An Sdn Based Approach to Measuring and Optimizing Abr Video Quality of Experience

The volume of video traffic is continuing to grow rapidly over cable networks. While a majority of IP video on DOCSIS networks today is over-the-top video from third parties, operators are increasing the amount of programming their subscribers can access on IP video-capable devices. Additionally many operators are pursuing IP video deployments that are at various stages from planning, testing, trialing to deploying. Most OTT deployments use Adaptive Bit Rate (ABR) technology (also known as HTTP Adaptive Streaming). And ABR has also been adopted by cable operators as the technology of choice as they begin their migration to IP video. ABR video was primarily developed to work as well as possible despite the network. This approach made sense for over-the-top content providers who have no control or influence of the network. Such a server-client method has so far worked well for OTT providers. Some of the challenges with such an ABR delivery method are masked because of end-users’ acceptance of a poorer userexperience since OTT providers are perceived to be lower cost options. Also the amount of ABR video on the cable network today, while significant, will still be dwarfed by the amount of ABR video that will come on to the network when cable operators migrate to a ABR-based IP video delivery method. Hence the combination of more ABR traffic along with higher user-experience expectations of a cable subscriber, may pose challenges to cable operators as they deploy ABR-based IP video. Another concern with ABR video is that different segments of the network can have similar network utilization levels yet there could be a large difference in end user Quality-of-Experience (QoE) between these segments of the network. Currently operators do not have good visibility to the actual QoE of subscribers; instead they primarily monitor network utilization levels to identify those segments that need to be upgraded. Operators need better tools to identify where QoE is below their service objectives so that they can target their network investments accordingly. In this paper, we present a novel SDN-based solution to solve some of the challenges that we anticipate will occur with heavy ABR usage. Our proposal will help operators to improve their visibility of QoE and optimize their network investments. We also present a technique to help operators improve aggregate QoE of all users on their network. Alternatively this method can be used to pack significantly more streams in the given bandwidth with comparable quality to conventional ABR. Our studies indicate that with this approach bandwidth requirements can be reduced by a third or more, thereby saving DOCSIS channels and HFC spectrum, and reducing the cost of the overall solution.