A good existing model for a data acquistion solution is digital lecture capture:
- data (lectures) are recorded along with the powerpoints etc
- data is converted to produce a set of outputs, eg mp3 recordings
- recording are placed in an archive according to collections eg all stats 101 recordings grouped together
- metadata is published to learning management system to allow students to search and find recordings.
Description of such a solution
XYZU is currently running a large-scale implementation of Podcast Producer for our lecture recording service. It is deployed in over 70 venues currently and captures almost 300 lectures per week during semester.
Capture is initiated on demand using a Java recording tool installed on the PC or Mac used for teaching. The tool automatically populates the recording metadata from the central timetable. This data can also be entered by the lectures at the start of recording.
Recordings can be captured in three qualities:
• video – 640×480@15fps
• slides – 1024×768@3fps.
Currently only one video stream is captured at a time.
Using custom PcP workflows, content is automatically branded, watermarked and processed into 3 formats –video suitable for computer/fast network (~200MB per hour), video suitable for ipod (~70MB per hour) and one audio-only (10MB per hour) It is then automatically posted for download through either RSS feeds or through the course page on our Moodle LMS.
Venues contain a headless MacMini with either an Epiphan VGA2USB-LR or a DVI2USB frame grabber. The Epiphan captures the video source going to the projector, subject to some qualifiers based on the configuration of the room’s AV equipment.
At the back end, we have a Production, Test, and Development environment. Our Production environment is all XServe and Promise kit consisting of:
• 1 application server
• 1 Podcast Producer server, 17 grid nodes sharing a 4TB Xsan
• 2 Delivery servers each with 3x10TB RAID volumes
Test is an Xserver with a MacMini grid node. Development lives on VMs with a MacMini as the grid node.
After the usual teething problems for a service on this scale – mostly integration with the diverse range of configurations of AV kit and overall tuning, the service has proven very reliable. According to our lecturers, “It just works”.
Running this scale of service on anything other than enterprise quality server hardware is not an option. I agree that you shouldn’t be considering Podcast Producer given Apple’s limited server offerings. We’re currently looking at alternative systems for lecture recording too. Another system you should look at is Opencast. It quite new now but has a lot of potential.