At a conference last week devoted to over-the-top (OTT) video, one expert discussed how adaptive streaming has become a key solution to Internet video delivery.
One issue facing OTT video is that while consumers may have sufficient last-mile connectivity, problems deeper in the network can lead to video stopping unexpectedly.
First-generation streaming technologies let the consumer select available speeds. More advanced techniques determine network throughput and set the video transmission rate automatically. But both use the same transfer rate for the entire transmission. As a result, if aggregate bandwidth drops, the video can stop suddenly.
The answer to that dilemma has been to adjust, or adapt, data rates on the fly.
“Adaptive streaming has emerged to deliver the highest quality video experience possible at any given moment,” Steve Francis, CEO of embedded software developer Avtrex, said at the OTT Video Conference in San Jose last Wednesday.
According to Francis, adaptive streaming is akin to the days of analog television, when the picture would get fuzzy rather than just drop out. When throughput drops, adaptive streaming reduces video quality but maintains the picture, which is what the user expects.
The entire experience can become more responsive. For example, traditional static rate video requires the user to wait thirty seconds or more to start the video as it fills the buffer. With adaptive streaming, it is possible to start the video with low quality immediately, and then increase the quality once the buffer holds enough video.
Proprietary approaches include Move Networks adaptive streaming, Widevine, and Adobe Flash Dynamic Streaming. More open approaches include Microsoft Silverlight Smooth Streaming made available under Microsoft Community Promise and Apple HTTP Live Streaming, which has been proposed to the IETF as a standard.
Adobe Flash is the dominant way of delivering video on the Web today, but Francis said that could change with the adoption of Smooth Streaming and Live Streaming.
Flash video needs to be delivered from special servers, which cost about 10 to 15 percent more for the content provider, Francis said. In contrast, the open approaches advocated by Apple and Microsoft use standard HTTP technology, making it easier to store and manage the video with existing Web management tools.
Under the hood
Adaptive streaming adjusts the quality of streaming video in one or more dimensions, including resolution, frame-rate, and encoding quality. Francis said that the content provider chooses what dimensions to adjust in order to provide the best user experience.
Video with lots of motion, such as sporting events, will adjust the image size but keep the frame rate constant. More picturesque video, such as products on the Home Shopping Network, will keep the images large but adjust the frame rate.
The video is broken into chunks 2-10 seconds long, so the client can switch between bit rates at any segment. The different versions of video are synchronized across these segments, which makes it easier to dynamically change between streams. The client measures the available bandwidth and selects streams based on the results.
Adaptive technologies require about double the storage of the highest quality stream.
Microsoft Smooth Streaming uses a “manifest” file, which describes the available rates and includes addresses for available streams. It lists one URL for each 2-seconds of the stream by default. Apple uses an M3U playlist file to do the same thing with a default length of 10-seconds. The specifications let the content provider change the segment lengths, but most implementations use the default settings.
Francis recommended choosing between the more open options advocated by Microsoft and Apple. “We like the Microsoft from an implementation point of view,” he said, “but if you plan to support the iPhone anyways, then the Apple approach could make more sense.”