The idea that dedicated CDNs are the only solution for high-quality video delivery over the Internet or fixed IP-based networks has been challenged by some leading industry vendors. The CDN is seen as essential by many emerging OTT providers if they are to compete with established cable- and satellite-based pay TV operators. This is because CDNs bypass the unpredictable component of the delivery infrastructure over which QoS cannot be guaranteed at present. Adaptive streaming has emerged as the best way of getting the best out of an unpredictable network, but on its own can only cope with limited variation in network performance. CDNs though are expensive because they must provide enough capacity to carry peak levels of video traffic, otherwise they too would fail to meet QoS requirements.
But Alcatel-Lucent is developing an alternative to CDNs in its Bell laboratories near Paris, by adding cache memory to IP routers to provide resilience against failure at the level of individual network hops or links. This could operate alongside adaptive streaming, but the network itself would play a greater part in QoS by being better able to absorb shocks and bursts of local congestion.
An IP network usually comprises a mesh of links connected to routers that decide which link to switch an IP packet to at each stage on the basis of the packet address or a label. Failure or congestion on any single link can cause delay at the destination, which can create video artefacts or even temporary loss of service. However if each router has its own cache storage, either in solid-state memory or a disk drive, it can continue to play out video onto downstream links in the event of a failure or congestion immediately upstream. In effect, this cache makes the network finer grained, with increased ability to cope with problems locally without disrupting the whole end-to-end path.
The practical benefit for operators lies in avoiding the need to dedicate part of the network to pay TV video delivery, instead allowing other data traffic including perhaps low-priority video, to share the whole infrastructure. This will save money given that most pay TV and telecommunication operators offer triple- or quad-play services including broadband Internet and voice over the same network.
Alcatel-Lucent has demonstrated that video performance can be just as good when capacity is shared with other data, providing efficient prioritisation mechanisms operate to ensure that it takes precedence. There is also the potential for some level of statistical multiplexing, reducing bandwidth below the level needed to provide full service to everyone at once in the knowledge that this will hardly ever be required. At least it gives operators a new option for balancing QoS against cost.
