Linux Certification Survey Results Summary

By Alan
Mead

August 9, 1999

ABSTRACT: The author surveyed members of the on-line Linux
community regarding themselves, their companies' uses of Linux and
their perceptions of and attitudes towards certification. Among
considerable variation there were some consistencies, including:
(1) respondents forecast considerably more use of Linux in the next
two years; (2) most people were supportive of Linux certification
but will not certify immediately; (3) the number one reason for
certification was to support Linux; and (4) at this time (when
companies are less likely to pay, or reimburse costs) potential
certification seekers have significant sensitivity to the cost of
certification but other factors may be more influential.
Considerable detail and some conclusions and limitations are
discussed.

Introduction

This report summarizes the results of a Linux certification
survey conducted from Thursday July 22, 1999 to Sunday August 1,
1999. The purpose of the survey was to measure some of the
attitudes and desires of potential consumers of a Linux
certification exam. Although the survey originated as a vehicle to
answer some specific questions which arose as part of the author’s
pro bono work with the Linux Professional Institute (LPI), it
evolved into a more general inquiry into why one would choose Linux
certification and how one would choose.

This summary was written to help Linux certification programs
meet the needs of the Linux community.

At this point in a research article, I would cite previous
research. I am not aware of specific previous surveys or other
research into attitudes towards Linux certification. The sort of
questions asked in the present survey might be considered
“marketing-oriented” and kept proprietary. If anyone who is reading
this knows of some research I can access, I would very much
appreciate a pointer.

The survey itself can be viewed at: http://www.conan.ipat.com/~cert/index.cgi.

Participants

Although participants were solicited on emailing lists, the
majority of the participants appear to have been attracted from a
Linux Today article about the survey. Although the survey did not
ask for personal details, Linux Today has published details of its
readership. According to Linux Today, readers hail from 152
different countries with a majority from the United States.
Fifty-six percent access Linux Today from the .com, .net, and .edu
domains. The typical reader is thirty-two, a professional, and uses
and prefers Linux. Although 80% work in computer-related
occupations, only 13% work as system administrators with 22%
software developers or programmers, 6% network administrators, and
4% technical managers (the other 35% are not discussed).

Among the 435 people who responded to the present survey, a
majority (56%) lived and worked in the US with many of the rest
coming from countries with large English-speaking populations such
as Canada (8%), Australia (7%), and United Kingdom (4%). In all, 41
countries were represented. Nonetheless, the vast majority were
comfortable with English for certification testing, 94% reported
that they would have “no” problem with English as a language for
certification exams (of the 107 respondents outside the US, Canada,
Australia, and the UK, 82% reported no problem). Of other languages
requested, French and German topped the list.

Length of experience with Linux varied, the average person had
between 19 and 24 months of experience but 18% had 12 months or
less and 35% had 4 or more years experience. Respondents were
mostly employees (49%), although many were consultants (18%) and
students (14%). Only about a quarter listed computer administration
as their primary job duty with consultant (23%–obviously some
employees worked for consulting firms) and programmer (20%) being
popular. Thirty-nine percent felt that they worked in a “Linux
field” (developing for Linux, doing Linux consulting or training,
writing about Linux, etc.). Currently, most (59%) respondents
reported that a Linux certification was not required although 64%
thought that a Linux certification would be required in some way
within two years.

Not surprisingly, 75% had a favorable or very favorable view of
Linux as an enterprise solution (most of the rest had at least a
somewhat favorable view). Only 27% of respondents’ bosses were
reported to have favorable or very favorable views towards Linux
although 32% held opinions which were favorable but who required
education or had questions. This trend held more so for
participants’ clients (i.e., for those who had clients), only 18%
held favorable or very favorable opinions and 39% had questions or
needed education.

How are companies using Linux?

Interestingly, the general trend for respondents outside the US
was to use Linux more often than US organizations.

Certification perceptions

About 81% of respondents did not currently hold a certification.
Virtually none held a Linux or Unix certification. Of the existing
certification programs, the Red Hat program was the most widely
recognized (71% recognition), followed by the Linux Professional
Institute (52%), Sair (20%), and Digital Metrics (15%). Thirty-five
percent had visited the LPI home page and of those 77% who had an
opinion, 69% had very favorable or favorable opinions of LPI as
compared to other programs.

When asked about certification intentions, 21% intended to
certify in the next six months and 50% in the next year. Asked
about the LPI program specifically, 8% will certify immediately, 9%
in the next six months, 15% in the next year, 40% are waiting to
see costs, and 8% will never certify.

The two characteristics of certification programs most highly
rated by respondents were quality exams (90% favorable or very) and
supports Linux (87% favorable). Of least (but still considerable)
concern were holding not-for-profit status (62%) and avoiding
conflicts of interest (54%). Of intermediate importance were
ensuring test security (80%), minimizing certification costs (80%),
and international cooperation (71%).

When asked why they would choose Linux certification, 68% chose
“supporting Linux” as one of their top three reasons followed by
enhancing promotion possibilities (38%), making more money (37%),
becoming a Linux trainer (24%), and impressing future
coworkers/bosses (22%). First choices (alone) were only slightly
different: supporting Linux (35%), enhance promotion possibilities
(13%), make more money (9%), impress future coworkers/bosses (8%),
and become a Linux trainer (8%) (the other 27% had other reasons or
skipped the question).

Factors inhibiting participation in Linux certification include
concerns about cost (63% listed as one of top three reasons), lack
of business necessity (30%), and lack of Linux training (29%).
Fifteen percent reportedly disagreed with the very concept of
certification. In choosing among Linux certification programs,
first priority would be given to quality programs (79%), and then
programs that are affordable (53%).

Although certification is thought to be of greatest benefit to
newcomers who have not yet had opportunity to prove themselves in
the market, length of time using Linux had essentially no
relationship to propensity to certify in the next six months or
year.

Implementation and Costs

When asked for the maximum acceptable fee for certification
(i.e., all exams together), the average was $263 but the most
popular was $100. Eleven percent felt the certification should be
free and another 11% felt that a fee above $500 would be
acceptable.

When asked about different scenarios for test deployment, people
made these ratings:

It is worth pointing out that these are completely hypothetical
scenarios constructed to examine people’s perceptions more than as
realistic possibilities for any certification program.

That said, there is clearly a divergence of opinion between
those who feel that the costs should be as minimal as possible and
those who would not object to paying substantial sums for a high
quality Linux certification. In the scenarios ratings, the least
preferred scenario was the least expensive. However, this scenario
also listed the possibility that the certification program might be
unsuccessful so it would seem that costs, although very important,
are by no means the most important influence on people’s
perceptions.

Several comments indicated a need to balance more exams (which
are thus easier to prepare for) with greater travel costs (because
more trips might be needed). Some comments expressed an extreme
desire to avoid long (3 hour) exams. If the exams are initially
available in English, non-English natives may find longer exams
particularly fatiguing.

Other Findings

A number of comments expressed a concern for the validity of any
certification exams and related horror stories of certified
professionals who could not “tie their shoe”. Several comments
suggested that certification programs are flawed by their nature
because they are not valid and thus make less qualified candidates
look better than they would otherwise, often to non-Linux people
who cannot determine that their Linux skills are inferior.

Several people adamantly argued for free certification just as
Linux is free of cost.

Several people pointed out that the training needed for
certification was, in many ways, more of an issue than the
certification itself. This training was seen as relatively
expensive, scarce, and time-consuming.

Several people indicated that the choice to become certified in
Linux would be easier if their employer would bear the cost. Many
indicated that their employer would not.

Distribution differences, and even the need for a Linux (as
opposed to UNIX) certification, were discussed by a few
respondents. No one expressed an interest in a particular
distribution; everyone who mentioned this wanted the certification
to hold over all distributions.

Over 95% of respondents rated the server as “very responsive” or
“responsive” even though well over 3000 hits were served, mostly
over a three-day period, and the hardware used was modest with
virtually no tuning for performance as a web server. The survey was
implemented as a perl CGI script which assembled survey pages
“on-the-fly”. Apache 1.3 was used on a host running Linux kernel
2.0.36. The host was an Intel 486 66 MHz PC with 24MB RAM and an
IDE drive. Internet connectivity was through a 10mb ethernet
connection to a 128K ISDN router. Users in geographically distant
locales were more likely to rate the server as less
responsive–probably reflecting transmission times and network
congestion as much as actual server responsiveness.

Speculation and Conclusions

• If not ubiquitous, interest in Linux certification is still
  wide-spread and there exists considerable support for one or more
  certification programs provided they meet the needs of the Linux
  community and promote Linux.
• Linux certification programs should strive to have the highest
  possible quality. I recommend that certification programs pursue
  educational and psychological testing best practice, especially job
  analysis and test validation.
• If the present trend holds, interest in Linux will continue to
  gain momentum and Linux will become more common than rare (i.e.,
  most businesses will use Linux for something and more businesses
  will use Linux extensively). Linux certifications will become
  easier to obtain and more meaningful as Linux becomes common.
• Currently, the strongest drive for most people to obtain Linux
  certification is to support Linux. These people do not have
  organizational support nor bottomless wallets. To meet the needs of
  the entire Linux community, the program should be as available and
  as inexpensive as is possible while maintaining high quality.
• A strong secondary market for certification are Linux-oriented
  consultants who seem to view Linux certification as important
  documentation of their skills for potential clients.
• Affordable Linux training was perhaps more widely requested
  than certification per se. If high-quality Linux certification were
  free but the training was expensive and took weeks then
  certification might remain uncommon.
• Certification is not an area where a quality organization’s
  monopolization would be bad (balkanization would be bad).
  Therefore, the programs’ recognition rates are an important issue.
  At the present time, Red Hat’s certification is the most widely
  recognized (probably because Red Hat is by far the most widely-used
  Linux distribution and because Red Hat ran a high-traffic portal
  site until recently). However LPI is a close second even though
  they have yet to field an exam. Other programs were more obscure.
  Thus even for people who are interested in certification, more
  information is needed abut the programs available.
• Because LPI is distribution-neutral, LPI could be poised to
  become the preferred certification organization. For other programs
  (i.e., other than LPI or Red Hat) to succeed, they need to become
  more widely recognized. Because over 53% reported a preference for
  LPI, other programs’ successes may also require that the LPI and
  Red Hat programs to become discredited in some way; perhaps through
  heavy marketing or FUD.
• Since balkanization would be bad, another alternative for less
  well-known certification programs would be to seek to become part
  of the LPI project. For example, SAGE and LPI apparently share
  board members. Proper consolidation could be a healthy outcome for
  professionals seeking certification and employers as well as the
  individuals providing the certification.
• For certification programs yet to enter the market with a
  product, the urgency is not great (few will certify immediately).
  Quality should be the number one priority.
• Although certification in general system administration is
  needed, certification in specialized areas such as networking and
  security seems equally important and urgent.

• Exam costs should be minimized wherever possible. However, other
  considerations such as quality and accessability might ultimately
  be more important. Respondents apparently preferred exams between
  an hour and two hours and between three and six exams total.

• Certification programs should document the validity of their
  certifications.
• A respondent or two complained that the survey did not measure
  any perceptions about the content of a certification exam. It is
  true that many interesting questions went unasked for the sake of
  brevity. However, I personally feel that it is important to base
  the content of a certification exam on an analysis of the job
  rather than mere opinion. LPI has conducted a job analysis and SAGE
  has constructed job-element-like task descriptions. It is unclear
  how other programs created their content specifications.

Limitations, Restrictions, and Disclaimers

• All percentages were rounded to the nearest whole
  percentage.
• The raw data of this survey is available to any interested
  party under the Gnu Public License (GPL).
• This survey was “unscientific” in the sense that no attempt was
  made to randomly sample a population. Thus no attempt was made to
  present confidence intervals or use statistical tests to determine
  statistical significance. Due to the degree of participants’
  self-selection, the precise population sampled is unclear.
• The author offered (and offers) his pro bono services to all
  Linux certification programs. He has performed paid and pro bono
  work for LPI; this survey was not a paid activity.
• The survey, data collection, analysis, and report represent a
  good faith effort by the author to contribute to Linux
  professionally. However, in no way can the author, any organization
  which might seem affiliated with the author or the survey, or
  anyone else make any claim nor warranty regarding these results
  even including that the results are accurate for any purpose. These
  results are presented as is without any guarantee of any sort.
  Under no circumstances may anyone be held liable for anything as a
  direct or indirect result of this survey or its results.
• This document is copyright (c) 1999 by Alan D. Mead but is
  freely copyable provided appropriate attribution is given.
• Linux is a registered trademark of Linus Torvalds in the USA
  and other countries; other trademarks are held by their
  owners.