Ideally data in a Multi-CDN (Content Delivery Network) setup is load balanced dynamically using real-time traffic information gathered throughout the delivery chain.
This paper provides the rationale and architectural guidelines for an online Multi-CDN distribution backend, abstracting the content publication and playout logic from the actual delivery networks used. It describes the essential technical practices for a Multi-CDN setup covering the role of the Origin servers and essential broadcast features like geo-fencing, HTTPs secured traffic, cache purging, etc.
Essential metrics like video player feedback are detailed and fall back scenarios are explained. All the elements above are deployed and tested during a pilot involving 5 European broadcasters sharing a Multi-CDN overlay to load balance their traffic over 3 different CDNs using dynamic switching algorithms.
Broadcasters are increasingly relying upon online delivery. Unfortunately, using the best effort network offered by the open internet is more expensive, operationally less reliable and offers insufficient traffic capacity compared to traditional broadcast distribution methods. By switching HTTP traffic between different CDNs, based on real-time performance data and business parameters, these limitations can be overcome. Stacking of CDNs improves redundancy, increases availability and capacity which should improve the audience’s quality of experience while driving costs down.
From a cost perspective, it makes sense to allocate as much traffic as possible to a single supplier, but this creates an operational risk as all traffic is run through a single arrangement. With a Multi-CDN setup, the content provider can switch to a lower cost CDN if the quality is good enough. Automatically applying real-time traffic data in combination with business rules enables a dynamic optimal choice of the data flow. It leverages fluctuating live capabilities of different networks and increases operational control.
As well as optimising for quality, the Multi-CDN load balancing solution can be used to maximise different bandwidth arrangements, i.e. CDN operators or peering relations. It will be able to fill the ‘pipes’ efficiently, to comply with different contractual commitments and switch networks dynamically when performance is not meeting the required standards. One does not have to bet on the service of a single provider; on the contrary, it will be possible to add promising new providers or remove low performers.
However, it is not all good news with this model, as the switching layer introduces new costs that can be substantial if it has to be operated by a single content provider. Also the upscaling from a single to Multi-CDN introduces extra complexities as roles and responsibilities get distributed over different organisational entities.
The technical load balancing solution should provide the tools to manage this situation effectively. But it will also impact the choice of partners as potential competitors need to complement each other and cooperate in a single service to work on common solutions.