Skip to main content

Network optimisation vital for OTT quality

The success of OTT video services in Europe driven by platforms such as the BBC iPlayer has put the spotlight on QoS and increased strain on IP networks. Not many OTT providers have optimised delivery infrastructures involving dedicated CDNs plugged straight into their own ADSL or fibre access infrastructures, with many relying on third-party IP WANs of some type. Furthermore the growth in use of cloud-based content hosting services from vendors such as Brightcove is increasing the network optimization challenge in some cases. Now the operator not only has to traverse a third party network but also relies on someone else to store the content and so only has access to one end of the link.

The good news is that these challenges have reinvigorated an old sector of the Information and Communications Technology (ICT) industry called WAN optimization, leading to creation of new tools designed or optimised specifically for online video delivery.

The task of WAN optimization is to ensure that applications meet QoS requirements, either by making the most efficient use of an existing end-to-end connection, or failing that by caching content closer to the user to avoid relying on the full end-to-end link for the final delivery. In the first of these two categories there are four types of optimisation: data compression, data packaging, prioritization and multicasting, all of which are highly relevant for OTT services. Compression is almost a separate category since it has nothing to do with the WAN link, but in the case of video makes OTT possible in the first place, with H.264/MPEG-4 becoming the accepted standard, enabling HD services to be delivered in around 7Mb/s, which is increasingly available for broadband access.

Data packaging optimization involves sending data in chunks of the most efficient size for the network conditional and bandwidth. This is part of the premise of adaptive bit rate streaming (ABRS), which breaks video streams into chunks that can be sent in parallel along separate paths, which tends to smooth out traffic and avoid sudden bumps of congestion at routing or switching points within an IP network. ABRS also allows the video resolution to keep in step with the capacity of the network or receiving device so that the user receives the best possible quality at a given time.

In the case of interactive services though, it may be more efficient to batch chunks of video up into larger packets in order to reduce the latency caused by waiting for acknowledgements from the receiving end. This comes into the realm of protocol optimization, where the objective is to overcome the accumulated delays waiting for acknowledgements of successive transmissions of small data packets. This is imposed by some protocols that were designed for LANs where link delay was not an issue because of the short distances involved. But when the same protocols are deployed across a WAN, the delays become a problem, making optimization necessary not just for video but also for conventional IT applications.

Finally, in the first category of optimization is multicasting, which reduces bandwidth consumption by only streaming live video once across a given link rather than unicasting it for each user that is watching. This leads on to the second category of WAN optimization, caching, which applies more to on-demand rather than live video. The idea is that if some stored content, such as a movie, is popular in a given locality, it makes sense to cache it on a server in that geographical area rather than repeatedly serving it from a central location. This again saves on WAN bandwidth.

For the greatest efficiency, caching should operate on the level of individual packets or units of data rather than whole chunks of content. Then, when a small change is made to some AV content that had already been distributed to local caches, only that change has to be sent over the network rather than the whole program content.

The growing importance of cloud-based services hosting applications and content has brought a new flavour of caching: cloud-based caching. This has arisen because most WAN optimization tools require deployment of some hardware or software at both ends of the link so that they can tune performance in both directions. But public cloud services are shared between many clients and so cannot allow any given one to deploy its WAN optimization tools onsite. Therefore, WAN optimization has to be deployed asymmetrically, from the receiving or client end only. One WAN optimization vendor, Blue Coat, has developed a system called Cloud Caching Engine, which optimizes performance by looking ahead and downloading multiple parts of a large video file in parallel and then holding them locally in cache so that can be played out as required. This avoids disruption to playout if there is a temporary glitch in the cloud service. More such tools inspired largely by the requirements of OTT video can be expected soon.