The telecommunications and technology industries have reached a critical turning point. Because of significant advances in digital sensor technology, digital television, broadband deployment, and inexpensive data storage, the volume of online digital content is now estimated to be more than 280 exabytes (280 billion gigabytes).1

As unimaginable as that number is, increasing volumes of real-time, “life-oriented” content such as Internet Protocol (IP) and Internet-based television, live security feeds from the home and workplace, telepresence applications, and commercial video are expected to cause the volume of online digital content to grow by a factor of 10 in just the next 24 months!1

This amounts to no less than the democratization of the Internet on a global scale, which is a very good thing, but the volumes have begun to overrun existing technologies. Since 2007, the quantity of created data has exceeded available storage space. Unless we can bend the supply curve to more closely match the demand curve, half of all information created in 2011 will become digitally homeless.1

For network technology designers, the severity and complexity of the environment have the ironic effect of simplifying our challenge: to create an extraordinarily scalable but cost-effective infrastructure that supports the rapid delivery of any type of content from any network device to any network device at any time.

The Network Challenge

Today, many different device types create video in diverse formats for delivery to an equally diverse collection of end-user devices, each of which (potentially) requires differing encoded bit streams. For mobile devices, informed estimates point to as many as 60 different encoding formats, and that number is growing.2 We’ve all had the frustrating experience of unsuccessfully attempting to view a video on our computers and trying in vain to locate the correct codec online to make it work. Clearly, there are too many formats.

There are four key challenges in delivering the any-to-any mobile vision for video (see the figure). First is the non-stop increase in long-tail videos by global content providers. Second is the proliferation of digital standards and how they are increasing the complexity of data centers. Third is the conflict between “network access” capacity and the demand for high-definition content. Finally, there’s the emergence of real-time video content.

Voice over IP (VoIP) was designed to improve interoperability and drive down network costs. However, new standards have increased complexity with more than 20 voice codecs and massive gateways throughout the network.3 While some of the more common standards for VoIP deployment represent alternate or evolutionary approaches to VoIP as a service, they also represent complexity that must be managed in an increasingly complex operating environment (Table 1). The same can be said for the world of video. There are as many standards for video development as there are companies wanting to play in the space.

Transcoding is the process of converting digital content from one compression format to another so it can be displayed on devices with differing requirements. For example, an H.263-encoded file might be converted to H.264 to facilitate its use on different user devices. Transcoding is the key to the creation of universally consumable video for all types of mobile devices.

As demand for mobile television increases, so does the demand for mobile bandwidth. With high-definition (HD) content becoming more the norm rather than the exception, customers will increasingly turn to their mobile devices as the preferred method for viewing while on the go, especially as the Millennial Generation matures and becomes the major consumer of content on the planet.

We’ve already seen the impact of the migration to HD. Late in 2008, HDTVs outsold standard-definition devices for the first time.4 Online HD content is also rapidly becoming the norm. Cost-reduction efforts have increased the use of telepresence based on HD technology as an alternative to travel. HD has improved the quality of the experience, making telepresence more acceptable. The trend also has carried over to voice, where HD audio is becoming available. People now expect a higher quality of experience from their network services.

The emergence of the all-IP network has accelerated the “HD everywhere” phenomenon. For example, the processing of HD audio requires up to 10 times more hardware resources than a simple 4-kHz plain-old telephone service (POTS) call. The most advanced conferencing systems today require more than 30 times the power for image processing than standard-definition videoconferencing. Our industry needs to deliver orders of magnitude more bandwidth and processing with the same power and cost budget.

The strongest challenge that confronts the industry is the unprecedented growth of real-time video. Networks are required to transport and process large amounts of video. Any-to-any video requires demanding, real-time transcoding either in the endpoints or the network.

As growth continues, bandwidth demand increases put pressure on mobile service providers. This opens up many opportunities for silicon and system providers. Endpoint management is clearly an important element. But as devices become more capable, other demands emerge. Philosophies and strategies on how to create and store content vary.

One idea leverages the rapid advances in the end-user device domain, appliance-based transcoding, whether in a specialty device at the edge of the network or within the user’s device itself. Another idea requires content to be pre-transcoded into a variety of formats and stored for immediate delivery to the customer. Other alternatives include on-demand transcoding within the network and delivery to the customer on a real-time or nearly real-time basis, as well as approaches that leverage scalable video coding. All of these methods will impact the network.

Mobile Content Management Strategies

Currently, providers must create and make available distinctly different versions of each content stream. The result is massive complexity on the encoding side, massive storage within the network, and a complex tracking mechanism to ensure that the correct version is transported when the consumer demands the download. Four strategies are possible for managing this challenge, and a combination of them can be used to deliver the any-to-any video network:

• Endpoint management

• Pre-transcode and store

• On-demand transcoding

• Next-generation video codecs

Endpoint management is the default mode of operation, but scalability and usability limitations have emerged. In a PC-to-PC environment, this strategy works well. However, the mobile environment requires a different approach. Older devices are still in widespread use, meaning they will only support a minimal amount of distributed content.

The industry’s solution has been to accelerate the deployment of higher-resolution devices that support varying bit rates, formats, and high-resolution streams. In addition, pushing video management complexity to the mobile devices has impacted cost and, more importantly, battery life. YouTube, for instance, has adopted this approach, so only endpoints that support YouTube’s format can view its content.

Pre-transcode and store requires content to be pre-transcoded and stored in every format that users will require. This approach is widely used today by video-on-demand service providers. Under the tenets of a pre-transcode and store strategy, infrastructure costs scale according to the volume of available content and the number of available formats, not on a per-usage basis.

Consider a global movie-on-demand provider. Every time that a company decides to offer its content in a new country or region of the world, it must first establish a local data center presence with copies of all content in all formats. Even with storage costs as low as $0.10 per gigabyte per month, the cost rapidly becomes overwhelming (Table 2).

On-demand transcoding requires an infrastructure that scales to usage or revenue and not content. Content is transcoded in real time. Movies on demand would be stored in a few formats and transcoded in real time for the various mobile device formats. The strategy works for non-real-time and real-time content. Silicon solutions must deliver the quality and performance required to support a large volume of transactions. Advances in multicore media processors now make this possible.

Ideally, content would be encoded once, and when decoded it would yield full-resolution video. If lower-resolution versions are required, a version of the encoded stream would be transported with limited additional processing. The encoded stream would be able to adapt to both available channel bandwidth and the capabilities of the requesting device.

Scalable video coding (SVC), an extension of the H.264/MPEG-4 AVC video compression standard, intends to do this. The Moving Picture Experts Group (MPEG) issued a call for proposals on SVC technology in 2003 and approved the standard in 2007. SVC standardization enables the encoding of high-quality video bitstreams that contain one or more subset bitstreams.

The subset bitstreams can be decoded similar to that achieved using the existing H.264/MPEG-4 AVC designs. Reduced processing can be achieved as the subset bitstream is derived by dropping packets from the larger bitstream. The dropping of packets can be used to dynamically reduce the resolution of a video frame to adapt it to network constraints or the supported format of the end device.

The SVC standard includes three additional scalable profiles: Scalable Baseline, Scalable High, and Scalable High Intra. These profiles are defined as a combination of the H.264/MPEG-4 AVC profile for the base layer. SVC is a new, network-friendly video codec that may well transform the way we architect the next-generation, video-aware infrastructure (Table 3). SVC-based products started to appear in 2008 and 2009.

Final Thoughts

Clearly, getting all content creators to agree on a common format for mobile video distribution is not realistic. Therefore, on-demand transcoding of content across various formats, bit rates, and screen resolutions is important if widespread delivery of content on a non-discriminatory, device-agnostic basis is to be realized. Ultimately, the transcoding process will be performed on the device, the data center server infrastructure, and the networks.

Key to the success of this evolving process are the semiconductors that will drive a transcoding success story. Powerful accelerators are required to handle the intensive processing that transcoding requires. Evolution in video standards continues, and hardware solutions must be ready and flexible for these new developments. Networking silicon providers targeting these concerns will be well positioned as on-demand transcoding evolves to satisfy the increasing demands of a media-hungry marketplace.

References

1. IDC White Paper, “The Diverse and Exploding Digital Universe,” March 2008

2. Akamai’s 2008 State of the Internet Report

3. http://en.wikipedia.org/wiki/List_of_codecs

4. www.informationweek.com/news/personal_tech/TV_theater/showArticle.jhtml?articleID=211600041

Jas Tremblay is the marketing director and segment lead for the SMB and enterprise markets for the Networking and Storage Products Division of LSI Corp. Previously, he was a senior project and development manager for CGI Consultant Firm in Toronto, Ontario, and a product and business development manager for the EMEA region for Nortel Networks. He received a bachelor’s degree in electrical engineering from the Université de Sherbrooke and a master’s degree in business administration from HEC Université de Montréal.