Chirag Dekate, Thomas Sterling
Department of Computer Science & Center for Computation and Technology
Louisiana State University, Baton Rouge, Louisiana – 70803, U.S.A.
Daniel Eiland, Ravi Paruchuri
Center for Computation and Technology, Louisiana State University
Baton Rouge, Louisiana - 70803, U.S.A.
Keywords: Optical Networks, High definition, Distance learning, High performance computing.
Abstract: Although distance learning has a history that spans many decades, the full opportunity that is implicit in its
exploitation has not been fully realized due to combination of factors including disparate experience
between it and its classroom counterpart. However current and emerging technologies are helping overcome
this barrier by providing significantly better interaction among the individual participants, thereby opening
new avenues for knowledge dissemination. LSU in collaboration with five other institutions has developed
effective methods that greatly extend the educational opportunities through combination of advanced
technologies and educational methodologies. LSU and its partners have tested these technologies in real-
time over the last two years. While further improvements are needed, this activity represents the current
state of the art in technologies utilized and the quality of content and experience delivered. The distance
learning initiative undertaken by LSU and its partners is driven by a vision for education, which aims to
deliver expert & top-quality educational content to locations irrespective of their economic or technological
Distance learning is as effective as the means to
emulate the local classroom experience is successful.
The motivation to realize the potential of distance
learning is to address critical challenges being
imposed by the realities of economics,
demographics, cultural diversity, and the rapidly
increasing wealth of possible topics and specialties.
Of particular importance to the work undertaken by
the authors represented in this paper is the
opportunity to dramatically increase the choice of
educational pursuit by students independent of
geography, financial circumstances, and stature of
the educational facility. Only by aggressively
addressing this challenge of choice can the full
potential of every student be realized for the benefit
of themselves and society as a whole.
Although distance learning has a history that
spans many decades through continuing education
programs and more than two decades through low
grade video, the full opportunity that is implicit in its
exploitation has not been fully realized because of
disparate experience between it and its classroom
counterpart. Recent advances in high definition
digital video over Internet have opened new vistas in
the quality of the distance learning experience with
advances in both visual quality and narrowing of
round trip latency for realistic dialog. As will be
discussed in the next section of this paper, a
transformative change in the future of higher
education may be achieved through the effective use
of a combination of these and other technologies for
a synthesis of the distributed teaching and learning
Louisiana State University in collaboration with
five other institutions has undertaken an important
experimental program in distance learning to
develop effective methods greatly extending
educational opportunity through the combination of
Dekate C., Sterling T., Eiland D. and Paruchuri R. (2009).
In Proceedings of the First International Conference on Computer Supported Education, pages 152-159
advanced technologies and educational
methodologies. Specifically, LSU has employed a
range of real time video qualities including the use
of uncompressed high definition for low latency
over long distances in real time. It has combined this
with an active web site for a multitude of material
dissemination techniques including on-demand
video download to provide a viable and effective
educational experience that rivals that of local
classroom teaching. To exercise these technologies
and methods in a real world distance learning
setting, LSU has developed a new first-year graduate
course: “High Performance Computing: Models,
Methods, and Means” which is intended as an
introductory treatment of the multi-subdisciplinary
area of high performance computing for the largest
diversity of student interests. This course was also
opened to advanced undergraduates as the
prerequisites were minimized for maximum
participation. This course was received in real time
by four other campuses for each of two cycles in the
Spring of 2007 and 2008.
The purpose of this paper is to present the
methods employed, the resulting experiences, and
the advances still required to fully achieve the dream
of the promise of distance learning as a mainstream
strategy. The next section of this paper discusses the
details of this promise that provides the long-term
motivation of our work and its importance to the
future of college education in the US. Section 3
provides an overview of CSC-7600, the computer
science course developed in part to develop the
methods used in conjunction with the advanced
distance learning technologies. Section 4 provides a
comprehensive description of the array of
technologies and their synthesis used for this
experiment. Section 5 then describes the strengths
and weaknesses experiences from the perspective of
the educational process. Finally, Section 6 briefly
discusses additional extensions to the techniques
employed that are being pursued to improve the
overall educational experience in response to our
initial results. This work has been funded in part by
the National Science Foundation (NSF) and by the
LSU Center for Computation and Technology (LSU-
The role of distance learning combined with the
advanced technologies and methodologies that
enable it will have a transformative impact on how
higher education is accomplished in the 21
Century. In the US diversity in demographics,
geographical economics, and the effects of world
competition is challenging effective delivery of
quality education, especially in rapidly changing
Science, Technology, Engineering, & Mathematics
disciplines, including computer science and
engineering. Such fields demand expertise and
experience in a diversity of sub disciplines for
effective education at the college level. These areas
are in constant change and require faculty who are
actively participating in related research to remain
current. Unfortunately, these specialties are under-
represented or entirely absent at many or even a
majority of US universities and smaller colleges. As
a consequence, students at these otherwise fine
institutions are deprived the opportunity to benefit
intellectually in these areas. While many such
students may not ultimately undertake such studies
in any case, they are deprived the fundamental
opportunity of choice. At a time when many young
students are still in a stage of personal development,
being deprived of such choice predetermines that
outcome of their potential evolution and overly
constrains the promise of their professional
A second critical factor is the cost of education.
This cost includes the development of new courses
as well as those advanced courses which must be
constantly updated. In recent years, the cost of a
student-seat-hour has skyrocketed in the US only
partially offset by increased student tuition raises.
Even these increased tuitions can prove a severe
inhibitor to economically challenged students in
their choice of college or even whether or not they
undertake higher education at all. While many
factors contributed to these escalating costs, one
major factor is the significant increase in the salaries
of faculty in fields for which there is a strong
industrial competition. High technology and applied
science fields are among these.
The application of advanced digital multimedia,
communication, and computer technologies may
alleviate and even transform higher education
through distance learning. When combined with
innovative methodologies of teaching, these
emerging technologies may deliver:
Higher quality education by making courses
available from national experts in specialized
fields, and investing more resources in to the
development of each such course,
CSEDU 2009 - International Conference on Computer Supported Education
Lower cost by amortizing the course
development across a wider range of institutions
and students,
Greater choice for students in pursuit of the
professional and personal growth through
access of the widest possible number of
excellent courses being delivered at their local
institution, and
Increase the number of highly specialized
courses by amortizing such courses over an
aggregate student body distributed across a
large number of institutions.
This last opportunity is a subtle but important
one. Often a course, which could be taught, is not
because enrolment is too few. Although the material
may important, the desire of the professor to teach it
is high, and the interest of the few students who sign
up also high the economics simply cannot permit the
realization of such a course. The exploitation of
distance learning may create a distributed student
body of sufficient capacity to justify teaching a
course and bringing new diversity and quality to
education. Ultimately all of these factors provide
unprecedented choice for students of the widest
variation in circumstances. This may be true
internationally as well with technologies spanning
national boundaries and language barriers reduced
through a worldwide community.
To achieve the promise of this vision for
freedom of choice and opportunity in higher
education, advanced technologies and pedagogical
methodologies have to satisfy key requirements:
Point to point high bandwidth digital
communication that is real-time, reliable, stable,
and bounded in cost for practical application,
Broadcast capability n-way to n-way,
High definition video streaming,
Low latency video and audio for real-time
dynamic interaction,
Active web site for access to all course
information and materials including but not
limited to course slides, schedules, problem
sets, reference materials, tutorial notes,
homework solutions, and wikis for frequently
asked questions,
On-demand downloadable videos of lectures
and recitation sections,
Self-tests and quizzes for frequent evaluation of
Back channel communication for set up and
management of networked sites, as well as
continuation of lectures with degraded
resources, and
High quality course material crafted with the
recognition of the strengths and limitations of
the media being used.
The remainder of this paper describes the
methods and experience of one consortium of
universities to address these challenges for a new
graduate course in high performance computing
provided in high definition video among multiple
campuses in the US and one in Europe.
LSU has developed a new course with the express
purpose of teaching it via distance learning using the
advanced technologies described in the next section.
CSC-7600 – High Performance Computing: Models
Methods and Means – has been offered as a first
year graduate course and advanced undergraduate
course. It has been so structured to serve four key
professional goals including:
1. Computational sciences – for students who wish
to focus on other fields that require the use of
high performance computing as a tool to
achieve the goals of the science or engineering
discipline being pursued,
2. Research in Computer Systems – for doctoral
students in electrical engineering and computer
science who wish to conduct research in this
and related fields of study,
3. Hardware & Software Developers – for future
engineers pursuing positions in industry
involved in the design of hardware or software
systems associated with HPC, and
4. Systems Operations – for future managers and
administrators of supercomputers and their
Because of the diversity of professional goals
and disciplines from which participating students
may come, this course required a minimum of
prerequisites including: user familiarity with a Unix-
like environment (e.g., Linux) and programming
experience with the C programming language.
Wavers were provided in most cases for those
without C background but with experience with
other comparable languages such as Fortran. No text
book was used but reference material from a number
of sources (Sterling, 2003; Gropp, 1994; Chandra,
2001; Hennessy & Patterson, 2003; Galvin, Gagne
& Silberschatz, 2005) was made available through
the course web site, some of which was developed
expressly for the course. All lectures were taught
using slides in electronic form. These were
developed using Microsoft PowerPoint and
disseminated via the web site in .ppt and .pdf
formats. The availability of slides downloaded to
remote sites early was a risk mitigation factor that
allowed the lecture to proceed even if the video link
was disrupted by using the back channel conference
call channel. It also allowed a higher quality remote
presentation of slides that was achieved through live
video capture and distribution of the same visual
material. Live demonstrations were done this way,
however that was not always sufficient for the
The one-semester curriculum was taught over a
fifteen week period including exams and holidays
and comprised six major parts:
A. Introduction and Throughput Computing
B. Clusters and Message Passing
C. Shared Memory Processing
D. Parallel Algorithms and programming
E. Operating Systems
F. Visualization and Advanced Tools
The course partitions the concept space for HPC
into three classes of parallel processing, two of
which are commonly referred to and a third essential
to a correct representation as it relates to scalability,
programming models, and system architectures. The
cross cutting theme is performance and throughout
the course performance implications are examined
as well as skills are developed using tools for
evaluating performance. The first part explores the
simple but important form of capacity or throughput
computing that exploits concurrent work stream of
independent and unrelated user jobs. The class of
architecture employed for this is a workstation farm
and loosely couple clusters. Condor is taught as a
framework for controlling this ultra coarse grain
parallelism and simple means of measuring
performance are introduced. Weak scalability is
employed to increase performance by increasing the
number of jobs as the system size is increased.
The second part provides in-depth coverage of
the very important message-passing execution model
used with commodity clusters and MPPs and
programmed with MPI. Collaborative computing is
offered the domain of weak scaling that applies
multiple processing elements to a single parallel
task. As the system scale is increased, so is the size
of the application problem yielding more work to do
and maintaining a constant level of granularity. The
third part covers the most difficult form of HPC,
capability computing in which a single problem of
fixed size is able to reduce its time to execution
through increased system scale. OpenMP is used on
a shared memory system to represent capability
computing and provide a programming model.
The fourth part is dedicated to reinforcing the
lessons of the previous three by investing substantial
time in additional programming experiences for
different classes of problem algorithms. The use of
instrumentation tools such as Tau and PAPI are
developed to measure performance improvements
and evaluate effective scalability. The course
considers four sources of performance degradation
including overhead, latency, contention, and
starvation and shows how they may be addressed
through different techniques.
The fifth and sixth parts focus on system
software and methods for using supercomputers to
practical advantage. This includes the important
domain of scientific visualization for conveying the
meaning of the resulting data to the end user. The
course included hands-on experience through
examples, exercises, and projects using a cluster of
SMP nodes dedicated to the course.
Technology is the key component to enable teaching
and distribution of course materials without any
geographical barriers. The HPC course was
distributed to different sites using different
technology methods for each site. The technology
method used were uncompressed HD streaming,
compressed HD streaming using a Polycom HD
device, Access Grid and SD streaming using Ncast
stream engine. The type of technology used at each
site was determined based on the available resources
at that site.
CSEDU 2009 - International Conference on Computer Supported Education
Table 1: Distance learning technologies used by different
participating sites.
Media Format Sites
Uncompressed HD Video
Masaryk University, Brno, Czech
Compressed HD Video
University of Arkansas –
Fayetteville, Arkansas
Access Grid
Louisiana Tech University –
University of Arkansas – Little
Rock, Arkansas
Web streaming (NCast) Back up option and also used for
recording of lectures to enable
post reviewing for students
Uncompressed HD video streaming: This
technology was used to stream uncompressed
HD video of the HPC course to Masaryk
University, Brno, Czech Republic over a private
10Gb optical network using a open source
application called Ultra-grid. Each HD stream is
1.5Gbps. Multiple optical networks had to
bridge connections and make allocations to
make the communication happen between LSU
and Masaryk University over 10Gb optical
network. The network partners were Louisiana
Optical Network Initiative (LONI), National
Lamda Rail (NLR), StarLIght and CESNET.
Two high-end workstations one to send and the
other to receive video were deployed at both
sites. The sender workstation consists of video
capture components and a 10Gb network
interface. The receiver workstation consists of
an Nvidia graphics card with a DVI out that is
capable of displaying HD resolution
(1920x1080) and a 10Gb network interface. A
HD camera and a HD display device (LCD or
Plasma screen) were deployed at both locations
to capture and display video respectively.
The HD camera was used to capture the video at
each location and fed the capture device on the
sender workstation. Ultra-grid application then
sent the captured video over 10Gb network to
the recipient site. The receiver workstation at
each site received network stream and Ultra-
grid displayed the far site video on the HD
display at each site. A different application
called Robust Audio Tool (RAT) was used to
send audio over the same network.
Pros: The video quality was extremely sharp
with almost no latency. The output resolution
was 1080i. The interaction was very good in
this technique.
Cons: Deployment of this technique was very
expensive, bandwidth intensive and involves lot
of manual intervention throughout the session.
The equipment costs a lot and the charges for
the network bandwidth usage are extremely
high. Not all sites have 10Gb network access to
participate using such a technique and even then
the site has to purchase the workstations,
capture cards, etc, which are expensive. Even
after making such an investment, they can only
collaborate with another site that had all these
and a 10Gb network, which is very uncommon.
Figure 1: The different interconnection technologies used
in the distance learning experience.
Compressed HD streaming: This technique
was used to stream compressed HD audio/video
to University of Arkansas at Fayetteville. A
Polycom HDX 9000 unit was deployed at both
the locations. This is boxed product that works
well and the stream is sent on commercial
Internet. At LSU, the same video captured on
the HD camera (used for the uncompressed
technique) was fed in to the Polycom unit and
the audio was picked from the room mixer/echo
canceller. University of Arkansas has a HD
video camera that fed the video to the Polycom
unit and the audio from the room mixer. Both
sites had HD LCD screens to display the far
Pros: Video quality was good with a little
latency. The output resolution is 720p. Very low
bandwidth utilization of less than 1Mbps, that is
affordable at all sites. The device can talk to
non-HD Polycom or any H323 unit.
Cons: The unit is expensive as it is a
commercial solution.
Access Grid: Access Grid is an open source
application developed to enable collaboration of
research universities. This application is
available at many universities and simply uses
low cost capture cards, cameras and PCs to
capture send and receive video to/from far sites.
This technology was used to stream to
University of Arkansas at Little Rock, Arkansas
and Louisiana Tech University at Ruston,
Pros: Very easy to setup and the application is
free. Anyone can install the AG on a PC and
participate in the class
Cons: Video quality was not very good as the
resolution is very low. There is a little latency.
Students cannot feel the interaction as opposed
to other techniques above. Application is very
Webstreaming: A boxed product called NCast
was used to send the video stream live over the
internet. Users can access it via a browser or
quicktime video. Also the same video is
recorded to enable on demand access to the
lectures. This feature was very popular as
students accessed the videos numerous times.
This technique also served as backup plan if AG
does not work
Pros: On demand access and very basic
requirements to access the video
Cons: Huge latency. Not suitable for interactive
sessions mainly due to latency factors.
In its first offering, this course involved more
students than any other computer science graduate
course presented that semester. In its second
offering, it received the highest student evaluations
of any graduate course of the computer science
depart in its semester. For this accomplishment, the
professor was awarded the Graduate Teaching
Award for his College in 2007.
Student and Faculty Perceptions. Technical
application is only one facet explored in the
presentation of the course. A study was conducted
through the first iteration in order to ascertain
student response to the various technical aspects that
were used during the class. This study was designed
to demonstrate both the strengths of using multiple
technologies to foster interaction as well as the
weaknesses proctored by trying to combine these
technologies as well as human factors and have them
mesh together seamlessly.
The study itself consisted of a combination of
anonymous surveys, journal entries and discussions
with the instructors as well as proctors from the
various remote sites. In order to maintain continuity,
only sites that made use of all of the various
technical applications participated in the study. This
included two remote sites: the University of
Arkansas and a satellite location on the campus of
Louisiana State University, as well as the local site
at LSU. The overall population comprised 38
participants: 16 (42.10%) attending the class from a
remote site and 22 (57.89%) attending from the local
Students who participated in the anonymous
survey revealed correlations between three distinct
areas of interaction and the technological aspects of
the course. Dependent on the strength of the
technology at the various sites, students
demonstrated a greater sense of interaction between
themselves and the other sites as well as with the
host site. These correlations offer keys to the
technical areas that need the most work to best
engage every student.
Video Quality and Interaction.
The technology
that had the most positive impact on student
interaction was the uncompressed video. Every
student participating in the course either agreed or
strongly agreed that they were able to clearly see the
professor. Considering that current research suggests
that video clarity is paramount in maintaining
student attention as well as retention in a distance-
based course the use of the uncompressed high
definition video offered students the ability to
engage the professor as well as the other sites
visually (Coventry, 1998; Fillion, Limayem, &
Bouchard, 1999; Pitcher, Davidson, & Napier,
According to journals kept by the proctors at the
individual sites, while video quality offered a
numerous amount of tweaks and readjusts
throughout the course, it more often than not
remained running and smooth once initiated. As a
result, any problems that may have been
encountered in regards to networking, camera issues,
and the like were often not experienced by the
student considering the amount of setup time allotted
to the course before it was presented each class day.
CSEDU 2009 - International Conference on Computer Supported Education
Table 2: Remote Site Students Opinion of Interactivity
Key: 1 – Strongly Disagree | 2 – Somewhat Disagree | 3 –
Not sure | 4 – Somewhat Agree | 5- Strongly Agree.
N Min Max Mean Std.
I could hear the
professor clearly.
19 1 5 3.63 1.342
I could see the professor
19 4 5 4.63 .496
The audio/video quality
did not distract me from
the course material.
19 1 5 3.47 1.124
I made use of the video
archives outside of
19 1 5 3.26 1.661
I felt comfortable
asking questions during
19 1 5 2.47 1.389
I felt like I could
interact with people
from other institutions.
19 1 4 2.11 1.100
Audio Quality and Interaction. Where video
problems were not perceived within the course,
audio problems abounded and offered a significant
deterrent to student interaction. According to Frater,
Arnold, & Vahedian, (2001) as well as Tan & Tan
(2006), audio quality stands out as the most
important technology within a synchronous distance-
learning environment to maintaining the ability of
students to interact with the professor and remote
sites (Frater, Arnold, & Vahedian, 2001; Tan & Tan,
2006). According to follow up comments offered by
students the audio often distracted them from the
material and was the technology that needed to
improve the most. The survey also demonstrated a
positive correlation between audio quality and the
students’ perception of their ability to interact with
both the host site as well as the remote sites.
According to proctor journals, much of the audio
problems stemmed from the varying types of
equipment being used. While every site made use of
the Robust Audio Tool to send and receive audio,
each site used their own mixers, microphones,
speakers, echo-cancelling equipment, etc. to process
the audio being sent and received. In synchronous
distance-learning situations, if one site is sending
bad audio then it often propagates to all the sites. For
example, there was often a problem when one site’s
echo cancelling was faulty. All other sites would
then have to either mute the offending site or
continue while hearing their own voices sent back to
them often to the point of complete distraction.
Asyncronous Technologies. Asynchronous
technologies were often called upon to provide a
firm brace to class. Whenever there were issues in
the live class offerings, online videos were provided
to allow for the continuity of learning. One of the
more interesting developments demonstrated from
the student surveys was the fact that more local
students regularly viewed the archived videos than
the remote students. Those who did view the videos
both locally as well as remotely commented that
they used the videos for study guides and to assist in
the assimilation of the entirety of knowledge given
throughout class. More important to the idea of
interactivity, there was a positive correlation
demonstrated between students who viewed the
videos and the perceived ability to interact and ask
questions during class.
Table 3: Local Site Students Opinion of Interactivity Key:
1 – Strongly Disagree | 2 – Somewhat Disagree | 3 – Not
sure | 4 – Somewhat Agree | 5- Strongly Agree.
N Min Max Mean Std.
I could hear the professor
19 4 5 4.74 .452
I could see the professor
19 4 5 4.95 .229
The audio/video quality
did not distract me from
the course material.
19 2 5 3.89 1.05
I made use of the video
archives outside of class.
19 4 5 4.68 .478
I felt comfortable asking
questions during class.
19 1 5 3.16 1.39
I felt like I could interact
with people from other
19 1 5 2.79 1.13
Distance Learning is fundamentally important in
furthering knowledge across man-made limitations.
Based on our experience, using the latest
technologies it is possible to enable distance learning
across wide range of partners irrespective of the
local technological constraints. The tried and tested
methodologies developed by LSU and its partners,
verify the fact that using the technologies such as
optical networking, compressed HD via polycom,
live streaming via web, can be effectively utilized
(measured by student experience surveys) to deliver
highly involved – technical content across
geographical boundaries. Furthermore through
offline (non-live) mediums such as podcasts and on-
demand web streaming (of recorded material), the
educational content can be disseminated to a wider
audience; more importantly reaching the end users
who do not have access the state-of-the-art
technologies. While a lot more needs and can be
done, this work demonstrates the critical strides
taken by LSU and its partners in developing the
essential elements and methodologies to deliver
expert content to remote sites irrespective of
capability limitations.
The authors would like to thank Ed Seidel, Andrei
Hutanu, Petr Holub in pioneering the efforts in
developing technologies for dissemination of
uncompressed HD content. The authors would like
to thank Amy Apon, Petr Holub, Box Leangsuksun
and Gigi Karmous-Edwards for their continued
involvement in the multi-institutional High
Performance Computing Course. The authors would
also like to thank Steven Beck for their insights
regarding iPod dissemination of content and Stacey
Simmons for their organizational and leadership
support. The authors would also like to thank Jorge
Ucan for developing and implementing a HD
Audio/Video workflow that helps simplify the media
editing activities. This work has been developed
partly under the leadership and support from NSF
Award 0634046.
Coventry, L. (1998). Video conferencing in higher
education. Heriot Watt University, Edinburg.
Fillion, G., Limayem, M., & Bouchard, L. (1999).
Videoconferencing in distance education: A study of
student perceptions in the lecture context. Innovations
in Education & Training International, 36(4), 302.
Frater, M. R., Arnold, J. F., & Vahedian, A. (2001).
Impact of Audio on Subjective Assessment of Video
Quality in Videoconferencing Applications. IEEE
Transactions on Circuits & Systems for Video
Technology, 11(9), 1059.
Pitcher, N., Davidson, K., & Napier, J. G. (2000).
Videoconferencing in Higher Education. Innovations
in Education & Training International, 37(3), 199-
Tan, S.-C., & Tan, A.-L. (2006). Conversational Analysis
as an Analytical Tool for Face-to-Face and Online
Conversations. Educational Media International,
43(4), 347.
T. Sterling, E. Lusk, and W. Gropp, Eds. 2003 Beowulf
Cluster Computing with Linux. 2. MIT Press.
Gropp, W., Lusk, E., and Skjellum, A. 1994 Using MPI:
Portable Parallel Programming with the Message-
Passing Interface. MIT Press.
Chandra, R., Dagum, L., Kohr, D., Maydan, D.,
McDonald, J., and Menon, R. 2001 Parallel
Programming in OpenMP. Morgan Kaufmann
Publishers Inc.
Patterson, D. A. and Hennessy, J. L. 2003 Computer
Architecture: a Quantitative Approach. Morgan
Kaufmann Publishers Inc.
Silberschatz, A., Galvin, P., and Gagne, G. 2005
Operating System Concepts. 7th. Wiley Publishing.
CSEDU 2009 - International Conference on Computer Supported Education