installed on Linux servers. Even
when Linux isn't the primary database repository, there are often
requirements to have the database client or Intersolve ODBC drivers
installed on Linux.
> OSS projects count on the continued good will and interest of the
> developers to ensure product continuity and rely on the notion that
> some new group of volunteers will pop up and shoulder the load in the
> event of the disappearance of the original developer or team.
That's why corporate sponsors often form nonprofit organizations and
make sure they are well funded by corporate vendor and user
contributions. The nonprofit organization is harder to railroad
through proxy votes. It is also easier to make sure that contributions
are accepted and shared, even when the contributions of one vendor or
author might conflict with the interests of another vendor or author.
For example, IBM created eclipse, but put it under the control of a
nonprofit organization. The organization quickly began taking
contributions from many of IBM's competitiors, including Borland, Sun,
and even BEA. Before long, many of the functions provided by WebSphere
but not included in Eclipse were being incorporated as Jakarta, Jboss,
and Struts. At the same time, BEA was porting many weblogic tool
plug-ins.
Ironically, OSS projects often facilitate the development of much more
commercial software because they help standardize the infrastructure
and architecture into a framework that others can work with.
Commercial software companies can often build on these frameworks and
standards.
> That gives OSS somewhat of a cost advantage over proprietary software,
> where the development is performed by paid personnel, but doesn't help
> at all when it comes to a need for guaranteed support and future
> development.
Again, this is offset by the support of corporate interests. Granted,
no one is going to become "Bill Gates Rich" coding, or even managing a
nonprofit organization that creates and maintains infrastructure
frameworks such as Eclipse, OpenOffice, or Linux, but often the
salaries and fringe benefits are comparable with those of comparable
rank in typical corporate organizations.
Many companies, like OEMs, actually don't like the idea of creating
another "Microsoft Monopoly". It's better to have an open
infrastructure like Linux and have both commercial and OSS
distributions like Novell SUSE and UBUNTU than to let Novell get
monopoly control of a Linux infrastructure element that becomes
"indespensible".
> Furthermore, it doesn't provide for any conventional
> promotional activity using advertising and sales calls. "No bucks, no
> Buck Rogers" was the early thoughts about manned space exploration and
Ironic isn't it, that NASA ended up generating more innovations and
technology in 10 years than most major corporations create in a
century. The transistor, the Jet, the modern computer, the integrated
circuit, teflon, temprapedic cushions, tang, and thousands of other new
technologies, were invented as part of the space program. I remember a
director of NASA once stating that if NASA had been able to patent
their innovations, and collect royalties on those innovations, that the
royalties would have funded BOTH manned space flights to Mars AND
interstellar exploration technologies. He also stated that if the
United States Government had been able to collect those royalties, we
would have no federal debt.
Not everyone things of their success in life entirely in terms of cold
hard cash. Mother Theresa had no personal wealth, and a very austere
lifestyle, but she was a success on HER terms. Martin Luther King was
never wealthy, and he got shot, but he was a success in terms of what
HE considered important. Mahatma Ghandi became the leader of one of
the largest countries in the world in terms of populations, but was
willing to die of starvation while wearing diapers, yet I think most
people would consider him one of the most "successful" men of the 21st
century.
The world is full of people who have gone above material needs on
Maslo's hierarchy of needs, to the point where self-expression and
making a difference have become more important than mere "cash in the
pocket". They could easily make more money, but they are more
interested in enhancing the lives of others.
> it applies equally to the idea of OSS, which is relegated to some kind
> of word of mouth and meager promotion effort due to lack of sufficient
> funds to do this effectively.
Linux is openly mentioned in ads by IBM, DELL, and HP in national
magazines, television ads, technical journals, and other national and
local media. True, Linus Torvalds didn't pay for it out of his own
pocket, but the plugs are worth millions of dollars. When you can
enroll someone else into funding your create humanitarian project,
that's real power.
> People who rely on proper function and continued product improvement
> are willing to pay for it. OSS cannot offer any guarantees.
Which is what makes Linux so extraordinary. OSS doesn't make any
promises about the future. There is what is there today. But what is
there today is driving Microsoft to make promises it hasn't been able
to keep. In 1992, Microsoft promised a "Better Unix than Unix" making
reference to SunOS 4.0 which featured X11 Windows, true preemptive
multitasking ,fully collaborative computing, clustering of desktop
machines into a "network is the mainframe" supercomputer, security that
exceeded the needs of most banks, insurance companies, and stock
brokers, performance and reliability that exceeded the needs of
companies like Dow Jones (a failure in the DJ news feed in 1987 caused
a stock market crash almost as bad as the one in 1929), utility
companies (even controlling nuclear reactors) and even the stock market
trading systems (many of the NYSE trading systems were Suns in 1992).
To this day, nearly 14 years later, Windows XP has still failed to come
anywhere near fulfilling that promise. They have YET to deliver a
system that provides the performance, reliability, security, stability,
flexibility, and performance of even SunOS, let alone Solaris 10, AIX
5.3, or SUSE Enterprise Linux Desktop 10.
|
|
