The State of Web Video

Web video has grown up fast. Ten years ago, quality and capacity problems plagued attempts at putting video online. Today, Internet users can experience watching video online at up to 720p with adaptive streaming so they never suffer from buffering or stuttering, and they can use advanced interactive features that traditional television can only dream about.

According to eMarketer, more people are watching online than ever before: More than 60 percent of the U.S. Internet population watches video content online at least once per month. Major events boost online viewership and keep users engaged with the content longer.

Impact on Broadcasters
For broadcasters, the Web is both a challenge and an opportunity. A recent Nielsen study showed that first, television viewers who also use the Internet are spending more time online, which means less time watching TV, decreasing the value of the broadcast channel. Secondly, 18- to 24-year-olds watch the most online video. Today's younger audience is tomorrow's mainstream audience and is a leading indicator of the future. Finally, online viewers prefer free content but will pay for certain types of content and for instant gratification. However, according to comScore, they are also more engaged with the advertising they watch.

Adaptive streaming
A recent comScore study found that consumers prefer the picture and sound quality of broadcast television to Web video. Historically, picture quality has not been a strong point of Web video, but this has changed over the past two years. Now, anyone with a broadband connection and a recent computer can watch Web video that exceeds the quality of SD broadcast television. The technology that makes this possible is HTTP-based adaptive streaming, and it promises to bring together broadcast-quality video with the transformative power of the Web.

Popular video sites like YouTube do not actually stream video. When a user plays a video, it is progressively downloaded to his computer like a file transfer. As soon as the user's computer starts to download the content, it starts to play it back. The benefits of progressive download are that it is easy and works with any Web server. The disadvantages are that different viewers can get wildly varying playback experiences; some of the content that the broadcaster is paying to transfer may be wasted if it is not watched, which is akin to burning money; and it does not work for live content.

Streaming overcomes some of the limitations of progressive download. A streaming server will deliver the exact portion of content that the user wants to play, so no data transfer is wasted. Streaming also handles live and on-demand content.

However, traditional streaming has some important limitations:

  • Broadcasters providing a stream with a single bit rate need to accommodate the bandwidth constraints of their viewers, which could be anything from a 3G cellular connection to fiber. This one-size-fits-all approach means that some users will not be able to stream the content well because the bit rate chosen by the broadcaster is too high, and users with greater bandwidth than the stream bit rate will not enjoy the highest possible quality.

  • Broadcasters providing a choice of streams will either confuse or disappoint users who try to stream at the highest bit rate even if their bandwidth is insufficient to do so reliably.

  • Streaming protocols are frequently blocked by firewalls, which the user may not have control over.

  • Many dedicated streaming servers are required to scale content distribution to thousands of geographically dispersed viewers.

  • Building out and managing a separate streaming server infrastructure can be expensive.

HTTP-based adaptive streaming is a better way to deliver content to viewers. The goals of HTTP-based adaptive streaming are to efficiently deliver high-quality video to any viewer and scale using standard Internet infrastructure.

HTTP-based adaptive streaming is not one technology; rather it encompasses a set of concepts related to the encoding, preparation and transmission of Web video. These concepts have implemented it in different ways and are called different names.

Fundamentally, they offer:
  • More efficient data transfer
    As soon as a viewer stops watching, the data transfer stops, saving transmission costs.

  • Multiple bit rates
    The content is encoded at multiple bit rates, which enables different bit rates to be delivered to a viewer depending on their bandwidth, delivering the best possible picture quality for that situation.

  • Adaptive bit rates
    The stream can be switched to an appropriate bit rate for the viewer's bandwidth and other conditions, so that they always see video that does not stutter.

  • HTTP transfer
    The content is delivered over HTTP, the lingua franca of the Internet. The content scales better because it takes advantage of existing Web infrastructure and guarantees access to a viewer because it traverses Internet firewalls with no reconfiguration.

All current approaches rely on encoding streams at different bit rates and breaking them into stream fragments at regular intervals. These fragments are stored on a server and delivered to a user's Web browser over HTTP.

The differences in vendor implementation come up when one considers supported codecs, transport wrappers, the chunking process and player heuristics (how the stream is switched).

The basic workflow consists of four main steps: acquire, encode, deliver and consume (see Figure 1.) For example, the server receives VC-1 or H.264 streams at different bit rates from a hardware or software encoder, creates stream fragments and delivers them to Web caches (edge servers), which are typically part of a CDN.

Figure 1 - A basic streaming workflow includes acquisition, encoding, delivery and consumption.

When a viewer watches the video on his computer through the media plug-in in the Web browser, the edge server delivers the most appropriate stream fragment to the viewer. Which bit rate is actually sent depends on both the network and the user's computer performance and is continually evaluated by player heuristics. This enables the quality of the stream to adapt in real time to changing network and local performance conditions, resulting in seamless and optimal video playback.

The scalability to handle millions of concurrent users comes from the use of standard Internet protocols, which Internet caches know how to handle. As a result, the load on the streaming server is low (see Figure 2.) This dramatically reduces the number of servers needed even for a large event such as the Winter Olympics, which had more than 20 million unique viewers.

Figure 2 - Microsoft IIS Smooth Streaming can adapt to changing network conditions
by varying the quality of the stream (click to enlarge).

This same approach can be expanded beyond Web browsers to work with STBs and mobile devices.

Finally, HTTP-based adaptive streaming enables the use of Web analytics tools to provide real-time monitoring from encode to viewer experience. This actually exceeds what is available in the broadcast realm because visibility into individual viewer metrics is possible.

The Web has forever changed the way that viewers consume short-form, long-form, live and on-demand content. To preserve, grow and be relevant to their audiences, broadcasters need to develop Web experiences that are differentiated from their traditional broadcast output.

HTTP-based adaptive streaming enables broadcasters to deliver content on the Web that looks and sounds great, thereby increasing the time that a viewer spends on their Web property, which translates into a higher value viewer to the broadcaster.

By David Sayed, responsible for Silverlight media marketing at Microsoft, BroadcastEngineering