The implementation of the Opencast Matterhorn lecture capture framework at the University of Cape Town.
Presented at the Opencast Community Unconference, Harvard, 6 June 2012.
2. Our Matterhorn journey
• Identified an emerging need on campus: demand for solutions and
proliferation of DIY solutions.
• UCT staff survey: 78% of lecturers would make some recorded
lectures available to their students, 60% would make some lectures
public (Dec 2010, n=176), 9% doing some DIY recording
• $50,000 allocated for equipment in 2011 (approx $60K in 2012), as
a side-effect of including lecture recording in a larger project
• Looked at proprietary solutions but either too costly or not aligned
with enterprise architecture
• Early proof of concept of Matterhorn from Nov 2010
• Week-long on-site training and consultancy with Entwine, Oct 2011
• Launched production pilot in Feb 2012 on Matterhorn 1.3
• Expecting demand to scale up (possibly to 50 venues)
3. Why consider lecture recording?
• Challenges addressed • Better student support within
courses
1) Students whose first
language is not English often • Value to students and staff not
struggle to understand the in the course
content of face-to-face lectures
– Across courses
2) Lectures are content-packed (same year)
with limited time for discussion – Same course,
3) Students are sometimes previous years
absent for reasons beyond
• Potential for wider
their control
dissemination of UCT’s
4) Student attention and expertise (Afropolitan focus,
engagement within a lecture OpenContent, OpenUCT)
can vary a lot
4. UCT survey
CET 2010 Educational
What proportion of your lectures Technology Staff Survey
do you record yourself?
1 to 15 Dec 2010,
176 respondents
If your lectures could be
automatically recorded, what your students the public
proportion of your lectures would
you like to be available to …
5. What it means for lecturers
• If you teach in an equipped venue, you may
choose to have your lecture(s) recorded
• You may need to clip on a lapel mic, but do not
need to do anything technical before, during
or after.
• You may choose how to share the recording:
a. Course cohort only (your students)
b. University-wide (any UCT students or staff)
c. Public (anyone)
6. lecture recording in action
Three channels:
•Audio
•Camera
•Screen
(VGA output)
Allows navigation
by slides or
timeline
Shows viewing
statistics
(hotspots)
Online playback,
or download
7. Matterhorn in Sakai CLE with LTI
A Matterhorn
tool is placed
in the Sakai
site using the
Learning Tools
Interoperability
(LTI) standard.
The LTI tool is
preconfigured
with the Series
ID.
8. Self-service scheduling in the LTI tool
http://opencast.jira.com/browse/MH-8315
Allow instructors
to schedule their
own recordings
and update
recording
metadata
(title, description)
9. Matterhorn Dashboard
We graph a set of
metrics (activity,
throughput, response
time) with mrtg using
REST queries and a
few direct database
queries.
View it live at
http://media.uct.ac.za/
mrtg/dashboard/
We also have Sintrex
monitoring of servers
and agents.
10. JIRA integration
We manage
recording
requests and
related tasks in
JIRA.
A custom script
will place the
LTI tool in the
right Sakai site
and also create
JIRA issues for
any workflow
failures in the
series.
11. Matterhorn Deployment
• 4 SLES VM instances (1 x admin node, 1 x
Red5 streaming server, 2 x worker nodes)
• mysql 5.1 database
• Storage on SAN, shared by nfs
• 15 Ubuntu capture agents in venues (small
Dells)
• Workflow for handling recording requests uses
JIRA integrated with Matterhorn (though still too
manual)
12. Venue equipment
• Custom-built Capture Agents (Dell
Optiplex 780 USFF)
• Epiphan USB2VGA
• Logitech C910 webcam
• MXL AC404 USB Conference Mics
• Integration with venue audio systems
• IP Cameras
13. Scale
• 15 equipped venues
• 8 active venues Feb-May 2012
• 1605 distinct users to date
• Peak activity:
– 28 recordings / week
– 25 concurrent streaming users
– 148 distinct users in 24 hrs
– 75 distinct recordings viewed in 24 hrs
14. Audio
• High reliability in simple venues (USB conference mic)
• 50% reliability in some venues with lapel mics:
– Flat batteries
– User behaviour
– Missing microphones
– Inadequate in-venue support
• New audio strategy: combination of mics for amplification
and recording-only mics (e.g. boundary mics), auto-
selected / mixed with Digital Signal Processor
• Involved in a project to redesign the A/V support services
on campus (design, installation, support)