By Jeffrey Tyre, ARRIS; Wendell Sun, Viasat
Since early deployment of IP-based video networks, various technologies have emerged to help cope with the variability associated with delivering video over non-deterministic, best effort IP networking. In managed environments, IPTV operators have traditionally used MPEG-2 TS as the transport mechanism for video over IP networks. Hyper Text Transfer Protocol (HTTP), by virtue of being the content transport protocol for web-based applications, is almost ubiquitously used for video delivery over the Internet.
Traditionally HTTP video was delivered by progressive file download. However, a newer technology called adaptive bit rate (ABR) streaming has become widely used. ABR streaming promises to enable videos to be delivered over unmanaged networks with a very high quality of experience, and is thus applicable to both Internet video environments and managed video networks that are seeking to extend the delivery of premium content to devices other than the television set. ABR streaming has emerged as a technology of choice for many types of video delivery. For managed-networks, Pay TV Operators migration to adaptive streaming fits into an overall strategic objective for a unified, video delivery IP network infrastructure supporting every device screen and all subscriber services using web and cloudbased technologies.
Unlike previous HTTP video technologies, such as progressive download, adaptive streaming introduces the ability to dynamically react to changes in network conditions by switching to a video encoded at a different bit rate. This ability to adapt in real time more accurately reflects the dynamic conditions of today’s networks, content, and devices. With users streaming more premium long form content, it is natural to expect that there will be fluctuations in the amount of bandwidth available during a two-hour movie, than say a 3-minute video clip. Adaptive streaming is a recognition of this fact and enables viewers to watch this premium content with a superior quality of experience (QoE).
Adaptive streaming works by leveraging the same content encoded in various bit rates—in a range that reflects the expected quality of the content itself, the network performance, and the screen resolution desired. For example, a video could be encoded in bit rates ranging from 300 kbps (low-quality, online video) up to 6 Mbps or higher (high quality streaming content to the TV). A typical video could been coded in as many as eight different bit rate profiles, depending on the range of devices and quality desired. Each of these files are then further segmented—or “chunked”—into short segments (typically two to ten seconds long) that are each precisely time-stamped.
As the video is delivered, the HTTP client maintains a communication channel with the adaptive bit rate server. The client (the viewing device) downloads these chunks as individual files, which are buffered by the client, decoded, and played out as a continuous presentation of video and audio to the viewer. During the viewing session, the client player monitors the rate at which the buffer is filling and can thereby infer the performance of the network.
If there is degradation in network performance, the client can request that chunks be delivered from one of the lower bit rate files. This is all seamless to the viewer since each source file is chunked and timestamped in the same, very precise intervals—so there is no visible interruption or hesitation when switching to a different bit rate. Likewise, if the player detects an improvement in performance, it can request HTTP file segments from one of the higher bit rates.
Since adaptive streaming video content is prepared and conditioned into multiple bit rate sources to allow the client player to dynamically select the appropriate bit rate source and seamlessly switch between different bit rate sources per broadband network status, it can be used by almost any type of multiscreen device, provide transport resiliency to a network’s condition and gives a much better user viewing experience.
For the video on demand (VOD) type of service, this is near perfect since VOD service is delivered to each end device individually in non-real-time fashion, e.g. IP unicast, and there is no concern for delivery latency and no demand to coordinate viewing experience among end users. The initial buffering delay most likely is acceptable, especially when there is a pre-roll ad play. The commercial application is very successful for Internet-based VOD services, such as Netflix, Hulu, Amazon, etc. However, the HTTP basedtransport has major shortcomings when it is applied for live or linear video delivery.