Beruflich Dokumente
Kultur Dokumente
EDUCATIONAL ENVIRONMENT
BY
GREGG FERRIE
Athabasca, Alberta
July, 2011
To the five pillars of my life: God, my wife, and my three children. Sometimes my focus gets
misdirected and without you life would be empty and without purpose, however you always
get me back on track. Vikki your steadfast support, love, strength and faith have seen me
through the most difficult aspects of my studies. Naomi, Daniel and Joel you have grown into
amazing and responsible adults and I couldn't imagine life without you. Thank you all for
School districts are under continual pressure to reduce budgets while providing state-of-the-
this with traditional Windows-based desktops and limited funding is becoming unsustainable
and most school districts are falling further behind in support of the curriculum. This essay
will provide a comprehensive and critical analysis of how inexpensive, energy-efficient and
sustainable and proven solution. As it is antithetical from the current standard, many school
applications. These desktops are energy inefficient and often poorly managed. They run
expensive commercial software applications which provide largely the same functionality as
their open-source counterparts. This essay will seek to quantify the educational, operational
and financial advantages in conjunction with the significant energy savings possible by
i
AKNOWLEDGEMENTS
I would like to express my sincere gratitude to the many people who helped me along this
journey including my wife Vikki and family, fellow students, the Computer of Computing
and Information Systems staff. Thank you for all the advice and support from Linda Gray,
my professors, particularly professor Lin, professor Dron, professor Graf and my essay
supervisor professor Huntrods. Most importantly, senior executive staff at School District
No. 63 (Saanich) Dr. Keven Elder and Joan Axford. Your support will always be remembered
and appreciated.
ii
TABLE OF CONTENTS
Significance................................................................................................................. 15
Assumptions................................................................................................................ 15
Limitations. ................................................................................................................. 16
CHAPTER IV - RESULTS................................................................................................. 43
System Requirements.................................................................................................. 63
iii
Energy Conservation and Diskless Clients. ................................................................ 67
Summary. .................................................................................................................... 73
Conclusion. ................................................................................................................. 84
REFERENCES ................................................................................................................... 85
iv
LIST OF TABLES
Table 3. Functional comparison between thin, diskless and thick clients .............................. 53
Table 4. Thick client versus diskless client boot process comparison .................................... 57
v
LIST OF FIGURES
vi
CHAPTER I
INTRODUCTION
and administration provided by the various provincial governments and territories. Although
there are some minor differences between the various jurisdictions, the challenges are
analogous and include providing equitable resources across a diverse student population to
facilitate student learning and assessment. In the province of British Columbia, the prime
goals of public education are: to develop intellectual development and critical thinking,
encourage and support human and social development while enabling career development
opportunities for the future [1]. Although these are lofty goals, the public education system
is faced with numerous and significant challenges which include: declining enrolment in
many districts, rising costs, salary and benefit increases, policy and regulation changes,
within the context of a very rapidly changing world. The latter two issues, in particular, are
influencing education ministries, policy makers, districts and teachers to rethink the
traditional role of the school and how students learning and achievement will be supported.
Until recently, education public education has been primarily delivered through
traditional instruction methods, where teachers provide the curriculum and lecture from the
front of the class. Students are required to listen, utilize rote learning and memorize
information from which they are challenged through quizzes, exams and other traditional
methods of assessment. Over the past 10 to 15 years there has been a gradual move by
ministries of education to offer students greater choice of curriculum via distributed learning:
1
utilizing computer-based technologies and learning management systems such as Blackboard
Teachers and schools are also being challenged to provide access to Internet-based
anytime, particularly external of traditional school hours. Synchronous with this paradigm,
educational ministries, school districts and schools is the rapid change of technology
including, but not limited to, the advent of the Internet, social networking, ubiquitous access
to resources, mobility, eLearning and a host of other influences [3]. The contemporary, under
30 generation, often called “digital natives” have grown up in a digital world, where
information, resources, access, multitasking are systemic to their experience [4]. There is
now recognition from educators that these students think differently, work differently and
Understanding this, the British Columbian Ministry of Education, along with other
jurisdictions, are rapidly moving to a 21st Century pedagogical model which considers
from any school who can offer the curriculum [6]. Although not exclusively, personalized
communication, social networking, digital resources, ePortfolios and other applications [7].
Ubiquitous access to these resources, both within and without of the confines of the
traditional school anytime is becoming predominant. Although the link between ubiquitous
access and environmental concerns appear to be diverse, there are threads of commonality
which is beginning to change the way educational technology is being viewed and delivered.
2
As students experience with connected activities is endemic to their experience, they
are also becoming increasingly aware and concerned about climate change and its impact on
the environment. Consequently there is increasing and persistent pressure to reduce energy
use and become more efficient in all aspects of public education. In British Columbia, school
districts are under extraordinary pressure to conserve energy and reduce green house gases
(GHG) and carbon dioxide (CO2) emissions. There is increasing demand to accomplish this
while maintaining modern technologies in support of student learning and achievement. This
topic is highly germane to today's public school system as budgets decrease, while
technologically astute students demand access to modern ICT-based learning resources with a
competing pressures teachers and schools are struggling to strike a balance which provides
The challenge districts are now facing is how to provide cost-effective, relevant,
current, sustainable, environmentally conscious and ubiquitous access to resources that work
for today’s educational outcomes and simultaneously anticipate the future. School district
technology plans are usually developed around a three to five year cycle to anticipate
Over the past eight to ten years, the current and most pervasive computing model is to
provide schools with three year old, off-lease, Microsoft Windows XP-based desktop
computers. These computers typically stay in the school environment until they are
completely obsolescent.
The Microsoft XP operating system was originally introduced in late 2001 and is the
overwhelming choice for desktop computers in K-12 schools. This has worked well for most
3
school districts as long as the Windows XP platform is supported and moderately current
desktop computers are available at low cost. According to w3schools, as of June 2011 the
Window XP operating system still maintains the overwhelming majority of desktop use
estimated to be 39.7%, while Windows 7 has gradually increased to 37.8%, Microsoft Vista
is at 6.7%, the Mac operating system is at 8.1% and Linux is at 5.2% [8] as shown in Fig. 1 –
OS Platform Statistics. MS Windows 7, released in late 2009 has not seen significant
adoption in K-12 schools, due to greater hardware requirements which necessitate more up-
to-date desktop computer systems [9]. Microsoft terminated licensing of the XP operating
system in July of 2010 while it has extended support until April of 2014 due to it significant
and current use. Microsoft considers Windows XP to be a legacy operating system and is
4
This move has been much relatively easy for home users and small business; however
large organizations and specifically school districts, have found this a serious challenge.
Many districts have moved some office and administration desktops to Microsoft Windows 7
while opting to maintain Microsoft XP in computer labs, libraries and classrooms for as long
as possible. The anticipation is, that funding or off-lease equipment will provide
opportunities to move to Microsoft 7 and later versions of Microsoft Office at a later date.
Many of the current technology plans are broad-based and discuss: district policies,
desktop computers with newer models and other ICT aspects but rarely address the cost,
The difficulty currently facing school districts can hardly be overstated. Many
districts have not articulated plans or committed resources which address the looming
challenge for transition from Windows XP to Windows 7 by the 2014 deadline. As districts
require at least two to three years to implement, most large technology plans including plan
development, consultation, budgeting and implementation this indicates the time left for this
process to occur is getting shorter and options are limited. Some districts are considering
XenServer or other technologies. However, many of these systems are not cost-effective,
even with academic pricing for a school district, and do not support accelerated video and
streaming sound required by many graphic applications. This model has been tried by school
Other alternatives to this traditional computing model are being explored by some
school districts. The Vancouver School District is making forays into this realm by replacing
5
some commercial software products with open source programs like OpenOffice, Microsoft
Publisher with Scribus, and so forth, nevertheless it does not address the unavoidable concern
about what happens when support is withdrawn for Windows XP and hardware have become
obsolete. Additionally, it does not significantly address issues around environmental concerns
and reducing GHG and CO2 emissions although some reductions have been made through
the use of programs like Faronic's Power Save software [10]. With funding being a continual
and yearly concern, Microsoft's eliminating support for Windows XP in 2014, educational
paradigm shifts such as personalize learning, distributed learning and mobility, Internet-based
model needs to be considered. The utilization of open source software in conjunction with
low-cost diskless clients can address many of the concerns that exist currently and anticipate
Linux thin client computing is a model which has been largely ignored by the K-12
educational community. The reasons for this are diverse and include: fear, uncertainty and
doubt about open source software, level of support requited, lack of technical knowledge for
poor documentation, lack of warranty and other factors. Not all of these however are valid
and the advantages and disadvantages will be further explored in this essay. Open-source
software and thin client technologies have evolved to be competitive, sustainable and highly
applicable for school districts. This essay anticipates the utilization of an energy-efficient,
Linux thin client computing model can provide significant benefits for K-12 education over
resistance to change.
6
Statement of the Purpose.
This essay will provide an overview of past directions regarding Linux thin client use
in an educational environment and will further seek to analyze current issues and trends as it
Research will be conducted into factors which may have impeded broad-based support for
this model including cultural bias, fear of the unknown, uncertainty, doubt, questions
regarding open source software efficacy and other issues which have been studied
extensively [11]. The purpose is to demystify the apprehensions around Linux thin client
technology, analyze the significant advantages while considering the disadvantages with a
view to provide districts and schools a viable and highly cost-effective, energy-efficient,
The author has been part of the K-12 educational environment since 1992 when
initially hired as a network engineer to administer and plan school-based Local Area
Networks for School District No. 73 (Kamloops). In 1996 the author was promoted to the
new position of Manager of Information Technology and given the opportunity to produce
the school districts first technology plan which included raising the computer-to-student ratio
students. During the subsequent four years there was a great deal of activity hiring additional
staff and upgrading districts servers and software, however elementary schools were often
left using outdated computers handed down from secondary schools, with the exception of
school offices. In 2000 a pilot project was conceptualized with the district systems analyst
7
with a proposal to utilize off-lease computers to provide a low-cost, open source-based
In the summer of 2001, School District No. 73 completed the building of a new
school, Pacific Way Elementary School and there was a requirement for technology to
populate the new school. In conjunction with the author, the district systems analyst, the
principal of the new school and the assistant superintendent, it was decided that off-lease 3-
year old computers, using Pentium III processors and 256Mb of RAM could be utilized to
provide the school with approximately 120 workstations. A Debian-based server running the
Linux Terminal Server Project (LTSP) would provide the bootstrap, operating system,
desktop environment, file and print capabilities [12]. Core software was included consisting
of StarOffice, Mozilla browser, gCompris children's software suite and other relevant open
source applications. This model utilized a school-centralized server, running Linux and LTSP.
The client boot process briefly outlined, involved loading the server-based operating
system with the Preboot eXectution Environment (PXE) through the client network interface
card (NIC), where the file server would respond with an appropriate network bootstrap
program (NBP). The server would load the NBP into the clients random access memory
(RAM), where users would be presented with a login screen and after an authenticated ID
and password and a highly customized desktop environment would be delivered [13]. As the
school was new, the network incorporated category (CAT) 5 cabling and utilized centralized
Ethernet switches running 100 megabit (Mb) to each client and 1 gigabit (Gb) on the
backbone to the school-based LTSP server. Under these optimal conditions the server could
easily handle the 120 clients with very low server utilization with most stations active.
8
The pilot was updated and refined for one year, after which an implementation plan
was created to resource and actualize the system for all 32 elementary schools. It was
therefore the implementation also included updating the infrastructure at all schools to CAT 5
data wiring, replacing older CAT3 cable where necessary, replacing older switches and
centralizing to facilitate at least 100Mb to each client and 1Gb on for the Local Area Network
LAN backbone, An electrician was provided with all of the necessary resources and time
with the task of upgrading each of the school LANs. This would precede the district analyst
who would subsequently replace the old Windows-based desktop computers with the new
basis. The initial pilot confirmed that the key factors which would enable success included:
• a very responsive LAN with as few hops between the clients and server as possible
• 100Mb to the clients and 1Gb on the backbone including the server
• homogeneous clients with consistent models of network cards, video cards, sound
In the fall of 2002, implementation began for 32 elementary schools scheduled for a
three year period. The full implementation was fully completed by the summer of 2005. Each
of the elementary schools was upgraded on a three week cycle. During the first week the old
computers were replaced, new consistent computers were installed and tested. The server was
setup, printing was established, user accounts were created, old documents and files were
transferred and preliminary image debugging was accomplished. During the second week
9
users were permitted to login and provided with training and orientation with the new
system. During the third week any issues which had not been discovered earlier were
addressed including just-in-time training and support. This process was repeated for the next
two years until all 32 elementary schools were fully converted. Throughout the
implementation there was continual development while rethinking regarding the efficacy of
utilizing off-lease or used computers. Although the systems were under much greater control
through centralized management and virtualizion, the age of the systems made them subject
to breakdowns after 2-3 years reducing user confidence in the system. As this was being
deliberated a Vancouver computer wholesaler contacted the district with an offer to donate
hard drives and RAM, which were the result of a merger with another company. The district
decided to accept this offer and purchased 512Mb of RAM for each system and spent the
next year replacing all of the older off-lease computers at the elementary schools.
The benefit of using dedicated diskless clients became apparent very quickly.
Utilizing new and more powerful computers greatly increased user confidence in the system.
Previously, support and maintenance of the older computers was continual and unproductive.
Emphasis was consequently placed upon responsive, remote management where incremental
the improved processing capabilities of the new computers, it became easier to offload
programs that had previously been run on the server to the clients when possible, and utilize
the distributed processing, incorporating technologies such as trivial file transfer protocol
(TFTP). As well, with the on-board video of the new workstations incorporating greater
accelerated video capabilities, it was possible to run better graphic-based software which had
10
been somewhat lacking in the previous iteration of the image. By the end of the elementary
school implementation it was necessary to reflect on the advantages and disadvantages of the
system and how the model could be scaled to secondary schools which were considered
schools located across a very large and diverse geographical area. Secondary school
curriculum included heavy use of office productivity software, had requirements for
aided modelling (CAM), video editing, graphic design, programming and required pervasive
committee was created to determine the viability of using the Linux-based model in use at
elementary schools into the secondary schools. After six months of discussion and
deliberations it was determined that a pilot would be accomplished at the smallest secondary
school with a principal who had previous experience and expertise with Linux and open-
source software. Subsequently a strategy was developed during the winter of 2005-2006 for
The secondary school image would incorporate updated open-source software, faster
diskless clients and better video chip sets. Over the winter, the electrician updated the LAN,
replaced the Ethernet switches and prepared the network. During the spring break of 2006 the
district analyst replaced all of the older Windows-based desktop computers with new diskless
clients, a new LTSP server and performed most of the preliminary work to transition files and
access for staff and students. Although not without its challenges the secondary school pilot
was deemed a success and refinements were continued throughout the subsequent months. In
11
the late spring of 2006 another plan was drafted to see implementation of the upgraded
secondary school image for the subsequent nine schools. Beginning in the summer of 2006
the remainder of the secondary schools were upgraded during times of low activity, which
included winter breaks, spring breaks and summer vacations and was finally completed by
the summer of 2009. Linux and open-source software have become the standard for School
Since the beginning of the secondary school implementation much has occurred. In
early 2008, the author was offered the position of the Director of Information Technology for
School District No. 63 (Saanich), accepted the new position and commenced work in Saanich
in February of that year. As part of the authors employment contract, district executive and
the Board of Trustees agreed to incorporate a new technology plan based on the experience
gained previously by utilizing open-source software and Linux diskless client technology.
Consequently, in the spring of 2008 an extensive District Technology Plan was drafted and
accepted by the Board of Trustees which included upgrades and support for the LAN and
technology, professional development and training, privacy and security and risk
dollars was allocated for the implementation and consultation began with the eight
The first task was to hire an experienced systems analyst and assist current district
technical staff to help transition to the new system. It was agreed that a new school, KELSET
Elementary school would serve as a pilot, based on the work done at the Kamloops school
district, with improvements and knowledge gained. An Linux LTSP server and approximately
12
120 updated diskless clients were purchased for a pilot which was implemented during the
summer of 2008. Throughout the 2008-2009 school year significant improvements were
made to the system, incorporating new open-source software, improved speed, pod-printing
capabilities and significantly the ability to manage the client for automatic turn-on and turn-
off based on a predefined schedule greatly increasing the energy efficiency of the model.
at the remainder of the elementary schools began during the summer of 2009 and continued
until the spring of 2010. A similar process utilized at Kamloops, was used at Saanich. An
electrician was hired and preceded the technology implementation for each school, upgrading
the LANs and replacing switches. Server closets were created, mini-server racks were
installed, the number of hops from the client stations was reduced and access greatly
improved to a minimum of 100Mb for clients and 1Gb for the server to central switches. An
implementation team, including the district analyst, two district technicians, the school
support person and the district teacher coordinator were on hand for the first week. With a
larger team it was possible to change out the old desktop computers on the first day, install
the server and new diskless clients on the second, transfer all files, create new user IDs and
setup printers on the third day. By the fourth day school staff would be assisted by logging
into the system and orientation provided. On the fifth day of implementation the district
technology coordinator would be taking groups of children through the system accompanied
by their teachers. Two staff would remain on-site for the next week, debugging assisting staff
and providing any support and training requested. By the end of the second week the schools
were operating efficiently and making good use of the new system. This process continued
until all eight elementary schools were converted to the new system by late winter of 2010.
13
Upon successful conclusion of the elementary plan, consultation was started with the three
The first middle school was implemented during the March spring break of 2010
utilizing one Linux LTSP server and 232 diskless clients. The middle schools have advanced
requirements and information technology staff spent a greater amount of effort refining and
improving the system for more sophisticated students and staff. As well there was additional
Windows-based software system in use at the middle schools which was not in use
The initial hope was to virtualize these packages using the open source Wine Windows-based
Application Programming Interface (API) [15]. Many programs function well under Wine,
however many do not and the aforementioned software would not operate properly after
much effort by district analysts. Consequently a hybrid model was developed where critical
computers connecting to the school network through Samba server interoperability also
loaded on the LTSP server [16]. The average middle school at the Saanich school district has
approximately 225 workstations for which approximately 15-20 units must continue to run
Windows XP until Linux versions become available or web-based versions are released.
Although this is an impediment it was not considered a significant enough obstacle to prevent
implementation for all middle schools. As a consequence, after the first middle school had
been thoroughly tested and refined, the subsequent two middle schools were implemented
14
Significance.
Diskless client technology is not new technology; however the manner in which it is
being implemented and deployed is new, enabling accelerated video and streaming sound.
Being able to scale very low cost, diskless clients puts the support emphasis not on the
desktop support, but on the use of the software and applications while providing constant
improvement and upgrades. The diskless clients that have been deployed at School District
No. 73 (Kamloops) and School District No. 63 (Saanich) and continue to be implemented
allow for multi-language support, ubiquitous access not only across the network, but securely
through the Internet with the use of software such as FreeNX. Another area of consequence
occurred after the Saanich school district hired an Energy Manager and there was a desire to
quantify the energy savings from fully managed diskless clients. Energy savings of 60
percent over the traditional model were quantified and realized through rigorous data
gathering and analysis as confirmed by BC Hydro engineers. Perhaps the most important
result from having BC Hydro confirm the achieved energy savings is the prestige associated
with greatly reduced energy consumption and reduction of GHG and CO2. As well, BC
Hydro is considering offering incentives to organizations who implement this model which
might include a portion of funding for diskless clients purchased [17], particularly if they
conform to EPEAT [18], including Energy Star [19], and 80Plus [20] standards..
Assumptions.
This essay assumes the reader is familiar with the K-12 public education sector and
the many challenges associated with public funding, while endeavouring to provide the best
the reader already has a familiarity with the Linux operating system, thin client technologies
15
and open source software, either as developer, a user, or a system administrator although
enough background information is provided for a broad perspective. Additionally this essay
assumes the reader is aware of global warming concerns and acknowledges there is a
significant challenge with climate change and the necessity to make changes.
Limitations.
diskless client model utilizing open-source software. Overcoming these obstacles is not
trivial and requires significant support from senior stakeholders. Perhaps the greatest of these
limited support for critical Windows-based applications, cultural bias, lack of executive
support and capital dollars. Although challenging, many of these deterrents can be overcome
There is still much work that can be done to improve this model and it may be an
interim step until web operating systems become fully developed making the need for local
servers redundant. In the meantime this model is the most elegant, cost effective model to
16
CHAPTER II
Server-based based computing models have existed since the first mainframe
computers were produced and “dumb” or text-based terminals were attached facilitating
display and data entry. These terminals were simple devices with no processing capabilities, a
screen for display, and a keyboard for data entry which were connected via RS-232 serial
connections [21]. In this model the monolithic server maintained the operating system, the
programs and data generated by the users and system. This computing model was dominant
until the advent of the personal computer introduced a new paradigm in the early 1980s.
Gradually most mainframe computers were replaced with mini servers connected to personal
computers (PC) with hard drives, hosting their own operating systems and providing a
graphical user interface (GUI), such as Microsoft Windows or the Apple Macintosh. In
By and large, over the next two decades, server functionality was primarily file and
print control devices that facilitated the sharing of data, controlled the network and other
centralized operations in what could be defined as a client-server environment. Since the late
1990s, this trend has changed gradually, where web and cloud-based servers provide access
to applications and resources, yet personal computers continue to connect to these systems
whether local or Internet-based servers. The PC or “fat” client is the antithesis to the early
terminal which has been alternatively called “thin” clients due to their heavy dependence
17
Although there has been a significant shift away from the early server-based, “green
screen” dumb terminals, the use of “thin” clients has evolved and persisted since the early
days of mainframe computers. A thin client in its strictest sense is a lower cost computing
and provide a screen output [22]. Thin clients normally lack any processing capabilities,
persistent memory or I/O devices support. Throughout that time there has been an ongoing
the user and one that is familiar for most personal computer users. This has been facilitated
through the use of products such as Microsoft Terminal Server [23], Citrix Terminal Server
and the Linux Terminal Server Project [12], to name a few, which utilize thin clients
manufactured by companies like Wyse, IBM, HP, Sun Microsystems and others.
The use of thin clients has been sporadic and not always brought the desired panacea
that was predicated by industry leaders such Oracle Corporation’s Larry Ellison who
attempted to build a business case in the mid-to-late 1990s around a thin client he called a
ridiculous device," where he saw the thin client rapidly replacing the user centric personal
computer [24]. Many technical and industry analysts also saw an eminent future for not only
the concept of thin client computing but a change in the very architecture for building
seamless network applications where the emphasis was on scalability, security with a smooth
migration to new technologies [25]. The reality was however the market share for the
functionality of thin clients and users desire to personalize their desktop environment.
Although the high expectation for the use of thin client technology has not been accepted to
18
the degree Ellison and others foresaw, there has been some successes in certain environments
and situations. Due to the scope of the topic this literature review will concentrate on thin
client use since the year 2000, with particular emphasis upon its application in K-12 and
post-secondary education and more specifically the significant benefits towards greater
energy conservation.
Historical Context.
In 1991 the late Dr. Mark Weiser envisioned the concept of ubiquitous computing for
which he coined the phrase “calm computing” where he anticipated the concept of a more
elegant form of computing. In his seminal work The Computer for the 21st Century Weiser is
quoted as saying “The most profound technologies are those that disappear. They weave
themselves into the fabric of everyday life until they are indistinguishable from it’’ [26]. This
vision of ubiquitous access and high mobility was the goal where researchers debated the
efficacy of how “thick” user’s clients actually needed to be, in order to realize the vision of
Dr. Weiser. There has been much discussion about how much processing power, disk
capacity, video capacity and so forth did the client require to provide sufficient access to
resources? Thick clients were by their very nature independent from the networked system,
incorporating their own operating systems, persistent memory, video, sound and other
capabilities which had the potential to introduce significant challenges for support, privacy
and energy management [27]. Satyanarayanan debated the potential for clients who were
functionally independent, facing all of these threats and could by their nature degrade the
vision of Weiser and others particularly regarding the hope that technology would disappear
into the environment, much like utilities such as electricity and the public switched telephone
network (PSTN). Calm technology could become the much anticipated, innocuous
19
technology in which people became little concerned as it became largely ubiquitous and
Although the notion of thin client computer, which incorporated a rich GUI-based
experience, was the expectation, the reality of very early testing of WAN-based, thin client
technologies recognized its inherent limitations. Vendors, such as Microsoft and Citrix and
applications service providers were proposing that thin client technology could deliver cost
effective, functional, secure, energy efficient and pervasive computers across a wide area
network. As well, it could also lower the total cost of ownership of an organization and
greatly reduce the maintenance and support required to maintain and manage personal
computers currently in use. In 2002 Lai and Nieh tested these propositions through the use of
evaluate thin client performance on a wide area network [28]. This was important research
due to the great differences in computing models. PC-based systems had the operating
systems and applications loaded on local hard drives, utilized powerful processors,
incorporated onboard video and sound capabilities with connections to the network for file
and print sharing, Alternatively thin clients utilised network boot technologies like Preboot
eXecution Environment (PXE) which provided the operating system, GUI environment,
applications and other functionality across the network. The thin client would have minimal
processing power, no persistent memory, and little video or sound capabilities with a single
point-of-failure being the network. Lai and Nieh tested six different clients across two test
bed environments and recorded results for data transfer of operations, letter latency, scroll
latency, fill latency, red bitmap latency, image latency, web data transfer, web latency, video
data transfer, video playback time and video quality. The results of the testing indicated that
20
in certain circumstances thin client technology could deliver acceptable performance even
when running across WANs and the Internet usually related to low-bandwidth requirements
such as text and data entry; however performance varied wildly across thin clients. They
concluded that the design trade-offs of the thin client capacity would not be appropriate in a
Science at Saint Louis University, considered the high computer-to-student ratio in the
computer labs and the need to provide student access in a “single cohesive environment” for
use in their course work [29]. They considered the common practises existing in-place for
facilitating student course work and proposed that thin client computing, utilizing open
source software could provide major benefits for the department and students. These benefits
included: a fully server-sided environment consisting of a server and thin clients where the
client acted only as a display and input device and all computation occurred on the server. It
could provide consistency in the image, be tightly controlled and easier to support. There was
also an assumption that student requirements were fairly basic and implemented open source
software would suffice. Prior to the thin client implementation the department had a very
Mac OS 9.1, Mac OS X and Red Hat Linux. During the academic year of 2003-2004 the
department implemented TightVNC as the basis for the system installed them on the existing
dual processor Dell workstations running Gentoo Linux. This allowed for up to 100 clients
per server access and the image was enhanced and supported over the subsequent year.
Remote users were permitted to access the system using TightVNC loaded on their personal
computers. A preliminary evaluation realized the following benefits including: consistent and
21
universal access, scalability, cost effective, open source software and a centrally controlled
system. Challenges included: universality while providing support on multiple and diverse
session management, security and OpenGL support. Other limitations included network
dependency and lack of local device support required by system users. Almost two years
after initial implementation, the overall analysis was, the thin client system had provided
significant benefits and more work needed to be done to overcome some of the challenges
and offered opportunities for continual improvement. Overall it was considered a success and
would be maintained.
Simultaneous to the work that was taking place at the Saint Louis University, Tolia, et
al at the Carnegie Mellon University School of Computer Science wrote a paper in 2005
entitled “The Seductive Appeal of Thin Clients” [30]. Their premise considered thin client
technologies promised many potential benefits however thin clients were not able to meet the
usability goal of “crisp interactive response” and their application was highly dependant upon
the application and available network quality. Similarly to an earlier paper written by
Satyanarayanan, the network being the single point-of-failure and network latency was
considered to be the largest impediments to wide-spread thin client acceptance. This paper
cautions organizations who were considering thin client technology to consider all aspects
before implementing. They surmised that management of personal computers often caused
organizations to consider reducing TCO through thin clients having the promise of lower
client costs and ubiquitous access to centralized resources. This is particularly attractive to
post-secondary institutes with large numbers of computer labs and computers. To measure
their supposition they tested the adequacy of thin client using an open source VNC client on
22
a local area network. They ran a series of programs, including The Gimp, OpenOffice
Impress and OpenOffice Writer across the LAN and traced packets for response times.
Latency times were recorded and analyzed where less than 150 ms latency was considered
greater than 5s Unusable. The results of their analysis showed that highly graphic
applications such as The Gimp perform poorly with moderate latency. Although less graphic
unacceptability to using thin clients in a networked environment. Their conclusion was the
promise of thin client technology posed a real threat to the high availability and graphically
In the same year Baratto, Kim and Nieh with the Department of Computer Science at
Columbia University, proposed a solution to the relatively poor graphics capabilities and
response of traditional thin clients, with technology called THINC or Thin-client InterNet
Computing. THINC considered the advantages of thin client computing while providing a
“high fidelity display and interactive performance in both LAN and WAN environments”
[31]. The difference between this model and previous attempts to improve the graphic
capabilities of thin clients was THINC leverages the inherent video driver while introducing
a simple virtual display driver which intercepts drawing commands at the device layer,
packetizes them and then relays them over the network to the client. The premise considered
the use of this technology to alleviate the need to change hardware, application software and
operating systems. Working below the window server and above the framebuffer at the
abstraction layer, THINC implemented a number of protocol commands for encoding display
23
updates which include: RAW for display raw pixel data at a given location, COPY for copy
frame buffer area to specified coordinates, SFILL for fill an area with a given pixel colour
value, PFILL for tile and area with a given pixel patter and BITMAP for fill a region with a
bitmap image. In order to manage network latency issues and possible network problems,
THINC introduced the notion of utilizing a command queue. The command queue is
designed to draw to a specific location, based on the order commands are received, thereby
freeing the problem for areas to become overwritten through subsequent commands. THINC
was tested across a widely used Citrix MetaFrameXP with Microsoft Remote Desktop using
Independent Computing Architecture (ICA) and Remote Desktop Protocol (RDP) on LAN,
WAN and Wireless 802.11 network environments. Additionally, it was tested within the
context of a LAN and greater distances over the Internet using a variety of clients including
PC, SunRay, Xclient, and a host of others to compare network response. Baratoo et al
concluded that using a new virtual display system, built around a virtual driver model,
optimizes the performance by up to 4.5 times and improves the graphic response for an
enhanced user experience. They also concluded that this technology worked well across the
Although thin clients had proved their merit for productivity software, web browsing
and other less network-dependant and low graphic applications, frustrations result when
attempting to use thin client technology for multimedia, gaming and other graphic intense
applications. With growing interesting in these areas, it was regarded essential that thin client
provide support to enhance the user experience. In 2006, De Winter et al considered a hybrid
technology using a combination of thin client protocol and video streaming to send the
graphical output of an application to a thin client device [32]. Traditional thin clients present
24
significant problems due to their low processing capabilities and lack of support for
Graphical Processing Units (GPU) with absolute reliance on network access. This model
proposed an additional driver abstraction layer between the graphical libraries and the device
driver layer and would determine if the application commands being passed were directed to
a standard video card device driver or one requiring GPU support. Additional information
regarding video requirements would be based on the amount of motion in the images, where
the software framebuffer exceeded a predetermined threshold where a pseudo device driver
would emulate a GPU video card and pass this along to the client. In this model the switch
between standard and enhanced video would not be apparent to the user. The researchers
utilized a test bed of PCs connected across a 100Mb LAN and developed a realtime desktop
streamer to compare personal computers utilizing traditional thin client protocols. A number
of parameters were measured including the influence on the bandwidth per frame for
different Group of Pictures (GOP) size and influence on delay for most determining
parameters. Four application types were also compared including: office applications, video
streaming, web browsing and graphic video gaming. The results of the research were
conclusive and the realtime videostreamer outperformed the classic thin client model
providing users a potential solution for enhanced video streaming and gaming. De Winter et
al proposed additional work on this model considering the significant energy efficiencies
Until 2006 most research was focused on enhancing thin client technology to reduce
TCO and provide enhanced usability, better access and improved graphics capability. In May
of 2006 Al Gore released the documentary “An Inconvenient Truth” which significantly
raised the awareness of climate change and global warming [33]. Interestingly there was an
25
increasing emphasis upon the energy savings which could be realized with thin client
comparison of PC and thin client desktop equipment” [34]. This study involved comparing
the environmental footprint of all aspects between PCs and thin clients and considered
including the benefits of both PCs and thin clients in an office environment for productivity.
A number of standards and criteria were used in the analysis such as the Environmental
Priority Strategy (EPS), Material Intensity per Service Unit (MIPS), calculations for energy
demands using Cumulative Energy Demand (CED) and ISO 14040ff for environmental
overwhelming advantages of thin clients for all metrics. Thin clients were smaller, lighter,
used less materials in manufacture, less e-waste, consumed less energy, had reduced GHG
and CO2 emissions, were quieter, required less physical space, required less maintenance and
In 2007, Przybyla and Pegah with the Ringling College of Art and Design, conducted
an analysis of technology use throughout the College including data centers and user
workstations [35]. They were concerned with the environmental impact of ICT as well as
energy consumption, affect on heating and cooling, e-waste and sustainability of their current
systems. Being high profile national College providing programs for visual arts and design
26
with a large-scaled integration of technology the responsible use of technology was
system throughout all campus buildings which had seen a near doubling of the number of
servers between 2000 and 2005 with a significant increase in the number of computer labs
which remained active 24 hours per day. The paper is primarily a discussion of the causes
and effects of the existing system with details on how the College could move forward to
reduce the environmental impact of ICT. Along there is scrutiny around how data centres
could be made more efficient there is significant discussion around the potential for use of
thin clients primarily as a focus for “eco-computing” which would facilitate less energy
consumption, less effect on heating and cooling, less maintenance and administration, longer
life expectancy and less e-waste. Although this is primarily a strategic directions paper it
does show the increased awareness for the need to become more environmentally responsible
in all facets of ICT particularly around workstation computing through the use of thin client
technologies.
Forrester Research conducted a study in 2008 entitled “Green Benefits Put Thin-
Client Computing Back on the Desktop Hardware Agenda” which considered the need for
organizations’ to re-examine thin clients with a view that this technology could have a
considerable effect upon reducing energy consumption, while decreasing GHG and CO2
emissions [36]. The basis for the analysis considered timing was opportune for business to
consider their next wave of desktop computer refresh with thin rather than thick clients due
to the growing corporate importance of green IT. The study considered reasons for thin
clients not attaining anticipated market saturation which had been predicted over the past
27
computer sovereignty. Organizations that had attempted to implement thin client technologies
were frequently held back from wide-spread use due to these and other factors. The promise
of thin clients as a solution for IT infrastructure had not realized the potential even with the
many acknowledged benefits. Although some interesting research had improved the
performance for graphic-based applications such as CAD, picture-editing and so forth, it still
did not meet the standards provided by PC-based computing. Although improvements had
been achieved for client response times, network latency, particularly over large distances,
such as the Internet, continued to hamper user experience. Also the absence of persistent
memory, such as a local hard drive and CD-ROM, rendered the thin clients unusable when
the network failed due to the single point-of-failure. Users who were accustomed to personal
computers were frustrated and saw thin client as a loss of personal control over their
environment, particularly when they could not download, install plugins or a range of
debatably useful, albeit non-corporate applications. These issues often served to generate
resistance by users. Interesting although some technical improvements had been achieved
over the past few years, many of the prominent issues continued to persist. The Forrester
study nevertheless considered the intensifying issue of global warming and climate change as
a way to engage users for change. Concerns which had previously not been of significance to
users might now be considered important to alter the cultural perception of this technology.
These included focusing acceptance based on: reduced power consumption, lowered
emissions and improved product life and sustainability. The Forrester study recommended
focusing the environmental advantages with the Chief Information Officer (CIO), the Chief
Financial Officer (CFO) and the Line of Business Officers (LBO) with an emphasis on:
understanding the realities of thin client computing, gaining an understanding of the broad
28
user context of the business, selecting appropriate vendors and engaging open dialogue with
important stakeholders in the business. The analysis considered the best approach for a
setting clear objects, open communication with stakeholders and most importantly
communicating the savings and efficiencies achieved through green computing models
Subsequent to the Forrester study additional research and analysis had been
accomplished into usability analysis in thin client acceptance by the School of Computing at
the Dublin Institute of Technology [37]. This study endeavoured to indentify successful
strategies for Linux thin client acceptance within an educational institute. It utilized the
Unified Theory of Acceptance and Use of Technology (UTAUT) model which aims to
explain user intentions and subsequent usage behaviour through four key constructs:
performance expectancy, effort expectancy, social influence and facilitating conditions) [38].
Although very little in the underlying technology had changed, the intent was to analyse user
acceptance for the use of Linux thin client and mitigate any changes for users through the use
of UTAUT. Two test environments were established using a mix of PCs and Linux-based thin
clients and users were allowed to login and experience the systems through a blind study.
recommendations for future thin client implementation was detailed. The results of the blind
tests confirmed a 92% satisfaction rating with thin client systems when users were not aware
of which workstations they were using. The UTAUT methodology assisted researchers to
distinguish between the purely technical and non-technical issues with a view to present thin
29
Organizations could consider user’s educational requirements such as thin client integration
with PCs, consistency with login procedures, storage options are identical to PCs and
Although there had been some sporadic use of Linux-based thin client technologies,
throughout the education community, much of the research previously done was based on
virtulizing the Microsoft Windows environment using Microsoft Terminal Server or Citrix
Terminal services. Linux and open source software had been generally discounted by the
There was a great deal of uncertainty about its efficacy, usability, support and other factors
with a view that although it functioned well in the server environment, however it was not a
viable option for corporate or organizational desktop users. In 2009 Braddock with the
Centre for Mobile & Converging Technologies at the Leeds Metropolitan University,
considered the viability of Linux thin client and open source technologies through the use of
the Linux Terminal Server Project (LTSP) [39]. Although not a peer-reviewed article,
Braddock deliberates the advantages of utilizing open source software including cost,
openness, availability and other aspects while considering its disadvantages which includes
lack of marketing, user perception, confusion around the model and so forth. He considers
the advantages of using Linux thin clients with a particular emphasis on LTSP. As well open
source technology with Linux distributions such as Ubuntu had seen steady and incremental
improvements since its initial creation in 2004 [40]. Ubuntu comes bundled with LTSP and
provides support for thin client technology inherently, making it a much more palatable
option. Braddock also introduced the use of diskless clients as an alternative to thin clients.
Diskless clients incorporate characteristics of both thick and thin clients together with faster
30
processors and onboard GPU which would support accelerated video and graphic
applications. This option in conjunction with optimum distribution showed potential for
providing the advantages of thin client technologies, open source software and usability
University compared the use of thin client technologies as part of the workstations provided
for students at the College of Engineering [41]. Approximately 50% of the lab computers
utilize Linux thin client technology and contrast personnel support requirements between
PCs and thin clients, mobility, security, system administration, financial implications,
previous studied the he reflects that thin clients offer significant and overarching advantages
in almost all areas of ICT particularly with energy conservation, however challenges for full
scale implementation include: resistance by users to adopt, licensing, support for peripherals
and limited graphic capabilities for graphic intensive software such as CAD, 3D Modeling,
video streaming and some web-based flash applications. His assessment is that thin clients
offer many technical advantages over PCs and that higher education has largely overlooked
Polit´ecnica de Madrid, focus on the significant benefits offered through thin client
technologies with an emphasis on Linux and open source software through the LTSP project.
They consider improved manageability and administration, greatly increased security, lower
TCO and considerable energy efficiencies possible with this model. In support of this they
provide a comparison of two metrics the first being the evolution of price differentiation and
31
the estimated hours of maintenance per week between PCs and thin clients. As well they
consider how new technologies are impacting post-secondary education including a shift
from the traditional models of lab-based computers to access remotely through distance
and are becoming a “resource barrier” for educational institutions. With greater concerns
about budgets, sustainability, support, ubiquitous access and global warming new models
need to be developed which include thin client, mobility and remote access. They present
many of the same challenges that have been previously discussed for thin client technology
however also stress that research in many areas will remove the remaining obstacles making
thin client technology the most attractive choice for educational institutions.
Over the past decade much of the research for thin client technology focused on a
comparative analysis between thick and thin clients, improving video performance and
reducing network latency. In almost every paper reviewed, there was an acknowledgment
that thin client technology could provide significant advantages over PC-based computing
• reduced maintenance
• reduced support
32
• standardization and consistency
• lighter
• quieter
Where open source software and systems were also utilized, such as Ubuntu and LTSP, thin
client technology provided additional rewards in reducing software licensing costs. Many
post-secondary institutes found that open source software could be effectively combined with
thin client in multipurpose teaching labs, although there has not been a significant uptake on
this model.
When thin client technology was utilized there was general agreement that it lacked
in three significant areas. Firstly, it did not sustain software which required high graphics
requirements such as computer aided drafting (CAD), photo editing, vector drawing and
other applications requiring a “crisp interactive response”. This was consistently the case
even improving video quality through developments such as THINC and driver abstraction
layer improvement [31]. It was still not possible to provide comparative results, across a
network, with a PC enabled with faster processor, more RAM, GPU and local persistent
memory. Secondly, due its complete dependence upon a single point-of-failure, the network,
33
thin client technology was subject to network interruptions and packet latency which limits
its use for any application requiring contiguous access such as streaming video or audio. To
lessen interruptions and breaks in a video or audio stream clients require reasonably large
memory buffers which would aggregate the video and accommodate network congestion,
latency or brief interruptions. One of the considered benefits of thin clients is greatly reduced
capacity including processing power, RAM and persistent memory. Typical thin clients have
low-powered and weak processors, greatly reduced RAM which is normally 128Kb to
512Kb, no GPU and no persistent memory. These features of thin clients are often considered
primary benefits but serve in hindering them for the core functionality required for streaming
video and audio. As a result, buffering video and audio streams is normally not an option,
resulting in thin clients suffering support for real-time applications, which users have come to
demand. Finally, if the client/server environment utilizes open source software combined
with thin client, there are an additional issues around lack of support for “industry standard”
software that runs on Windows-based systems such as Microsoft Office, Adobe Creative
Suite and others. There are some possibilities for substitutions; however there are some
educational environments which require the use of these specific products and the use of
Beginning with Dr. Mark Weiser’s vision of ubiquitous computing in the early 1990s
and Larry Ellison’s notion of the networked computer in the mid-1990s, thin client
while there has been much research to overcome the challenges of thin client, it has failed to
34
deliver on the key issues of enhanced graphic capabilities, network consistency and support
for application software. Due to these limitations pervasive use of thin client has not been
realized and is only utilized sporadically. Interestingly thin clients are gaining greater
attention due to concerns about global warming and its impact on the environment. Even
withstanding this important development, adoption for thin client will continue to be
restricted since the essential problems have not been substantively addressed. This essay
concludes that traditional thin clients cannot solve the outstanding issues and that it requires a
The purpose of this survey was to perform an historical analysis of thin client
technology. My analysis has indicated that thin client technology has not realized its
promised potential due to a fundamentally flawed model. Diskless, not thin clients can
provide the functionality which would combine the advantages of both thin and thick client
computing solution which would come much closer to realizing the vision of Dr. Mark
Weiser’s vision for ubiquitous computing. This model has the potential to offer greatly
reduced energy efficiencies and reduced GHG and CO2 while lowering TCO and providing
the “crisp interactive video” which PC-based users have come to expect. As well, the many
issues regarding network latency and video performance can also be addressed through the
use of greater client capacity available in diskless clients. For many organizations,
particularly education, this model comes much closer to the concept of utility computing
35
CHAPTER III
METHODOLOGY
Since its advent in the early 1980s the PC has gradually become the primary
computing device in most organizations, and although bringing powerful technology to users,
it is has brought with it increased complexity, security concerns, added costs, additional
staffing and often times significant instability. When computers were first introduced,
mainframe or mini computers hosted systems using terminals connected with serial
connections. The advent of networked PCs became more pervasive and powerful, with
software becoming more sophisticated. The result has been the increased complexity of
computer based information systems becoming more difficult to maintain, costs escalate and
network and ultimately the Internet, opening PC to vulnerabilities such as malware, viruses,
Trojan horses, and other security concerns [42]. Information technology managers, and
information technology departments, are finding it increasingly difficult, time consuming and
In K-12 and post-secondary environments, limited financial resources are often times
inadequate to even maintain outdated systems [44]. In 2008 School District No. 63, had
approximately 2500 PCs, connected to local area networks (LANs), which are in turn
connected to a district wide area network (WAN) spanning 18 sites and ultimately the
Internet. The IT department had a staff of three centralized technical support people and nine
site-based support people primarily responsible for maintaining the system. The challenge for
the information technology (IT) department was to not only maintain the over 100 plus
36
servers, but the network infrastructure and the 2500 networked PCs. Like many school
districts that cannot afford new computers, most were received by donation from Computers
for Schools (CFS) or purchased “off-lease”, resulting in the hardware being approximately
three years old before the school district integrated them. The primary operating system
continues to be Windows XP for which Microsoft has indicated will no longer be supported
by April 8, 2014 [45]. Realities include IT staff being dedicated to providing support to this
outdated technology, and rarely able to keep up. It has been widely recognized that capital
budgets are under funded [46]; consequently, it is possible to only upgrade a small
percentage of PCs in the system on an annual basis. The choice districts are often faced with
is, allowing the PCs to become progressively obsolete or find a more cost effective,
sustainable solution.
secure in an enterprise environment. Utility programs such as Ghost and Deep Freeze
certainly assist, but they also add to the total cost of ownership (TCO) and require IT staff
intervention whenever new programs are added or security patches and upgrades are
required. As well, the PCs, in a large organization are difficult to secure from a software
license standpoint. Regardless of policies or acceptable use agreements, users still manage to
subvert systems, by installing unauthorized or illegal software, which can compromise the
organization. Schools have been subject to audits from the Business Software Alliance (BSA)
and levied fines or payments due to member companies. Typically, teachers purchase
software products for home and then assume they can use them on their school computers
and vice versa. Obviously, a school district or university cannot be seen to uphold high
37
ethical standards, if its users are installing software, which violates software license
Added to these IT-based pressures is the need to reduce GHG and CO2 from ethical,
legal and moral responsibility. The province of British Columbia introduced Bill 44 in 2007
call the Greenhouse Gas Reduction Targets Act [47]. This Act called for the provincial
government and all aspects of the public sector, including school districts, to set reductions in
GHG by 33% by 2020, 80% by 2050 and 100% carbon neutrality by 2010. In early 2010
School District No. 63 (Saanich) has hired an Energy Manager sponsored by BC Hydro. The
energy manager has since created and implemented a Strategic Energy Management Plan
(SEMP) outlining energy reduction strategies for 2010 until 2020. The SEMP plan calls for
annual energy reductions of 5% per year until we have reached 33% of our 2007 energy
consumption. This is an ambitious objective which involves changes in culture, practice and
committed resources. Currently the district has achieved these objectives largely due to the
diskless client initiative which has seen reductions in energy of ICT energy of approximately
All of these increasing pressures and emerging issues indicate that districts must
either find the necessary funds to implement new and current technologies to facilitate the
teaching learning process in the 21st century or risk becoming irrelevant. With funding as the
and novel ways must be found to maintain currency in a rapidly changing technological
environment. This essay considers the efficacy of Linux diskless client technology which
38
Comparing Thin, Thick and Diskless Clients.
In my survey of the peer-reviewed papers and literature for the past decade most of
the research has been focused on the relative merits of thin over thick computing models
although there is frequently a muddying of terms relating to what constitutes a thin, slim or
diskless client. Strictly speaking thin clients could be defined as computing devices which
have little processing capability, limited graphic capabilities, reduced RAM and no CD-ROM
or hard drive. Thin clients do not generally have local operating systems and as a
consequence rely completely on the network and server for their functionality. All of the
manage too many simultaneous connections often topping out at 25 to 30 concurrent client
connections. The server side software sends screen output to the client and the client
responds with keyboard or mouse input [48]. Having a small amount of RAM, a weak CPU,
no persistent memory and the client is completely dependant upon the server and the client
cannot function in a connectionless state. Their advantages include high energy efficiency
due to small power supplies, low maintenance, greatly reduced support and much lower total
cost of ownership.
Thick clients by their nature are networked computers having most of their resources,
including operating system and many applications installed locally. They generally have lots
of RAM (2Gb or greater), a powerful CPU, GPU graphics resources, sound capabilities, hard
drive, CD-ROM or DVD drive and network card [49]. Although they can function
distributions. Users tend to prefer thick clients as they offer more control of the desktop and
39
unless prevented through an administrative policy, can install software locally. As thick
clients have their own operating systems and applications they can continue to allow users to
work even when disconnected to the network making them not fully dependant upon
centralized resources. This functionality along with fast GPU capability has led to their wide-
spread and continued acceptance by both organizations and users. The primary disadvantages
of thick clients include increased cost, higher energy costs, greater support and maintenance
Between the thin and thick clients lies a sort of no-mans-land which has been largely
unexplored and overlooked from my literature survey. By definition diskless clients have
crossover capabilities which have characteristics of both thin and thick. They incorporate all
of the advantages of thin clients having no hard drives or CD-ROMS, more energy efficient
power supplies, require less maintenance and support, cost less and do not require client
desktop support. Unlike true thin clients they have many of the benefits of thick clients which
include more RAM, faster CPU, GPU graphics support and support many of the applications
users want and require. Using ten client characteristics as indicated in Table 1 - Comparison
characteristics of thin, diskless and thick clients, I contrast the three computing models
40
Table 1. Comparison characteristics of thin, diskless and thick clients
Client Type
Characteristics Thin Diskless Thick
Low Watt Power Supply 3 2 1
Energy Efficiency 3 3 1
CPU Capacity 1 3 3
GPU Capability 1 2 3
Cost 2 3 1
Availability 1 1 3
App Location 1 2 3
Support 3 3 1
Overall TCO 3 3 1
Score 18 22 17
The number coding is an attempt to visibly rate each characteristic per category.
Negative characteristic have been weighted lowest with a 1, characteristics that might be
considered mid-way or nether positive or negative have been weighted with a 2, while
positive characteristics have been weighted with a number 3. Using this system it is possible
to rate characteristics of 1 for poor, 2 for medium and 3 for good; providing a crude
Thin and thick clients have very comparable scores, respectively 18 and 17 with
diskless clients scoring a marginally higher 22. Obviously this is a very simplistic way to
consider the relative strengths and weaknesses of these different models however it does
serve to underscore the aspect of diskless clients having many of the strengths of both thin
and thick clients while very few of their inherent weaknesses. Taking in to account the
prolonged and pervasive support for thick clients it is important to discuss what makes thick
clients the overwhelming choice over the obvious comparative overall advantages of thin
clients.
41
Although thin clients have proven benefits for IT departments which include reduced
cost, support, maintenance and TCO they have not been able to overcome the disadvantages
in providing support for graphic applications, limited number of clients per server and
functionality when network connections are lost. As a consequence users have continued to
expect and demand clients which afford them access to both networked and local applications
organization users are less concerned about the energy consumption of their clients and the
cost to purchase and support them. As thin clients have not produced the panacea which
Larry Ellison predicted in the early 1990s and researchers have not been able to overcome
the deficiencies of thin clients a different client computing model must be considered.
The previous comparison of computing models begs the question: what can diskless
clients offer to the organization which neither thick nor thin can provide? Diskless clients
offer the best possible compromise for large scale computing environments where TCO,
energy reduction, sustainability and functionality are prerequisites. Thin clients were not able
to subvert the popularity and functionality of thick clients while thick clients cannot be
sustained without significant increases in funding and support. As neither model provides the
foundation to meet all of K-12s objectives and requirements diskless client technology
42
CHAPTER IV
RESULTS
The use of Linux diskless client computing can dramatically alter a school districts
ability to provide staff and students with a secure, stable, and cost effective computer-based
hardware replacement while becoming more complex and support intensive. Districts
struggle to maintain and support systems which constantly require maintenance on users
networked personal computers. These computers, often running Microsoft Windows and
Windows applications, require vigilance with virus updates, software upgrades, security
patches, regular maintenance, and support. With increasing costs, reduced budgets and
Microsoft’s intention to cut support for Windows XP in 2014 it leaves districts with an
impending crisis. Add to this provincial requirements to greatly reduce energy consumption
by 2020 will require significant change by providing relevant and cost effective information
systems.
The primary goal of Linux diskless client technology is; the user’s personal computer
becomes secondary to the actual business processes, and enables users to perform their
required tasks while not having anxiety about viruses, crashing, access, and support. Diskless
client technology has relevance in providing educational environments with lower total cost
of ownership, reducing maintenance and support while greatly reducing energy consumption.
My research and practical application has show the efficacy of this technology as a viable
funding being the major limiting factor, valuable school district resources can be better
43
utilized in system development, support, and user training rather than purchasing hardware,
software licenses and maintaining systems. Although the issue is extensive, I briefly discuss
the methodology behind Linux diskless client technology which can effectively reduce the
total cost of ownership, streamlining and improving user access to the information and data
they require. As well there is a discussion around how diskless client technology is an
organizations require more detailed and complex analysis of information and data. The
objective of a superior information system is to provide accurate and reliable access to data,
information and knowledge that enhances the organizations ability to attain their core goals.
The challenge of school district information technology (IT) managers is to provide users
with computer based information systems that are cost effective, where the technology
becomes discreet to the business processes. Ironically, the advent of the personal computer
(PC), although introducing powerful processing capabilities to the desktop and a user-
friendly graphic user environment (GUI) has moved away from the efficiency and simplicity
of the mainframe with its centralized host model. Organizations have found it necessary to
technology director of a mid-sized school district. The cost of providing CBISs with ever-
increasing requirements to staff and students, while users networked PCs continues to age is
technology integration using Linux diskless client computing as opposed to the model of
44
“thin” client computing that has become an industry byword. This, in turn, does not imply
that the model developed at School District No. 63 is necessarily a panacea, but it has been
proven to realize the expectations of a modern CBIS and provide significant benefits beyond
often times inadequate to even maintain outdated systems [50]. At School District No. 63,
there are approximately 2500 PCs, connected to local area networks (LANs), which are in
turn connected to a district wide area network (WAN) spanning 18 sites and ultimately the
Internet. The challenge for the information technology (IT) department is to not only
maintain the over 100 plus servers, but the network infrastructure and the 2500 networked
PCs. Like many school districts that cannot afford to populate the system with new
computers, many are received by donation from Computers for Schools (CFS), or purchased
off-lease, resulting in the hardware being three to four years old before the school district
integrates them. The primary operating system continues to be Windows XP, no longer
facilitated with three technicians based out of the central office and another nine equivalent
located at school sites. The choice the district is faced with is, allowing the PCs to become
Thin client computing is a technology that although not new, is seeing resurgence and
incorporating the GUI interface users expect on their PCs however as shown earlier faces
significant barriers to user acceptance. It is in fact a hybrid model, diskless clients which can
offer an acceptable alternative between thin client and fat clients. I consider the history and
45
relevance to the implementation at School District No. 63 and the methodology around the
effective use of Linux “diskless” client computing in the organization, due to its reduced
costs relative to proprietary systems, its scalability, effectiveness and energy efficiency.
An organization, that requires client access to the corporate Intranet and the Internet,
can take advantage of this technology, as long as well thought out planning and adequate
resources are available. These resources include knowledgeable staff, adequate budgets,
testing and refining, staff preparation and training. All of this can be clearly defined in a
CBIS model designed to integrate the Linux diskless client technology and the business
processes of the organization. Educational institutions have successfully used thin client
technology to provide cost effective, reliable and effective systems to both staff and students.
The Linux diskless distributed computing model provides many additional advantages, and
includes:
• centralized administration
46
As well, educational institutions that have implemented this technology also note
benefits that include greater user reliance on the system, software standardization, single
point software installs, and the ability to change and enforce organizational policies.
In 2008 the author was hired as the Director of Information Technology for School
District No. 63 (Saanich) on southern Vancouver Island. Saanich school district provides
publicly funded educational services to approximately 7,500 students. The district has a
school board office, facilities plant, eight elementary schools, three middle schools, three
secondary schools, one alternative school and one distance learning school with a total of 27
sites. The author was hired, in part, due to the successful implementation of diskless client
technology at School District No. 73 (Kamloops) where he was employed since 1992 and
had been involved with the development of diskless client technology since 2001 where
After starting with School District No. 63 the initial task was to create a detailed three
year technology plan which included significant changes to the district network
professional development, security, privacy, risk management and a business continuity plan
[51]. The plan was atypical to many district plans in that it includes a comprehensive scope,
budget and project lead for accountability. Core to the plan was significant upgrades to the
district local area networks and wide area networks, complete replacement of all 2500
Windows workstations with diskless clients, replacement of all aging CRT monitors with 19”
LCD monitors, elimination of all Windows-based servers with Linux LTSP servers,
virtualization of redundant servers, elimination of all personal inkjet and laser printers with
47
energy efficient pod printers and multifunction devices. Project costs are included in Table 2
Multifunction devices and pod printers were ancillary to the technology plan and
monies were provided out of existing operating budgets. As well training and professional
development costs of $227,500 were budgeted separately out of operational funds and not
included in the capital budgets for the diskless client implementation. The Saanich school
district was fortunate to have had capital reserves of approximately four million dollars of
which 2.1 million, due to land sales, which were made available for this project.
In the June 2008 School Board meeting the board of trustees passed a motion to
direction [52]. The vision of the strategic plan recognized the need for students to be
technologically literate to compete and excel in the twenty-first century. With this mandate in
the summer of 2008 School District No. 63 (Saanich) piloted a system at KELSET
Elementary school, based on Ubuntu 8.04 Long Term Support (LTS) Hardy Heron [53]
incorporating the Linux Terminal Server Project (LTSP) [12]. A programmer analyst had
been hired 2 months earlier, with a degree in computer science and extensive experience with
Linux and open source software to develop and improve the image. Initial work began
immediately taking the standard Ubuntu/LTSP installation and developing custom menus,
48
installing pertinent applications, determining security role types and other details specific to
elementary schools. The version utilized was Ubuntu 8.04 LTS and LTSP 4.
From the authors previous experience at the Kamloops school district it was
important to ensure that all factors necessary to make the pilot a success were put into action.
This included ensuring the LAN at the school was optimized for a diskless client system
including a centralized server closet with switches located centrally. Although not strictly
necessary the client response time has shown improvement by having the server to central
switch connect at 1 Gb and all other switches co-located centrally connected at 100 Mb. As
the school was relatively small independent runs were made to each work station ensuring
In order to achieve a better user experience all existing workstations were replaced
with the same diskless client workstation, a new 19” wide screen LCD monitor and new
mouse and keyboard. It was considered important to provide the same environment and
experience to all users rather than attempt to repurpose older computers. This was done to
improve consistency for the image, reduce maintenance and support, diminish eye strain
manifest with the older CRT monitors, create a common experience and reduce the energy
consumption by using Energy Star [19] monitors and 80 Plus power supplies [20]. The
hardware initial hardware requirements for both the diskless clients and servers have evolved
since the initial pilot and specifications will be provided in a later section of this chapter.
The initial pilot at KELSET elementary school lasted through the 2008-2009 school
year where it was refined and enhanced. A great deal of user feedback and recommendations
were generated through the pilot project. Successive fixes, upgrades and enhancements were
49
applied in response to user feedback. By the end of June 2009 the project was deemed
During the summer of 2009 an implementation plan was developed for the remaining
seven elementary schools commencing in the fall of that year. A schedule was created with a
successful completion of all elementary schools by February of 2010. Subsequently the three
middle schools were scheduled and implemented beginning in the spring of 2010 and
completed by the fall of 2010. During that time the image had undergone continued
enhancements and improvements. Each school saw successive improvements in the image
During the fall of 2010 it was determined the upgrades would be applied to the early schools
including all of the major enhancements including Ubuntu 10.04 LTS, upgrades to LTSP,
software upgrades and so forth. Each of the schools was scheduled over an eleven week
period. It was initially thought that the upgrades would take the better part of the morning,
however the time to upgrade a typical school server with all upgrades and enhancements took
approximately one hour. Work commenced at approximately 7:30 am at each site and was
complete by school start time at 8:30am. The entire upgrade for all schools was completed by
Since that time a detailed analysis has been ongoing for the implementation of the
three secondary schools. Software requirements, image enhancements and user requirements
have been completed. A detailed implementation plan and schedule was developed and
implementation of the first of three secondary schools began in the early summer of 2011.
The second and third are planned for completion by March of 2012. Details of the software
50
being used and how we have coped with critical Windows-based software will be outlined
To date the implementation has been deemed a success by schools, administration and
senior staff, albeit with continued development of the image which is now in version three.
Overall user satisfaction is high, site-based technical support has been virtually eliminated
and many of the problematic issues normally inherent in a PC-based system have been
resolved. Implementing schools have achieved compliance to district acceptable use policies,
desktop security issues have been effectively eliminated, hardware failure and obsolescence
has been greatly reduced, software licensing issues are centrally administered and user
acceptance has been increased due to the high reliability of the system. As well, energy
consumption has been reduced by 70 percent for information technology use for a total of ten
percent overall. Detailed analysis of the energy efficiencies will also be addressed later on in
this chapter. The result of this implementation is far reaching and significant not only for
information technology but has significant implications for educational outcomes, energy
As indicated in earlier chapters there is ambiguity with the terminology between the
three models of network client computing often all mislabelled “thin” client computing. The
three models often termed “thin”, “diskless” and “fat” clients share some similarities but
some important difference as well. The first model requires the use of a Terminal server
whether provided by Microsoft Server Terminal Services, Citrix Presentation Server or Linux
Terminal Server Project. Both the Microsoft and Citrix server solutions have client
51
system having a minimum requirement for client hardware specifications. Clients that run
Terminal Services are not required to have a great deal of processing power. For example, a
Pentium with 256 MB of RAM (and sometimes less), network interface card supporting PXE
Therefore, it is relatively easy to integrate Terminal Services into a network that has older
computers and equipment, or commercial thin clients provided by vendors supplied by Wyse,
HP, Lenovo, and so forth. A thin client by this definition, is a client computer or client
software in client-server architecture networks, which depends primarily on the central server
for processing activities, and mainly focuses on conveying input and output between the user
functionality independent of the server. A thick or fat client does all processing on the client
and requires local secondary storage and only passes data for communications and storage to
the server. Such a client still requires periodic connection to the server and server resources,
but can provide many functions independent of the server including accelerated video for
high-end applications such as video, CAD and other graphics intensive programs. Many if
not all of the users applications are often loaded on the client. Additionally, the fat client will
always have a hard drive for the loading of a full local operating system and application
Diskless clients are a form of hybrid between the thin and fat client models, which
process data, using their own CPU and RAM to run software, but do not store the data
persistently [54]. The server handles the task of handing off the OS kernel, storage of data
and some if not all applications. In contrast, thin clients have all processing being
52
accomplished by the server. The thin client may operate an embedded operating system or
client software, which allows simple input/output tasks to communicate with the server, such
as drawing a dialogue box while waiting for user input. While the fat client has a full
operating system complete with secondary storage, such as a hard drive, to load applications
and simply use the server for data storage or print services. Diskless workstations can be seen
as a very elegant compromise between the two models where they combine the best of both
The differences between the three models are highlighted in Table 3 – Functional
53
As can be seen from the above table, diskless clients offer very low unit and support
costs, distributed processing which reduces the number of servers required, accelerated
graphics for high-end applications and very low administration requirements. At School
District No. 63 we have implemented this model successfully at eleven schools. Briefly, then
the model can be summarized by considering the theory of operation, the system
including initial boot up, running an operating system (when the Linux server passes the
kernel to the client), which is generally expecting access to secondary storage and subsequent
client applications. This section describes the theory behind how a diskless workstation
operates, focusing on the School District No. 63 evolved Linux-based, custom, diskless
workstation. Over the evolution of our implementation we have used both Etherboot and
Preboot Execution Environment (PXE), with mixed results. The original implementation
utilized recycled desktop computer and operated the Etherboot process, as it allowed the
system to determine which network card was installed in the PC. This was necessary as the
client base was wide and varied, and most client computers included floppy disks or hard
drives. Many educational organizations have attempted to repurpose aging PCs however this
can be problematic due to the huge variance of model and the poor energy efficiency. Our
initial implementation for booting these diverse workstations is shown in Fig. 2 – Diskless
54
Initially determine which NIC is installed (via lsmod and/or) Knoppix (3c905, eepro100)
/usr/ltsp/etherbootX.X.x/
Finds Card & downloads kernel
./bin32 all images for network cards
make bin32/3c905-tpo.fd0
(.fd0) is the device.
make bin32/3c905-tpo.lilo
cp 3c905-tpo.lilo /tftpboot/lts/ltsroot/lilo/b
Each machine is assigned an IP and kernel based on its MAC address. On the server:
/etc/dhcpd.conf (contains DHCP settings)
With the opportunity to have a uniform client with consistent motherboard and on-
board chipset the preferred method is PXE boot. Since that time we have standardized clients
to one particular make/model, with the same motherboard and very specific chip sets for
network interface card (NIC), video and sound. As well, these diskless clients do not have
floppy drives, hard drives or other types of secondary storage. Fig. 3 – PXE boot with sample
55
Assuming that pxelinux.0 and pxelinux.cfg are installed in /tftpboot...
ip=dhcp
create a symlink to the boot file above for the mac address of the system prefixed with
01.
ln -s via-c3.cfg 01-00-15-f2-15-c2-69
note it is easier to create a file called "default" for the majority of the clients and only
use symlinks for special-case computers, this ensures consistency.
edit /etc/dhcpd.conf file to ensure it knows to load pxelinux with:
filename "pxelinux.0";
Since our initial pilot of the KELSET elementary school this has been further
simplified by later versions of the Linux kernel and LTSP. Simplified we utilize copy the
kernel image through the use of a customized, created by the LTSP project, called ltsp-
update-kernels [55]. This script copies the kernel, ramdisk and performs other housekeeping
duties. Next the bootloader utilizes SYSLINUX which delivers the operating system via the
PXE boot process [13]. The process for booting up a diskless client is very similar to that of a
“fat” Linux client and could be compared in Table 4 – Thick client versus diskless client boot
process comparison.
56
Table 4. Thick client versus diskless client boot process comparison
Thick Linux Client Diskless Linux Client
1. Computer starts up, BIOS initializes 1. Computer starts up, BIOS initializes itself
itself
2. BIOS loads the bootloader {1} from the 2. BIOS loads the bootloader {3} from the
disk and runs it network {4} and runs it
3. Bootloader loads the kernel and 3. Bootloader loads the kernel and initramfs
initramfs {2} from the disk and starts the from the network and starts the kernel.
kernel.
4. Kernel starts up, mounts the root 4. Kernel starts up, mounts the root
filesystem, and passes control to the first filesystem {5}, and passes control to the first
userspace process: /sbin/init userspace process: /sbin/init
5. Finally, init is responsible for starting the 5. Finally, init is responsible for starting the
rest of the system. rest of the system.
{1} early versions of the bootloader utilized Linux LOader (LILO), this has since changed to
{2} initramfs or "initial RAM file system" is a small archive containing drivers and programs
that are needed to get the system started, such as storage drivers [56].
{3} GRUB is a bootloader used primarily for desktops and servers that boot from a local
hard disk. For embedded or networked systems, the most common bootloader is now
• The bootloader loads the kernel and initramfs using the TFTP protocol
{5} on a system with a local hard disk, the root file system is a partition on the disk. On a
By contrast these two processes it can be noted that booting a diskless client is now
57
very similar to booting a fat client which has a local operating system, apart from programs
which are loaded from the network instead of a local drive. Early versions of the process
required a custom complied kernel which included a network driver to boot from the
network. In current LTSP versions the initramfs that is loaded alongside the kernel includes
the network driver, therefore a standard distribution kernel is sufficient and requires very
little customization [57]. This has greatly simplified the boot process particularly if
Early versions of the system required that every workstation variant be determined
through a clear text file with its IP and MAC address stated. A script determined which card
was present then downloaded the appropriate kernel customized for each specific client
/usr/ltsp/etherbootX.X.x/
Finds Card & downloads kernel
./bin32 all images for network cards
make bin32/3c905-tpo.fd0
(.fd0) is the device.
make bin32/3c905-tpo.lilo
cp 3c905-tpo.lilo /tftpboot/lts/ltsroot/lilo/b
Figure 4. Diskless client customized kernel script
As can be assumed this was time consuming and required a great deal of additional
kernel configurations. With recent versions of the Linux kernel having all network drivers
incorporated there is also no need to utilize this kind of script. Each school server now
utilizes a MySQL database containing MAC addresses, IP addresses and other relevant
configuration data. Fig. 5 - Diskless client MySQL configuration database is taken directly
59
| dns2 | varchar(15) | YES | | NULL | |
| lease_time | int(6) | YES | | NULL | |
+------------+-------------+------+-----+---------+----------------+
The DHCP server and BIND DNS server are both configured based on this database.
Although still requiring detailed knowledge of Linux, the Linux kernel and LTSP the whole
boot process has been significantly improved and update for current Ubuntu server
distributions.
Network booting the diskless clients on an IP based network requires a server on the
network that supports both DHCP and Trivial File Transfer Protocol (TFTP), to provide
configuration information and allow transfer of initial boot files. The Dynamic Host
Configuration Protocol DHCP provides address discovery with the ability to also pass other
initial configuration information to the booting host. TFTP is a UDP based protocol which
allows the client to request a file one block at a time. As PXE booting requires a PXE server,
which is essentially an enhanced DHCP server, the PXE server offers all the information
The basic boot process can be expressed as follows: initially the network boot code gets
control after the basic POST checks have been performed and the hardware initialization has
been accomplished. When the network booting is included in the BIOS, a BIOS set-up option
will generally enable transfer of control to the network boot code, either instead of booting
from the hard drive or before hard drive booting is tried – this depends on BIOS settings
capability. In the case of a network card boot ROM, the boot ROM will be found during the
BIOS scan for card initialization ROMs and the ROM initialization code will pass to the
BIOS boot code to take control as soon as the booting process begins.
60
Once the network booting code has control, either it first sends out a DHCP or a
bootp request to the network to obtain an IP address, as well as other network card set-up
information, and gets the pathname of the boot file to load. PXE booting is similar, except
that PXE expects to get some PXE-specific information back in the DHCP query, including
the PXE version in use. Having obtained this information it initializes the network card, and
The first file retrieved over the network is generally either a second stage boot
loader (e.g., pxelinux in the case of a PXE boot), or a specially tagged kernel
(Etherboot/Netboot). A second stage boot loader will load a configuration file over the
network (such as a file from the pxelinux.cfg subdirectory in the case of pxelinux) to control
the next sequence. It will then use TFTP to download a kernel image, and optionally an initial
ram disk image. Once the kernel has been retrieved, and uncompressed, control will be
transferred to it, and the booting process can continue. Then X-Windows is loaded, the
preferred desktop environment (Gnome in our case) is set-up and configured along with
setting for video and sound. The environment is very similar to the set-up for thin client with
the exception that much of the processing ends up on the client rather than the server, thus
lowering the server CPU utilization to that of server/fat-client architecture, with all the
benefits of the centralized administration and management of thin client architecture. With
the use of diskless client architecture in this manner it is possible scale the system greatly.
With a well architected network infrastructure, diskless clients with adequate RAM
and processing capacity clients load within 30 seconds of a cold boot to login screen.
Typically a standard middle school has up to 250 workstations. With one Linux/LTSP server,
250 standard diskless clients and an optimized boot loader process this server can handle
61
typical loads with maximum utilization running well under ten percent. In order to reduce
server load at school startup we have developed a script which loads clients based on
predefined categories such as: administration, library, teaching labs, teacher computers and
student computers. Clients are prebooted and available for staff and students when required.
For example administration computers will boot up at 7:45am, library computers at 8:30am,
lab computers at 8:45am and so forth. Additionally each of the groups is staggered to startup
clients at 30 second intervals further reducing server loads. Further energy reductions are
maximized by shutting clients down at predetermined times. Users are able to startup any
Along with the boot process many other enhancements have been incorporated into
the diskless client image. With up to 300 clients loading off of one Ubuntu server it would be
assumed that load times would be sluggish at both boot up and application load, depending
on the number of clients active and those requesting application programs. There are many
factors that go into optimizing the system for large numbers of diskless clients. These include
ensuring the optimization of the network infrastructure; diskless client capacity and
enhancing the method key applications are loaded. As diskless clients are a hybrid model
incorporating the advantages of thin and thick clients it is possible to integrate features which
make diskless clients perform much better than true thin clients.
LTSP adds functionality to the system through a feature called local apps which can
allocate load from the server to the clients and is a form of distributed processing [58]. True
thin clients do not have the intrinsic processing capability or RAM to run applications client-
side. Diskless clients however, can take advantage of local app functionality. Simply put,
62
local apps allow system administrators to move some CPU intensive applications such as
Firefox and Open Office from the server to the client if they have the capacity. Provided that
client requirements are met, it is possible to make use of the functionality of LTSP to “hand
off” CPU intensive applications from the server to the client, thus greatly reducing the load
on the server. Ubuntu includes a package called "ureadahead" as part of the core server
system. This program records files that are loaded as part of the boot process and preloads
them automatically. As part of our customized image a custom ureadahead configuration was
created that includes the files used by Firefox and OpenOffice. Once a user has logged into a
client and is ready to launch one of these programs from the desktop a typical load time for
either Firefox 4.x or Firefox 3.x is four to five seconds, dramatically improving user
satisfaction. It is possible to preload other applications however we have noted that the
browser and word processor are the most heavily used applications. With the boot process
optimized and typical application load being greatly reduced it is possible to reduce the
Although this only touches on many of the enhancements made to optimize the
system it serves to point out that with all aspects of the implementation carefully thought out
it is possible for Linux diskless clients to perform as well as thick clients with the advantages
of thin clients. Although brief this provides some basis for the theory of operation, boot
process, local apps and client scheduling. It is important, however to define the basics for the
System Requirements.
The system requirements for the diskless requirements can be dealt with simply in
three categories: the network, the server and the diskless client. As the clients are heavily
63
dependant upon the server and clients utilizing localized processing, the network requires
sufficient bandwidth for handing off the kernel to the client as well as any applications which
can be run locally rather than server-side. Therefore, our experience has shown that the
server has to have gigabit connectivity across the backbone with a minimum of 100 megabits
to each client from the switch workgroup switches. This also assumes the server will have a
gigabit NIC while the clients have auto-switching 10/100/1000 NICS as indicated in Fig. 6 –
The server hardware components are designed to complement the Linux LTSP server
requirements and can easily support up to 300 diskless clients with the network specifications
noted above. The recommended server specifications for large diskless client implementation
64
• Tier 1 2U Server 2 way AMD Opteron 6128 HE (Dual CPU Sockets)
• 12 core CPU
Current large site server specifications recommend up to twelve core processors with
processes. Smaller sites with less workstation requirements can function with less capable
servers. Single server per site implementation is possible due to the processing capabilities of
the diskless client computers. Most true thin clients have very limited processing capabilities
and small amounts of RAM, however the hybrid diskless client incorporate enough
processing speed and RAM to run the operating system locally as well as most, if not all
applications. Fig. 7 – Diskless client server utilization, illustrates the CPU utilization for one
server and approximately 250 diskless clients. As a note, the large spike in utilization at
22:00 hours is the backup process, which takes place throughout the night. Peak utilization
time is from 8:00am to 3:00 pm and utilization rarely gets above 40% under a full load. Since
65
this initial snapshot was taken two years ago recent tests and upgraded servers have seen
The cost for each diskless client is approximately $250 for each unit, including
keyboard and optical mouse. Clients become appliances at this cost, where the primary
components are a case with low wattage 80Plus power supply, a motherboard and RAM. The
With the combination of a robust and fast local area network, capable Linux/LTSP
server and powerful, low cost diskless clients it is possible to build relatively support-free,
66
low cost and sustainable server client architecture. All of this is of great significance to IT
departments and users however, ancillary is the opportunity for diskless clients to greatly
reduce energy consumption, greenhouse gases (GHG) and carbon dioxide (CO2).
In performing its primary business, which is the instruction of students, the district's
annual budget is approximately $71m with $1.34m spent on maintenance, and $1.06m spent
on utilities including electricity, natural gas and other fuels. This is further broken down into
$.491m being spent on electricity, $.425m on natural gas, $.0134 on fuel oil, $.009m on
propane with $.13m spent on water. Saying this, the district contributes significantly to
(equivalent) for natural gas and 5.8 GWh/yr for electricity. Fig. 8 – School District No. 63
usage. The emphasis is upon the electrical energy used by the district.
years. Early projects initiated by the Facilities department included, a detailed lighting,
heating, ventilating and air conditioning (HVAC) analysis for potential energy savings, while
becoming a BC Hydro PowerSmart partner in 2002 [59]. The district embarked on numerous
lighting upgrades, including installing energy efficient T8 and T5 fluorescent lamps with
electronic ballasts in all buildings, upgraded various energy controls and direct digital control
(DDC) system to improve control and comfort enabled by the HVAC units. Many of these
systems utilize the data network and provide control software, data collection and analysis.
The district central office - incorporates a small data center currently housing eight
servers which are the result of previous consolidation and vitalization. Three years ago there
was no district redundancy or fail-over site. Each of the eight elementary schools and three
middle schools had two to four aging tower-based Windows servers which functioned as
DNS, File & Print servers. The three secondary schools have an average of six to seventeen
tower-based servers each running various versions of Windows server software performing
Additionally the eight elementary schools had an average of approximately 100 desktop
personal computers per school running Windows XP for a total of 780 units. Each of the
three middle schools each had an average of approximately 250 plus desktop personal
computers also running Windows XP for a total of 790 units. The three secondary schools
have an average of 300 desktop personal computers also running Windows XP for a total of
900 units. Each of these workstations had a minimum 120 plus Watts per station not
68
including monitors. Approximately fifty percent of existing systems utilized 17” CRT
monitors.
The district technology plan called for a ninety percent replacement of the Windows-
based desktop client systems with Linux diskless client systems. This included replacing
existing Windows-based server operating systems and associated software with Linux and
LTSP. In School District No. 63 the focus on energy savings was through the use diskless
clients which are fully managed through scripts and utilities implemented on each school
server. The clients are “lower” power computers running AMD Opteron or Athalon
processors with 1-2 Gb of RAM. They consume slightly more power than a conventional thin
client but much less than a personal computer. Table 5 – Client energy consumption
The district energy manager undertook an energy audit which provided energy data from
our inventory of approximately 2450 personal computers. The BC Hydro IT Survey Totals -
Strategic Energy Management Report [60] found that the current inventory of Windows-
• 40% of old computers left on 24/7 with no sweep, sleep, or power management
enabled
69
• 80% of power used when idle due to larger inefficient power supplies and no power
management implemented
• Average active operating time was averaged at approximately 3.6 hrs/day for 192
days/year
Typical power consumption values for thin clients was taken from the Desktop Energy
Consumption Report, by Wyse Technology Inc. [61] while PC and diskless client analysis
was accomplished using a WattsUp Pro ES Power Meter and the results were averaged
across all schools [62]. The results of this shows the energy savings on the eight elementary
and eleven middle schools implemented are show in a school-by-school comparison in Table
70
Table 6. School District No. 63 diskless client energy savings
ECM Savings
No. of Energy
School Diskless Clients Savings
per School kWh/yr
Elementary Schools
Brentwood 72 28,685
Cordova Bay 70 27,948
Deep Cove 91 36,327
Keating 87 34,737
Kelset 98 39,117
Lochside 80 31,935
Prospect Lake 64 25,556
Sidney 108 43,116
Total Elementary Schools 670 267,422
Middle Schools
Royal Oak 211 84,266
Bayside 224 89,450
North Saanich 221 88,500
Total Middle Schools 656 262,216
This total accounts for approximately ninety percent of the total workstations per
school with a conversion rate of approximately 9:1 for diskless clients to Windows PC
workstations. The savings generated from implementing the diskless client technology has
electrical rate is current at $.08 per kWh/year. The estimated savings per year of the diskless
client implementation is 529,637 kWh/year * $.08 for a total yearly savings of approximately
$42,370. Adding to this the additional 750 workstations yet to be installed provide for a total
of 2076 diskless clients for an approximate savings of 829,205 kWhr/year * $.08 for a total
estimated annual savings of $66,366 per year. With the initial cost of the diskless clients
71
being approximately $519,000 and the anticipated accumulative electrical saving of $331,682
over a five year period indicates that the cost overall for all clients can reduced to $187,317.
With an anticipated end-of-life for diskless clients estimated to be six to eight years due to
the few moving parts and processing requirements this makes the clients almost cost neutral
in energy savings alone. This is a significant savings, not only in client cost, but more
importantly in the reduction of GHG and CO2. With an estimated reduction in energy
consumption for IT client costs the district will reduce total electrical energy consumption by
These significant cost savings are not only the result of using low watt, energy efficient
power supplies in the diskless clients but highly managed power and power down procedures
through scripts and cron jobs. Additionally the image incorporated the Linux kernel “laptop
mode” for power management, black and blank screen savers and are scheduled to turn on
and off on through a managed schedule thus further reducing energy consumption when
clients are inactive [63]. A schedule regarding power up and turn off times is worked out
with the school through scripts and Command Run ON (cron) jobs. Each location or purpose
of computer is individually addressable through the database. Typically, the office and staff
computers are turned on first, and then the labs and library are turned on shortly before
students are expected in the lab. Users can also manually turn machines on at any time.
Clients are turned on using the wake on LAN protocol, which uses a special packet sent over
the network, addressed to the MAC address of the machine. Users are warned that the
computer is going to be turned off, and are provided a chance to leave the machine on. 15
72
With these features in place overall client energy consumption has been reduced by
approximately fifty percent per year. Adding the replacement of CRT monitors with LCD
monitors further reduces this to approximately 70 percent savings on client computing. All
diskless clients utilize 80 Plus power supplies and all LCD monitors are Energy Star
compliant.
Summary.
Diskless client technology is not new technology; however the manner in which it is
being implemented and deployed is new, enabling accelerated video and streaming sound
greatly enhancing the user experience. Being able to scale very low cost, diskless clients puts
the support emphasis not on the users PC, but on the use of the software and applications.
The diskless clients we have deployed and continue to implement, allow for multi-language
support, ubiquitous access not only across the network, but securely through the internet with
the use of software such as FreeNX. Additionally we are seeing greatly reduced support and
maintenance and significant energy savings. The emphasis is upon continued development,
enhancement of the image and improved software support. There is still work that can be
done to improve this model and it may only be an interim step until web operating systems
become fully developed making the need for local servers redundant. In the meantime I
believe this computing model to be the most elegant and cost effective one to provide large
73
CHAPTER V
School districts in British Columbia, and elsewhere, are faced with increasing
pressures resulting from decreasing budgets, increased support and maintenance for their
need to maintain software currency for student learning and changes in educational
paradigms such as 21st century learning and personalized learning. As well the inevitability
for the end-of-life of Windows XP in 2014 and an aging hardware inventories put increased
pressure making for difficult choices in the future. Typically many school districts have
operated a decentralized model where schools manage capital and operating budgets which
are loosely controlled by the central board office. The consequence is; many schools
purchase off-lease or used computers which already near end-of-life are less reliable and
District No. 68 (Nanaimo) are faced with the considerable task and expense of upgrading
In early 2011, IBM was engaged by the School District No. 68 board of trustees to
provide consultative services which would evaluate the district current technology with
recommendations for remediation [64]. The report was presented to the district trustees with
from school to school, as some schools have computers in classrooms, while other
74
schools only have computers in labs. This has resulted in have and have not schools”
“Although there is district level direction for the purchase of computers through
Computers for Schools, given the schools are primarily responsible for the purchase
technology related budgets are based on maintaining the status quo from previous
term plan, and having the various departments such as Technology, Program,
establish annual budget requirements based on a long term plan enterprise rather
The Nanaimo school district is a mid-sized district serving 14,500 students with 31
elementary schools, seven secondary schools and two alternative schools [65]. With 40
schools and approximately 5,000 computers district-wide that are largely three to eight years
old, this presents a significant challenge to the district. With the majority of these computers
still running Windows XP, security issues such as virus infections [66] and rootkit infections
[67] are getting progressively more problematic as support for the aging operating system
nears end-of-life and users fail to upgrade with the latest security patches. The
recommendations from IBM include: spending approximately five million dollars over the
next five years upgrading client workstations. Unfortunately Nanaimo does not have the
capital dollars to dedicate to these upgrades and leasing is generally not an option for school
districts due to pressure on operating budgets. The report highlights the problem that is
75
endemic with school districts all over British Columbia; they must upgrade aging equipment
if they are to provide current and relevant technology while not having the financial
resources to make it possible. Unfortunately the experience in Nanaimo is not new or unique.
Many school districts across British Columbia face similar challenges with a lack of
resources and “the lack of a district, centralized technology plan” [64], as the traditional
performed by post-secondary institutes; the three primary failures of thin client systems were
considered to be:
Over the past decade a variety of approaches were researched and developed to solve
these issues with thin clients such as: reducing user expectations [29], Thin-client InterNet
Computing THINC graphics enhancement [31], adding a driver abstraction layer providing
GPU emulation [32], analysis and improving client acceptance strategies [38], and so forth
with a great deal of optimism but very little tangible results. While there was some minor
progress with providing a richer graphic experience there was very little advancement for
application support and network latency issues. Although thin clients have the promise of
lower TCO, reduced support and reduced energy consumption they have proved
disappointing as the core disadvantages have not been overcome. Windows-based thick client
76
systems have proved unsustainable primarily due to cost constraints and support issues. An
alternative must be found or existing systems will continue to decay and become irrelevant.
District No. 63 (Saanich) have show that the utilization of Linux diskless client in
solution which is affordable, sustainable, maintainable and very energy efficient. It provides
curricular requirements. The diskless client solution being implemented employs primarily
open source software including Ubuntu [68] as the core distribution, Linux as the base
operating system, LTSP terminal services [12] and Gnome desktop environment [69]. Table 7
- Typical diskless client open source applications, lists the base open source software
The standard open source software compliment provides schools with much of the
provide additional functionality using commercial software where open source falls short.
Some commercial software is available for native Linux; such as Smart Technologies
Notebook software [70], while others are not. When possible, district analysts have been able
to get many closed source Windows-based applications operational, using Wine emulation
software [71]. Wine is included with the standard Ubuntu image and makes use of
78
compatibility layer functionality, providing support for Windows-based programs on Linux-
based systems. The School District No. 63 implementation uses Wine to support software
such as: All the Right Type typing tutor [72], Math Makes Sense [73], Google Sketchup 8
[74], Smart Notebook software and other programs. With Wine it is possible access software
emulation. Many of these programs are utilized in special education providing applications
for students with learning disabilities. When it is necessary to offer specialized programs
such as text-to-speech software like Kurzweil [75], speech-to-text software such as Dragon
NaturallySpeaking [76], or literacy software for example Clicker 5 [77] the district provides
Windows-based laptops to provide systems for students with special needs. This ensure that
the few users with the specific commercial software required for special education needs can
be completely accommodated.
For a typical school utilizing 120 workstations, 110 of these make use of Linux
diskless clients while the remaining ten will be Windows-based laptops. This 11:1 ratio
provides a fusion of diskless Linux clients and Windows laptops for elementary schools
preserving the centralized management. Fig. 9 – Typical Elementary school diskless desktop
snapshot provides a graphical view of the customized Gnome desktop staff and students see
when logging into the system. Using this hybrid approach offers schools the specialized
software they require to provide educational outcomes for the present while anticipating the
future.
79
Figure 9. Typical Elementary school diskless desktop snapshot
Since the initial pilot at KELSET elementary school in 2008, the Linux diskless client
image has seen continual development and refinement. With the primary focus on improving
the image, rather than providing support and maintenance, many enhancements have been
brought to the system. As with thin client technology, access to users desktop is available
from any client within the school as well as their customizations, files, documents,
bookmarks and other data. Moreover, user’s desktop environments are available securely
from home through NoMachines NX client technology [78] for which the benefits are
significant. Previously when staff and students desired to work on word processing or other
files from school servers, they were required to attach them to an email message or copy to a
flash drive. Predictably this was problematic and resulted in duplicate files, overwrites and so
forth. With secure remote access via the NX client to the entire desktop including
applications and data files, users no longer have to be concerned about redundancy issues.
80
Providing remote access to school-based programs has resulted in staff, students and parents
not having to purchase software which is in use at the school. Offsite school login is
available from Windows, MAC or Linux computers, thereby facilitating ubiquitous access
and approaching the kind of “calm computing” which Dr. Mark Weiser anticipated in the
possible to implement multilingual capabilities within the system. When users login to the
system they are presented with the choice of pertinent languages including French and
English. After logging in users menus and applicable applications such as Firefox, and
OpenOffice are also available in the language of choice. This is a huge benefit to English as
with Computers (iTalc) desktop management system for teachers, using the Festival text-to-
speech synthesis software system pervasively, improving print support and a host of other
improvements.
incorporating Linux diskless client technology into schools is, onsite support and
time was previously spent in simply maintaining the system, the diskless client model is
centered on continual improvement. Diskless clients become essentially appliances and can
be managed out by site staff. Fixes or updates are performed centrally and affect all users
with even major upgrades being affected within minutes, rather than days greatly minimizing
system downtime.
81
Other benefits include minimal effect of virus, rootkit or other malicious infections as
Linux is not targeted to the extent that Windows-based systems are. Although users have
some flexibility in their desktop environment they cannot download and install software and
control over the system remains the responsibility of the central IT department. The common
experience of most IT departments is reflected in the report from IBM to the Nanaimo school
district where “the gaps identified in the engagement reports are mainly a result of the IS
Department spending the majority of their time in reactive mode …” [64]. Rather than
limiting IT’s ability to provide services and support due to funding and other factors, diskless
client technology has provided a modern and responsive school-based information system for
a very low cost. The following quote from School District No. 63s Superintendent of Schools
underscores user and acceptance from students, parents and staff [79]:
Thanks go to everyone who is making the new thin (diskless) client (Linux) system
work so well, particularly to our Information Technology Director Gregg Ferrie, our
teacher leader Brock Simmonds, and our knowledgeable IT crews in the board office
and in the schools. Teachers’ uses of modern web-based solutions for sharing
information with children and colleagues is one of the real benefits of our new open-
source web-based approach. The positive effects on teaching and learning make the
As with any significant organizational change there are intrinsic barriers and
resistance. The success of the Linux diskless client implementation at both School District
No. 73 (Kamloops/Thompson) and School District No. 63 (Saanich) was not achieved
82
without overcoming many difficulties and challenges. For any school district planning to
implement significant technological change there are several potential impediments and can
• insufficient funding
• poor communication
• inadequate planning
• insufficient training
• inadequate support
With the diskless client implementation at School District No. 63 (Saanich), these
process. Immediately after the author was hired in early 2008 a comprehensive technology
plan was developed, based on experience gained from the Linux diskless client
regarding training and support. Once complete and approved, presentations were made to
senior district staff, followed by school administration staff and school teaching and support
staff. All questions, concerns and potential problems were discussed and responded to.
83
forward with the pilot at KELSET elementary. During the pilot, feedback was garnered from
school, admin and executive staff on a weekly basis and any concerns As a consequence the
image was improved and developed based on teacher and student requirements. To date, 11
schools have been successfully implemented with three more secondary schools to be
converted over the subsequent eight months. The will span the full scope of schools in the
Conclusion.
Linux diskless client technology although not a panacea combines many of the
advantages of both thin and thick client technology. It is a hybrid technology which has
proven successful when risks are anticipated and well managed. School districts are under
increasing pressure from budget constraints, curricular changes, technology advances and
environmental concerns and have not found satisfactory solutions to providing school-based
technology is relevant for the present and anticipates the future of web-based operating
systems and applications. It accomplishes this for less than one half the cost of desktop
providing ubiquitous and pervasive access. Furthermore school districts faced with the
challenge of having to reduce lowering GHG and CO2 levels have an elegant, effective
solution. Most importantly, diskless client technology can achieve workstation energy
districts. Linux diskless client technology provides the first step toward utility grade
computing and is enabling technology that should be given more than casual consideration
considering the many issues that face school districts in British Columbia.
84
REFERENCES
15-Oct-2009.
http://www.bced.gov.bc.ca/dist_learning/.
[3] J. Seely-Brown, “Minds on Fire: Open Education, the Long Tail, and Learning 2.0,”
[4] M. Prensky, “Digital Natives, Digital Immigrants Part 1,” On the Horizon, vol. 9, no. 5,
[6] Ministry of Education, “21st Century Learning,” Ministry of Education, 2011. [Online].
Available: http://www.bced.gov.bc.ca/dist_learning/21century_learning.htm.
[7] A. Turker, I. Gorgun, and O. Conlan, “The Challenge of Content Creation to Facilitate
http://www.w3schools.com/browsers/browsers_os.asp.
Available: http://windows.microsoft.com/en-us/windows7/products/system-
requirements.
http://www.faronics.com/enterprise/power-save/.
85
[11] J. Kavanagh, “Resistance as Motivation for Innovation: Open Source Software,”
Communications of hte Association for Information Systems, vol. 13, no. 1, pp. 615-628,
2004.
[12] LTSP, “Linux Terminal Server Project,” LTSP, 2011. Available: http://www.ltsp.org/.
education.
[15] Wine, “WineHQ - What is Wine?,” Wine User Guide, 2011. Available:
http://www.winehq.org/docs/wineusr-guide/what-is-wine.
[16] Samba, “Samba - opening windows to a wider world,” SAMBA, 2011. Available:
http://www.samba.org/.
[17] BC Hydro, “Faronics Power Save Power Save Power management software - BC
http://www.bchydro.com/ecatalog/eCatServlet?cmd=product_details&catId=179&prdId
=2265&pid=&rid=&eid=.
[20] Plug Load, “Ecos Plug Load Solutions - 80 Plus,” Plug Load Solutions, 2011.
Available: http://www.plugloadsolutions.com/80PlusPowerSupplies.aspx.
http://en.wikipedia.org/wiki/Dumb_terminals#Dumb_terminal.
86
[22] LINFO, “Thin client definition,” The Linux Information Project (LINFO), Feb-2006.
Available: http://www.linfo.org/thin_client.html.
http://technet.microsoft.com/en-us/library/cc727363%28WS.10%29.aspx.
Available: http://news.cnet.com/Ellison-resurrects-network-computer/2100-1001_3-
233137.html.
[25] J. Golick, “Network computing in the new thin-client age,” netWorker, vol. 3, no. 1, pp.
[26] M. Weiser, “The Computer for the 21st Century,” Scientific American, no. 1991, 1991.
[28] A. Lai and J. Nieh, “Limits of wide-area thin-client computing,” in Proceedings of the
computer systems - SIGMETRICS ’02, Marina Del Rey, California, 2002, p. 228.
centralized, graphical computing environment,” ACM SIGCSE Bulletin, vol. 37, no. 3,
[31] R. A. Baratto, L. N. Kim, and J. Nieh, “THINC: a virtual display architecture for thin-
client computing,” ACM SIGOPS Operating Systems Review, vol. 39, no. 5, p. 277, Oct.
2005.
87
[32] D. De Winter et al., “A hybrid thin-client protocol for multimedia streaming and
Network and operating systems support for digital audio and video - NOSSDAV ’06,
[33] R. Klein, “Gore’s `Truth’ turns up the heat,” The Boston Globe, 24-May-2006.
Available:
http://www.boston.com/news/nation/washington/articles/2006/05/24/gores_truth_turns_
up_the_heat/.
[35] D. Przybyla and M. Pegah, “Dealing with the veiled devil,” in Proceedings of the 35th
[36] E. Davis, A. Parker, B. Gray, and O. King, “Green Benefits Put Thin-Clients Back on
Studies in Thin Client Acceptance,” Special Issue on ICIT 2009 conference - Applied
[39] R. Braddock, “Linux in Education - Distributions, Dilemmas, and the charm of LTSP.”
88
[40] Wikipedia, “Ubuntu (operating system),” Wikipedia, 2011. Available:
http://en.wikipedia.org/wiki/Ubuntu_%28operating_system%29.
[41] M. R. Ritschard, “Thin clients: make them work for you,” in Proceedings of the ACM
SIGUCCS fall conference on User services conference - SIGUCCS ’09, St. Louis,
[42] M. R. Ritschard, “Thin clients: the key to our success,” in Proceedings of the 34th
[43] B. O’Donnell and R. Perry, “Thin Computing ROI: The Untold Story.” IDC, Nov-
2005.
[45] Microsoft, “Microsoft Support Lifecycle,” Microsoft Help and Support, 2011.
Jul-2011].
[46] TriCity, “EDITORIAL: A ’net loss for education,” Tri-City News, 19-Apr-2011.
Available:
http://www.bclocalnews.com/tri_city_maple_ridge/tricitynews/opinion/120140219.html
[47] B. Penner, “Bill 44 — 2007: Greenhouse Gas Reduction Targets Act,” Government of
89
[48] K. Chung, P. Zimmerman, and S. Chandra, “What is thin client (lean client)?,”
http://searchnetworking.techtarget.com/definition/thin-client.
[49] Whatis, “What is fat client (thick client)?,” Whatis.com, Jan-2006. [Online]. Available:
http://searchcio-midmarket.techtarget.com/definition/fat-client.
[51] G. Ferrie, “School District No. 63 (Saanich) - District Technology Plan 2009-2011.”
[52] SD63, “Regular Meeting of the Board of Education of School District No. 63
[53] Ubuntu, “Ubuntu 8.04.4 LTS (Hardy Heron),” Ubuntu, 2010. Available:
http://releases.ubuntu.com/8.04/.
[54] DisklessNode, “Diskless node - Wikipedia, the free encyclopedia,” 2011. Available:
http://en.wikipedia.org/wiki/Diskless_node.
[55] Ubuntu, “Ubuntu Manpage: ltsp-update-kernels - Updates the LTSP server kernel,”
http://manpages.ubuntu.com/manpages/lucid/man8/ltsp-update-kernels.8.html
90
[58] Ubuntu, “Ltsp-Local-Apps,” Ubuntu Wiki, 2008. Available:
https://wiki.ubuntu.com/Ltsp-Local-Apps.
[59] BC Hydro - Power Smart, “BC Hydro - Power Smart,” Power Smart, 2011. Available:
http://www.bchydro.com/powersmart/.
[60] J. Armstrong, “BC Hydro IT Survey Totals - Strategic Energy Management Report.”
https://www.wattsupmeters.com/secure/products.php?pn=0.
[63] R. Sarraf, “Laptop Mode / Laptop Mode Tools FAQ,” Laptop Mode Tools, 2010.
Available: http://samwel.tk/laptop_mode/faq.
[64] A. Berger, C. King, and D. Eli, “School District No. 68 - IBM Part 3 IT Strategic Plan
[65] SD68, “SD68 Schools,” School District No. 68 (Nanaimo/Ladysmith), 2011. Available:
http://www.sd68.bc.ca/Education/schools.asp.
http://vancouver.24hrs.ca/News/2009/01/14/8015156-sun.html.
Available:
http://www.computerworld.com/s/article/9218722/Windows_XP_PCs_breed_rootkit_in
fections.
91
[68] Ubuntu, “Ubuntu - Server Edition,” Ubuntu Server Edition, 2011. Available:
http://www.ubuntu.com/server.
http://www.gnome.org/.
http://smarttech.com/us/Support/Browse+Support/Download+Software/Software/SMA
RT+Notebook+collaborative+learning+software/SMART+Notebook+software/SMART
+Notebook+for+Linux.
[72] Ingenuityworks, “All The Right Type 3 Plus,” Ingenuityworks, 2011. Available:
http://ingenuityworks.com/catalog/product_info.php?cPath=41_45&products_id=50.
[73] Pearson, “Math Makes Sense,” Pearson Math Makes Sense, 2011. Available:
http://www.mathmakessense.ca/.
http://sketchup.google.com/download/.
http://www.kurzweiledu.com/default.html.
http://www.nuance.com/dragon/index.htm.
http://www.cricksoft.com/uk/products/tools/clicker/home.aspx.
92
[79] K. Elder, “Update form the Superintendent.” School District No. 63 (Saanich), Nov-
2010.
93