Sources
1. The Problem
The Linux Operating System has many advantages, particularly
advantages related to the control it gives its users over their
computing environments, but Linux does have one disadvantage: There
is no way of knowing how many people might be using it.
Not knowing how many users an operating system has causes many
problems, especially for companies who work in the Linux market. For
instance, companies cannot plan properly without knowing if they are
dealing with thousands or millions of potential customers. Red Hat
Software addresses this problem. This paper is compiled by Red Hat
Software for Red Hat's use. We are pleased to publish it for your use,
but we make absolutely no assurances about its accuracy or usefulness
for any purpose.
Notice that the terms "Linux users" and "Linux machines" appear
interchangeably because we estimate that for every computer that has
more than one user, there is at least one user who is running more
than one Linux machine. In fact, there may be either many more Linux
users than machines or more Linux-based machines than users, we simply
have no way of knowing.
2. Use of This Paper
This paper is our attempt to examine as much what is currently
known about the size of the Linux market. Our information is based on
estimates that are drawn from the most reliable sources that we can
find. Having said that, it should be noted that these figures are
estimates. If you feel that you have better data than that which is
shown here, we'd be thrilled to hear it. You are welcome to use this
paper for any purpose you see fit provided you:
- Acknowledge Red Hat's contribution
- Refer to the numbers found here as "estimates"
- Include the date of this estimate as shown above
3. Linux Use Data Sources We Did Not Use
We originally collected this data for our own marketing purposes.
Besides the sources of data we eventually chose, there are several
others that are certainly valid and useful but did not meet our needs.
This is not to be taken as any form of comment on the validity of that
work, and if you wish to prepare your own estimates, you will want to
choose from all available data.
a. Internet-Based Linux Counters
The most well known of these is the Linux Counter, a voluntary
registration system managed by Harald Tveit Alvestrand. More
information can be found at http://counter.li.org/
b. Sales of Linux CD-ROMs
While this does not give an answer itself, it does corroborate some
of the other data we do use. Its strength is primarily that it
indicates that there is a healthy and growing market for Linux
products. The problem with this data is that we do not know if each
CD-ROM is being used to install many Linux computers, or conversely if
most CD-ROMS are being sold as updates to users who are already running
Linux from a previous installation.
Our estimates based on our research during the year was that
US-based suppliers of Linux CD-Rom products sold 750,000 units in 1997.
This compares to our estimate of 450,000 units sold in 1996.
While US suppliers represented (in our opinion) the majority of the
Linux products shipped during the year there were significant numbers
of non-US-based CD-ROMs manufactured and sold in countries including
Germany, Japan, France, the UK, Taiwan, and China.
c. Downloads by FTP
One of the great mysteries of Linux usage is that there is no way
to know the number of Linux users who download the product by anonymous
FTP. These users range from students and users in developing countries
whose primary interest in Linux is its low cost to advanced engineering
groups who prefer to download these tools to guarantee that they are
using the very latest versions of all the tools and utilities.
For example, off of Red Hat Software's FTP site alone, there were
100,000 copies of Linux downloaded in the last 12 months. These
included 382 downloads (that we know of) to mirror sites. These sites
allow further local anonymous downloads.
d. Internet based RSA DES Challenge II
In order to try to beat The RSA DES Challenge II, distributed.net is
using computers' idle processing time all over the world. The
statistics recording which OS the volunteers' computers were running
are using is interesting if not very conclusive of anything. Thanks to
John Winters for pointing this one out to us. More info can be had at:
http://desstats.distributed.net/oslist.html
http://www.distributed.net/des/
Win32 (95/NT) 111,788,489
MacOS 48,878,795
Linux 29,834,793
Solaris 21,619,282
4. Data Sources We Have Used:
Magazine Surveys
Because of the problems of counting numbers of copies of Linux
distributed as shown above, we chose instead to base our estimates on
counting the number of Linux users against the number of UNIX users.
While far from an exact science, the number of UNIX users has been
tracked and estimated by more credible organizations using more
sophisticated techniques than we are pretending to use.
a. Linux Journal(s)
While the rapid growth in the distribution and paid subscription of
the Linux Journal over the last two years (see their web site at
http://www.ssc.com) indicates a healthy growing user base for the Linux
OS, its total circulation serves only as a minimum base number of
users, in the US.
- 1995: 10,000
- 1996: 25,000
- 1997: 45,000
- 1998: 57,000
Note the biggest growth in Linux magazine subscriptions is outside
of the US. In 1997, no less than five (that we know of) monthly Linux
magazines were founded in Japan, Poland, Germany, Yugoslavia, and the
UK.
b. UNIX Review, now Performance Computing Magazine
(circulation 90,000+), a US-based Miller Freeman publication,
acknowledged that in a late 1995 survey that 10% of their readers used
Linux. In their 1997 survey, the percentage of Linux users was 26%.
c. iX Magazine
(circulation 35,000+), a leading German UNIX magazine, found in a
recent 1996 survey that 34% of their readers used Linux. Today they
acknowledge that Linux is used by 45% of their readers, making it
-the- most popular OS used by their readers.
d. Sys Admin
(circulation 25,000+), a Miller Freeman publication serving UNIX
system administrators, found in a late 1996 survey that 21% of their
readers were using Linux. Today the survey reveals that 34% of their
readers use Linux.
e. Dr. Dobbs
(Circulation 155,000), is another Miller Freeman magazine with a
large international readership. Dr. Dobbs hired Wilson Research Group
of San Carlos CA, to conduct a significant survey of the UNIX OS usage
among their readers and found that of those using UNIX, about 43% were
using Linux.
5. Datapro's "1997 International User Ratings Survey of UNIX and NT"
Linux has jumped from being the 7th most commonly installed version
of UNIX in the survey sample to being the 4th most commonly installed
version of UNIX in just 12 months. Linux is, according to Datapro,
trailing only Solaris, HP/UX, and IBM's AIX on a worldwide usage by
"managers and directors of IS working for large organizations." In
several major world markets, specifically Germany, Brazil, and
Australia, Linux is installed as often if not more often than -any-
other UNIX category OS in large enterprises.
Their data indicates that NT is not making inroads into UNIX sites
and that NT is popular as a replacement for Dos/Windows and Novell
Netware installations.
Linux rates first in overall satisfaction in Datapro's ratings
survey of all the major UNIX OSes and NT.
While this survey does show a rapid increase in the use of Linux
in the corporate environment, 14% of all the sites surveyed used Linux,
the survey is limited as a guide to the total numbers of Linux users.
Linux use follows the PC model, where the users are making their own
purchasing decisions. There has been little effort to date to sell
Linux as an enterprise solution to the type of corporate managers
surveyed here.
For anyone interested in "Linux in the enterprise," Datapro's
study is filled with fascinating facts, figures, and analyses. While
pricey at $1,000 per copy, I recommend it for anyone looking for
serious data on the commercial market of Linux products and services.
6. Linux Users as a Percentage of UNIX Users
For all the hype other OSes are receiving, UNIX use is still
growing at a 10% annual rate, and the current UNIX industry figures are
that there are 6,700,000 active UNIX users. Taking the above
publications as representative of this industry, we come up with a
population of Linux users of between 26% and 45% of the total of UNIX
users, or between 1,742,000 and 3,015,000. This compares to last
year's numbers of 600,000 and 2,000,000 respectively.
Dr. Dobbs' survey (see above) is also very enlightening on this
subject. 42.5% of their UNIX using readers responded that "they would
be developing applications for Linux in the following 12 months"
compared to 12.3% who said that they would be developing applications
on SCO UNIX.
However when the question was changed to inquire "which UNIX OS
would you or your company purchase in the next 12 months," the
percentage of Linux users fell to 34.9% and SCO buyers grew to 12.7%
This is evidence of the hidden usage phenomenon that understates
the actual market share the Linux OS enjoys.
7. Red Hat Software's Customer Data
Our own experience selling and supporting Linux has highlighted an
important fact that bears on the above numbers. It is our experience
that the majority of our customers have not used UNIX previously.
Last year our registration system survey indicated that 56.2% of Linux
users had not previously used UNIX; conversely only 43.8% have had
previous experience with UNIX. This year, the numbers showed (and yes
we find this hard to believe ourselves) that 81% of those registering
their Red Hat Linux products had not used UNIX, while 19% had.
We are somewhat suspicious of the rapid change in this ratio and
speculate that the change is a result of other factors, such as the
fact that our UNIX-using customers are not registering their new copies
of Red Hat Linux or that the UNIX-using customers do not need the free
support that registration qualifies them for.
So while we are willing to concede some growth to the ratio of
non-UNIX users over UNIX users, we will use a 60/40 ratio for the
purpose of our low estimates and a 70/30 ratio for the purposes of our
high ratios.
Non-UNIX users would have had no reason to read or subscribe to any
of the UNIX magazines listed above. This would mean that the above
estimates of Linux users among the total population of UNIX users are
only 40% of the actual total.
8. The Low and High Range of Linux users
Total number of UNIX users using Linux plus the number of users
represented by the ratio of non-UNIX users among Red Hat Software's
customer base.
Low estimates: 1,742,000 + 2,613,000 = 4,355,000
High estimates: 3,015,000 + 7,035,000 = 10,050,000
These numbers of course do not include the users who have
downloaded Linux at no cost, and the Linux users in countries and
markets with whom we do not currently have much contact.
9. Red Hat's Official March 1998 Estimate of Linux Users
In summary, based on the above methodology and our own rough feel
for these numbers, we estimate, to within a margin of error of +/-25%,
that there are between 5,000,000 and 10,500,000 active Linux users.
It is interesting to compare this figure to our previous estimates
of the numbers of Linux users. Comparing our new estimates to our old
ones, we find that, in the last 12 months, the increase in the number
of new Linux users is:
End of 1993 100,000
1994 500,000
1995 1,500,000
1996 3,500,000
March 98 7,500,000
10. Sources and Disclaimer
Our thanks to all these publications without whose support this paper
would have been much less complete. All errors and misinterpretations
are mine alone.
You can find this document, and more about Red Hat, online at:
Red Hat Software, Inc.: http://www.redhat.com
Other References
Linux Journal: http://www.ssc.com/
Performance Computing (formerly UNIX Review): http://www.unixreview.com/
iX Magazine: http://www.heise.de/ix/
Sys Admin: http://www.samag.com/
Dr. Dobbs: http://www.ddj.com/
Datapro: http://www.datapro.com/
Copyright 1998 Red Hat Software, Inc. Please send email to
counting@redhat.com if you have any conflicting or corroborating data.