I Want My VOD: Internet Video-on-demand

Summary: A single Internet Video-on-demand system that delivers a wide range of distribution options (from on-campus users to alumni overseas) and access restrictions (from authorized groups to unlimited viewing) does not exist. Until is does, multiple products and services will have to be integrated, behind the scenes, to provide a seamless experience for users (producers and consumers) throughout the lifecycle of each digital media asset. Evaluations of seven options and recommendations for further initiatives are included.

Note: Please read “Lost and Found: A Primer to Digital Media Asset Management” first; much of the information in this paper assumes prior knowledge.

1. Context

Video-on-demand is one of the greater challenges for Web delivery. Comprising a large amount of data in its raw form, broadcast-quality video needs to be compressed and handled very efficiently in order to be viewed as a dependable service by Internet users.

In its first generation, Web video was compressed with a variety of proprietary compression methods (Cinepak, Sorenson) and contained within proprietary wrappers (QuickTime, Real, Windows Media). The compressed derivatives were low resolution, filled with artifacts, and couldn’t be played without interruption unless supported by a expansive system of servers located around the Internet, close to users.

Ultimately, future Internet Video on Delivery (iVoD) delivery systems will support standard formats and be integrated within the complete media asset lifecycle, linking with external metadata from creation through to storage, discovery, and delivery. Unfortunately, that level of integration is not yet ubiquitious.

In this paper, we will evaluate tools and techniques for distributing iVoD: broadcast-quality video assets created by faculty, staff, and students at Dartmouth. We will not focus on the production of those assets — other than the acquisition of metadata — but instead on their access and distribution, which is the traditional focus of video-on-demand via CATV systems.

For an overview of packaging video assets into a single program stream on the Web, see the Public Webcast Discovery Project report.

2. The Framework

A. Lifecycle support

For this paper, we will define the media asset lifecycle as having five phases: create, annotate, store, search, and distribute.

Figure 1. The Asset Lifecycle, Simplified

Some video-on-demand systems excel in multiple parts of the asset lifecycle while others provide support for some areas. Few, however, cover the entire cycle.

Again, see “Lost and Found: A Primer to Digital Media Asset Management at Dartmouth” for a deeper discussion about the asset lifecycle.

B. Distribution range and audience restrictions

Current restraints in Internet technology affect two aspects of iVoD video: distribution range and audience restrictions. The result is that different options have evolved for certain needs: whether the audience is on campus or worldwide, and whether or not the asset needs to restricted to a selected group of viewers.

 Figure 2.  Distribution Range

Distribution range includes local (on-campus course sessions) to distant (video news releases for worldwide viewing). Fortunately, the on-campus network environment provides a high level of control, so that a predetermined number of on-demand video streams can be dependably delivered to concurrent users.

For robust off campus distribution, however, either a network of edge servers or a peer-to-peer network is required to assure that sufficient storage and bandwidth is available for delivery of the broadcast-quality video asset without interruption.

 Figure 3. Audience Restrictions

Audience restrictions can vary from closed (students in a single course) to open (a worldwide alumni audience). Open, unrestricted usage is simple; it is the default for most media players.

Restricted playback for pre-defined audiences, however, is a challenge, especially since there are no widely-accepted open standards for doing so. Some architectures have implemented robust digital restrictions management (DRM) systems -such as Apple Fairplay and Microsoft Windows Media – but they currently require accessing the asset with a specific player.

So, by integrating these three factors, we now have a model to help us make a preliminary evaluation of various iVoD systems.

 Figure 4. iVoD Evaluation Model

3. Context

At this time, there is no single, robust iVoD system that provides a full distribution range with effective audience restrictions. Various solutions, however, can be combined to cover primary needs until single open-standard solutions becomes available.

For that reason, it makes sense to create a delivery master (a primary “derivative”) in a format with the greatest potential for preservation and transcoding (or repurposing) and also to record external information about the asset (“metadata”) so that it can be repurposed in the future.

Then, when an Internet VoD system is selected for a particular function, two processes need to be developed: an on-the-fly automated process for transforming essence to an appropriate new version (“derivative”) when requested, and a system for the exchange of metadata between an asset management system and the VoD delivery system.

Metadata is the most challenging piece of this puzzle. None of the iVoD systems we’ve investigated have robust support for ingest of pre-defined XML schemas, let alone exchange of that data via XML Web Services or defined API protocols. This may change in the near future, but a set of schemas that serves Dartmouth’s general media asset management needs will go a long way toward being ready when an optimal iVoD system becomes available.

Note that the Library of Congress, with Google and other partners, has launched a World Digital Library initiative. As its core, “the Library will develop a plan for identifying technology issues related to digitization and organization of WDL collections. These might include presentation, maintenance, standards and metadata schemas that support both access and preservation.”

4. Options

We will now evaluate seven Internet VoD solutions against these three requirements: support for the full asset lifecycle, robust distribution support for local and worldwide users, and flexible audience restrictions from small groups to unlimited access. Each factor will be ranked from 1 (low) to 3 (high) with preliminary rating based on the sum of those totals.

A. Akamai Edge Platform

Akamai was one of the first Internet video distribution networks and, as such, is used by the major programming services that moved early into Internet VoD, such as CNN, WGBH, etc.

A network of hundreds of “edge servers” — mirrored servers located at physical locations around the internet — cache and deliver content when requested by nearby users.

Akamai’s Edge Platform does not provide an end user interface, only distribution of the assets — streams and downloads — that can be integrated into locally-hosted Web presentation applications. Interface and management of user restrictions needs to be developed locally. Other Akamai services provide wider support of the lifecycle.

Figure 5.

Pros

• Tested thoroughly through years of service.
• Powerful customization possible with local Web applications.
• Supports a wide variety of binary and text formats, including mp3; architecture agnostic.

Cons

• No integrated digital rights management; external players and development required.
• No integrated metadata search; external development required,

Preliminary rating

1 (lifecycle) + 1 (restrictions) + 3 (range) = 5 (total)

B. Brightcove

Brightcove is a new entry into the Internet VoD space and is currently in prototype development. With funding from major media corporations, Brightcove may find a combination of cross-platform support and strong digital restriction management acceptable to the television and film industry.

The two layers of service provided are distribution (publisher features, affiliate aggregation, consumer search and access) and monetization (advertising, pay-per-view, syndication, licensing)

Pricing is primarily oriented toward paid media: based on commission. For free media: pricing is based a combination of bandwidth plus stream launches.

The Brightcove player supports Flash video codecs, but support for MPEG-4-AVC playback, for example, an external player such as WindowsMedia or QuickTime is required.

Figure 6.

Pros

• All interface elements are integrated in a proprietary Flash interface for cross-platform support.
• Auto ingest is provided for essence, derivatives, metadata via batch APIs.
• All metadata is accessible to external systems via APIs.
• HTML documents for each channel integrate text transcripts.

Cons

• All derivatives must be pre-compressed; no live transcoding available (although expected in Q2/Q3 of next year).
• Support for limited compression formats.
• Full transcript search not available at launch.
• Lower accessibility of Flash interface (as compared to HTML) for screen readers and other alternative interfaces.

Preliminary rating

3 (lifecycle) + 3 (restrictions) + 3 (range) = 9 (total, with an important caveat: the service is not yet available)

C. DTV/Broadcast Machine

“The Participatory Culture Foundation, is a nonprofit organization that is developing a TiVo-like combination RSS feed reader, Bittorrent client, and video player to go with their Broadcast Machine video publishing package. The goal is to put together a complete and OSS platform that lets nontechnical producers broadcast their videos to nontechnical users without using lots of bandwidth.”

DTV is the player — currently for Mac with Windows and Linux in final development — based on the VLC player, which supports a variety of streaming and downloading formats. DTV presents channels published on individual Web sites and aggregated, via RSS, in its channel guide as well as through other videologging aggregators. DTV supports MPEG-4, MPEG-2, and QuickTime streaming (RTSP).

Broadcast Machine is the set of scripts that can be posted in any PHP-enabled Web account. It allows for the aggregation and syndication of live or recorded videos (via URL or Torrent) via categories and tags, including playlists. Channels can be restricted to pre-authorized users. (There is currently no batch ingest of access information, but the system is open PHP, so custom development would be minimal.

Broadcast Machine can syndicate public assets hosted in the OurMedia section of the Internet Archive at no cost.

This project has received funding from Mitch Kapor. Partners include Creative Commons and Current TV. Press coverage has included CNET, Wired, Business Week, 

The system is in a similar market space as Brightcove (corporate) and the Open Media Network (public broadcasting).

Figure 7.

Pros

• All interface elements are integrated in a single application window
• DTV plays all compression formats supported by VLC
• Assets are identified by URL and so can live on any host (including QTSS, Internet Archive, Video Furnace)
• All metadata is accessible to external systems via APIs.

Cons

• All derivatives must be pre-compressed; no live transcoding available 
• Use of uploaded transcript for search isn’t clear

Preliminary rating

2 (lifecycle) + 2 (restrictions) + 3 (range) = 7

D. Flash Video Streaming Service

Flash Video wraps video encoded in the On2 format into a Flash .flv file. That asset can then be distributed in several ways: integrated into a Flash .swf skin, or streamed or downloaded as a .flv file from a Web server, a Flash Server, or the Flash Video Streaming Service. A Creative-Commons licensed player is available for download delivery.

In order to play Flash video consistently at a distance, delivery via an edge server network is required. Three companies offer the service now. For asset lifecycle support, however, an interface and management system needs to be developed external to the player, which is exactly what Google Video and Brightcove do.

Figure 8.

Pros

• Nearly ubiquitous player installation.
• Authentication available through VitalStream (tokens, IP Addresses, referrers).

Cons

• Requires proprietary compression and wrapper formats.
• Complex user interface for producers only partially simplified in Flash 8.
• Lower accessibility of Flash interface.
• Management and interface must be delivered externally.

Preliminary rating

1 (lifecycle) + 1 (restrictions) + 3 (range) = 5 (total)

E. Google Video (Beta)

Google launched the beta version of this service last spring; after its first past six months, thousands of independent videos are being hosted and delivered for free.

Google uses Flash Video to deliver assets to the browser window.

Figure 9.

Pros

• Free.
• Support for wide portion of life cycle, including metadata (transcript) search.
• Cross platform playback (via Flash).
• RSS feeds are in development for integration in external interfaces.

Cons

• Advertisements may be presented next to videos.
• Does not currently support .mp3 file playback.

Preliminary rating

2 (lifecycle) + 1 (restrictions) + 3 (range) = 6 (total)

F. Ourmedia.org/Internet Archive

Ourmedia.org is a free, “do-it-yourself mechanism that lets users anywhere in the world upload material, view or share media (depending on the license), rank their favorite works, and offer commentary and tutorials.”

Ourmedia provide an interface to many digital file formats —  video, audio, photos, text and software — and acts as a front end for the Internet Archive (IA). Ourmedia provides Mac and Windows applications that streamline file upload; files are then posted on the free IA servers, and distributed via the IA network.

Ourmedia is being co-led by Marc Canter, a recognized leader in social networks and blogging. The online community service is powered by the open-source Drupal community system.  The Internet Archive “was founded to build an Internet library, with the purpose of offering permanent access for researchers, historians, and scholars to historical collections that exist in digital format.”

Of course, media can be posted directly to the Internet Archive, as a number of schools have done, and accessed directly from there.

Figure 10.

Pros

• Highly-established archive storage.
• Simple, cross-platform upload tool.
• Supports a wide variety of binary and text formats, including mp3; architecture agnostic.

Cons

• Desktop player required.
• No support for legal file sharing, such as BitTorrent, to reduce distribution latency.
• No support for integrated digital rights management; external players and development required.

Preliminary rating

    3 (lifecycle) + 1 (restrictions) + 3 (range) = 7 (total)

G. QuickTime Streaming Server

QuickTime is an early media file architecture that has grown to integrate many different types of media assets and compression formats; the QuickTime wrapper architecture was adopted as the basis for the MPEG-4 standard. The Streaming Server system delivers pre-compressed videos and audio, but no search or metadata management.

Figure 11.

Pros

• Integrated deeply with Macintosh platform.
• Flexible architecture with support for large number of formats, including mp3.

Cons

• Periodic problems that break playback on Windows platform.
• Needs edge server system for robust off-campus distribution.
• No rights management available, other than manual password “keys”.

Preliminary rating

2 (lifecycle) + 2 (restrictions) + 1 (range) = 5 (total)

H. Video Furnace

Video Furnace was acquired by Dartmouth as an IP Video replacement for Cable TV service; Dartmouth has implemented that multicast feature for campus network users. Video Furnace also provides unicast on-demand streaming.

Figure 12.

Pros

• Locally-controlled within Dartmouth network.
• Uses session-based Java player with no caching or installed player to improve security.

Cons

• Compensation model based on bandwidth used for Internet VoD delivery.
• On-campus only.
• Authorization system needs to be developed in-house through available hooks.
• Does not currently support mp3.

Preliminary rating

2 (lifecycle) + 2 (restrictions) + 1 (range) = 5 (total)

I. WGBH Forum Network/Open Media Network (Kontiki)

Service that distributes videos from Boston-area non-profit partners. Now expanding to a national scope in partnership with the Open Media Network (based on the Kontiki Grid Platform). Evaluation of future service to be considered.

Figure 13.

Pros

• Utilizes network of client machines; no dedicated server network required.
• Access control, network security, desktop security, and usage control tightly integrated.
• Supports mp3.

Cons

• Kontiki currently supports only Windows (active-X); Mac expected in next quarter.
• Distribution only; on-the-fly transcoding not integrated.

Preliminary rating

2 (lifecycle) + 1 (restrictions) + 3 (range) = 6 (total)

Recommendations

Given the variety of overlapping feature sets provided by each of the above services and products, two factors need to be considered. First, what are the prioritized use scenarios that need immediate attention, and what resource is available? Second, which services and products that meet those needs most effectively integrate open-standards and interoperability. With that information, deeper evaluation of top solutions can be undertaken.