Category Archives: Uncategorized

Institutional Collaboration

Our SharePoint space is now working. A number of documents and a discussion area is there and the code will be added very soon (just removing usernames/passwords now).
If you’d like access to this site please e-mail peterg@soton.ac.uk with your details and I’ll send you an invite to register/access the content.

Caring and Sharing

Southampton is just starting its second academic year using the JISC funded e-Assignment system. This will be with the majority of assignments from 3 Faculties along with pilot assignments from all others during the year. We’ll also be linking e-Assignment to a new feature of the University student administration system so that final marks can [...]

AV workflow in the University of Reading

by Dr David Wong
The workflow provides for an end-to-end solution where users submit their audio / video file on a university web page, and after a few minutes receive an email containining instruction on how to link the transcoded media file on their web page.
The systems components consist of:

Workflow server (Ubuntu), provdes a platform for NSSDropbox 2 (from University of Delaware).
A modified version of NSSDropbox 2 using LDAP security, provides upload facility (2.0GB limit per file). Each submission causes an email sent to a designated email account
Script (Perl) in workflow server scans email account for new message,  extracts uploaded file out of NSSDropbox’s internal storage, writes it to a designated directory, emails administrator a notification of the upload, and emails the user instruction on how to link the transcoded files from their web page
Mac OS XServe, running Episode Engine, scans the designated directory regularly for new files, does transcoding, and writes transcoded files to an output directory
The output directory is mounted on a web server, making the files available on the web
Templates in University CMS, and VLE (Blackboard), enable end users to link the transcoded files from their web page
JW Player

File formats:

Audio: accept mp3, wav and wma, and output to mp3
Video: accept mpg, mp4, avi, and output to mp4 h264 AVC (Levels 1.2 and 3), and AAC audio encoding. Some success with other input formats wmv, tod, mod etc.

Advantages

Episode Engine’s preset formats and workflows transcode files that conform to h264 standards.
JW Player provides for a large number of platforms.
Very low load on servers. 
Apart from staff costs, main expenditures are XServe and Episode Engine.
Multiple workflows based on media type or requirements (e.g. teaching and learning, audio only, video).

Issues

Episode Engine is not easy to use, nor flexible; its warning and error messages can be confusing and far from instructional. In cases where it is stuck with a job that’s difficult to transcode, it could hang and requires intervention by aborting the job. It’s possible to modify / create workflows which requires root access, and I didn’t pursue that.
Episode Engine strips out metadata contained in file.
Workflow server (running NSSDropbox 2) is inside the firewall to minimise security issues. Off-campus users can upload if they log in using VPN.
Not really an issue, but for people new to Perl and scripting (or those thinking about implementing NSSDropbox 2), we did spend a lot of time (weeks) modifying NSSDropbox 2 code, adding extra bits (and removing/disabling others), and hours of testing. And time spent learning Perl.

Migration test case
Two years ago, most of the univeristy video files were hosted externally. With Episode Engine and Helix, we began hosting the assets in-house. By mid September 2011, we have about 700 AV files stored on a disk attached to the Helix streaming server (a majority of these originally migrated from another project dedicated to teaching and learning). Most of these files are in mp4 (video) or mp3 (audio) formats as for about 1.5 year we stop transcoding file in Real format and also ceased streaming, giving preference to wider standards. This paves the way for a relatively easy introduction of the “AV Dropbox” scheme (easy in the sense that we didn’t have to re-transcode the files)
Migration consists of requirements analysis with stakeholders, and commissioning of servers (workflow & upload using NSSDropbox, and web) and storage areas (Netapp, one for upload and one for transcoded files). Other issues came into play including the use of a “friendly URL” (we decided on avcontent.reading.ac.uk), and proxying the workflow (or not, being the motion at the moment), and re-transcoding a small number of old files to mp4 / mp3.
Going live was an event that demanded some added attention as updating the CMS and Blackboard templates (to point to new server URLs) meant currently available AV files needed to continue to work with the new URL. A great deal of checking went in to ensure going forward does not result in going back swiftly! Apart from some templates that were non-standard, the go-live went unnoticed and was a quiet success.
Next step
JW Player remains an attractive asset. With HTML5 (and mobile platforms) becoming mainstream, we are looking at what’s possible. There is significant requirements from the corporate office to provide for these outlets. A bought-in solution is unlikely, nor something that potentially gives security concerns. Cloud solution maybe possible, so long uploaded files are not stored on remote servers.
Zend.to appears to be attractive with regard to workflow.

Dr David Wong Multimedia: Corporate Information Systems Support IT Services, University of Reading

Permalink

| Leave a comment  »

APEL Paper Accepted

The pineapple team have had a paper on the tensions of APEL accepted for publication by the journal Research in Post-Compulsory Education. The lead author is Harriet Dismore and the paper will be published later this year.

Your technical help is requested…

Hello fellow Steeplers.
You may remember that I’m working on a project to improve reporting on the impact of podcasting (Listening for Impact @ JISC). I am in the process of writing an application to do a lot of the heavy data analysis from various sources (Apple spreadsheets, Apache Log files, Google Analytics, etc), but I’ve got a little bit of a challenge that I expect some of you can help with.
I am trying to identify a range of User Agents that appear in our hosting logs, but that aren’t identifiable using the tools I presently have. I have documented the latest on this on the LfI website and would ask you take a look at User Agent Analysis – Part 2: Name those agents to see if you can help us (and ideally save me a lot of time on Google and experimentation!).
Thanks!Carl

Permalink

| Leave a comment  »

What is Google playing at?

I said earlier it hasn’t been quiet lately. Another hot topic received a shove recently when Google announced they were removing support for the h264 video codec from their Chrome webrowser in favour of their own WebM format (oh, and Theora). This simply means that HTML 5 tag content will not playback without needing some third party software/plugin (where said content is in the globally dominant and arguably superior h264 format).
I can only conclude this is not a good thing.
Why? Well here’s a few observations resulting from a community discussion of this topic…

Desktop browsers that are not Internet Explorer or Safari still account for

Thoughts on Apple in Podcasting

Whilst it’s been a while since I last wrote, things haven’t been quiet. When Apple announced in Nov 2010 that Xserve was to be discontinued three months later, shockwaves went out through Apple’s enterprise community. Having gone through the tech equivalent of the 5 stages of grief and spent some time discussing this with colleagues near and far, I thought I’d jot down some notes on current thinking amongst us.

Apple have killed off the Xserve, it’s not coming back however we wail.
They have no current product that will suffice in a datacentre for any serious OS X based computing power/capacity (for various issues of density, resilience, connectivity, politics, etc).
There are only three services that exist only on OS X Server: Software Update, Netboot and Podcast Producer. Every other service has an equivalent or alternative on either Linux or Windows, and they have many acceptable options for datacentre deployment.
From a podcasting perspective, of those three, we’re only interested in Podcast Producer, so I can ignore the other two and rely on our talented SysAdmins to provide equivalent services for other functions (Directory, Apache, etc).
Podcast Producer has a range of faults/issues, including:

Capture is too unreliable and expensive to be deployed further than a pilot setup
The Library requires too much development work to be done on it to allow for edits to be made to metadata, etc
The workflow is QT dependant, and has a major bottleneck on the render task as that is reliant on GPU capacity (something Xserves are weak on)
To get a usable range of output codecs, you have to install alternative compression software/applications, several of which have workflow-process like capabilities duplicating functionality in PcP
There’s no indication these are going to be addressed/improved (see below on Apple and roadmaps)

Because of the above, Oxford has failed to seriously use PcP as intended from the outset, and we know we’re not alone.
To support a transcoding service scaled for university wide use, we created a cluster of Xserves backed by Xsan.
Xsan has been ok (1 fault in 2 years of operation – almost zero dataloss), but it requires qualified SysAdmin maintenance (not cheap or easy) and two Xserves (or at least, two machines with FibreChannel cards in, thus ruling out MacMinis in the datacentre) to act as Metadata Controllers. Whilst these machines can perform other tasks, it’s not recommended to put anything too intensive on them lest their response times decline.
Xsan has some weaknesses:

Only 85% of a volume capacity can be used before serious performance degradation kicks in.
This means that for a 16 x 1Tb drive chassis, two drives are given to Metadata, one is left as a common hot spare, and the remaining thirteen are in a RAID 6 configuration for speed and resilience. That gives us a Volume of 11Tb theoretical space, which after formatting is 10Tb. 85% of that is 8.5Tb of usable diskspace out of 16Tb of theoretical capacity, or if you prefer, little over half of the total storage purchased is usable!
Licences, whilst far cheaper than the Linux or Windows equivalent (Xsan is just Quantum’s StorNext FX SAN product rebadged), are not cheap, even on educational pricing, and you need one per machine attached.
FibreChannel is not a cheap technology to introduce (minimum £3000 investment), even more so if you want to attach video editing workstations to a central processing cluster.
All told, this make Xsan expensive to buy and expensive to maintain.

With Xserves now on a dying path, that gives Xsan about 2-3 years to live.
OS X Server is a separate issue from Xserve, but not entirely if your policy is that servers should operate in a datacentre.
There is no roadmap for OS X Server:

This isn’t news, Apple don’t do roadmaps in general, you take that as given whenever you buy from them, but this lack of future planning upsets enterprises whose strategies are often like supertankers and need long periods to turn around in.
There has been no announcement or even hint of OS X server existing beyond 10.6.n; again not surprising, but when you pull out the rug from customers with the end of an entire product line, it would help to reassure them their services might continue in another form. This is unusually bad PR for Apple, and lends credence to the idea Apple is withdrawing from the enterprise and/or education market to a degree.
If OS X Serve does continue, the only solution we can imagine that would make that make it viable in a datacentre is to virtualise it on non-Apple platforms. Indeed, this has been nicely campaigned for by Dave Schroeder (http://appleopenletter.org/). However, as Universities have been investing in virtualised infrastructure already, existing Mac hardware isn’t going to fit into that and thus becomes redundant anyhow, and Xsan becomes surplus to requirements also.

The above issues aside, the biggest problem from this announcement and subsequent (near) silence, is simply the trust that has been broken between Apple and clients that have bought into Apple’s enterprise products. This has been a vivid reminder of common Apple behaviour (consumers have known this for years) and undoes nearly all the work Apple have done in the 9 years of developing a server product/solution.
So, we conclude that:

Our Xserve investment is now EOL’d and we have to switch platforms. This is exceptionally unfortunate in current economic times where higher education is especially hurting.
Podcast Producer is no longer an option for centralised podcasting services and needs to be discounted from future thinking.
Xsan is redundant along with the Xserves (even though the hardware and tech will likely continue to run happily for the next 2-4 years – ignoring the security concerns of unpatched software).

As we’re now onto Acceptance, we look to the future… a future that may contain Matterhorn, Mediamosa, Cisco, Linux and more, but not Apple (apart from on a desktop or two ;-)
More on our future planning when we’ve finished evaluating our practical options.
Carl

Permalink

| Leave a comment  »

e-Assignment where are we so far?

So another big gap since our last update. We performed several more tests of submission and marking over the summer and created our pre-production and production environments. Unfortunately we were still lagging behind in functionality for the administration of assignments, as such we decided not to ‘go live’ with the system in semester 1.
The School’s [...]

Listening for Impact

Little bit of cross publicity here…
JISC are currently funding the Learning Technologies Group at Oxford to do a quick investigation on the impact of podcasting. This is where my day to day efforts are currently focussed, and since I’m dealing with the technical approaches side to this analysis, some of you may be interested to learn what we’re doing. 
If Log Analysis, Analytics, SEO and Stats are your bag, pop across to the project website (Listening for Impact) and keep an eye on the postings over the next 5 months. We would appreciate your feedback.
Carl

Permalink

| Leave a comment  »

Workshop at UCL – 22-11-2010

For those tracking what’s going on, but not noticing the wiki or the mailing list, here’s a brief reminder of our (last?) Benefit’s Realisation Workshop for Steeple. We’re gathering at UCL on Monday 22nd November for a 1 day workshop to cover a range of topical issues. More details can be found at http://www.steeple.org.uk/wiki/BR/Workshop_Nov_2010_UCL.
See some of you in London :-)
Carl

Permalink

| Leave a comment  »

Copyright © 2007 Institutional Innovation (p). All rights reserved.
Institutional Innovation