Ensuring quality in IP video delivery systems
The growth of video delivered over IP networks is showing no signs of slowing down. The industry shift towards IP, combined with consumer demand for the highest quality video streamed to any device, at any time, anywhere, is forcing service providers and broadcasters to adapt to stay competitive in this evolving video landscape.
HTTP-based adaptive streaming was developed to enable high quality video delivery over the internet, and proved to be an efficient method for delivering content to smartphones, tablets and connected devices. Today it is increasingly replacing traditional IPTV solutions for delivering even the prime video services for the living room.
To ensure the best possible user experience, video service providers must address bandwidth, latency and packet loss in order to avoid quality issues such as buffering, slow responsiveness, low resolution and glitches. HTTP adaptive streaming is based on two main technologies that impact the quality management: the use of TCP as the transport protocol and the provisioning of content at multiple quality levels.
TCP is a bi-directional protocol, allowing clients to adapt to changing network conditions by requesting a suitable quality level. The TCP protocol also offers inherent retransmission capabilities, which enables it to efficiently deal with packet losses, preventing noticeable glitches.
However, to maintain a certain bandwidth capacity in the presence of packet-loss requires over-provisioned networks. Another issue impacting the available bandwidth is latency. Despite not being a major issue with user interactivity, increased latency will lower the available bandwidth. This may prevent high bitrate streaming (UHD, HD) over long-haul networks.
To ensure optimal video delivery, service providers need to consider the following techniques to address these challenges.
Resource management: properly tracking and allocating bandwidth is required in order to deliver high or continuous quality content to a large number of viewers, without over-provisioning the network. This is challenging for adaptive streaming sessions, which consist of hundreds or thousands of small fragments instead of a single continuous stream. A solution is to use ‘virtual sessions’. Deploying an agile software defined network management solution makes it possible to allocate resources, provide load balancing and monitor functions that are scaled on demand.
Proximity to client: Originating streams as close as possible to the viewers makes it possible for operators to ensure high-quality video delivery. A distributed hierarchical network stores the most popular content closest to the end user and the least valuable content deeper in the network. In addition to improving quality, caching can be a cost saver by reducing upstream bandwidth requirements.
Multicast of HTTP live streams: Popular live content often cause peaks in network traffic. To overcome this issue, distributed caching servers can fan-out streams closer to the end user. However, this may still be inefficient in larger networks. There are several ways multicast can be used to deliver live streams, at least to a nearby cache. File fragments can be delivered over multicast, or video can be delivered as several synchronised transport streams. The first option requires less intelligent caches, but consumes more network bandwidth. The latter option requires intelligent caches that can segment, encrypt and re-package content, but requires only one format to be delivered through the network. Pushing live streams over multicast to edge caches is also an efficient way to minimise the end-to-end latency for live delivery, by avoiding intermediate cache traversals. For some live events, like sports, short latency is crucial.
Measure and Analyse: Since quality decisions are made by the clients, the only way to verify the quality experienced by end users and to evaluate the impact of any optimisations introduced is to measure and analyse the traffic that was actually delivered. This can be challenging if the delivery of every little chunk needs to be monitored. A video aware analytics tool that abstracts the data at a meaningful level is key.
HTTP-based adaptive streaming enables service providers to deliver premium content to a growing number of devices and over different network types. Yet, the issue of quality is still top of mind. By considering the techniques outlined above, service providers can address these issues and keep pace with industry demand.