On the Impact of Redirection on HTTP Adaptive Streaming Services in Federated CDNs

Abstract

Abstract. HTTP Adaptive Streaming (HAS) refers to a set of novel streaming services that allow clients to adapt video quality based on current network conditions. Their use of existing HTTP delivery infrastructure makes them perfectly suited for deployment on existing Content Delivery Networks (CDNs). Nevertheless, this leads to some new challenges, related to the distribution of content across servers and the latency caused by request redirection. The federation or interconnection of CDNs proliferates these problems, as it allows content to be distributed across networks and increases the number of redirects. This paper focuses on the second problem, assessing the impact of redirection on the Quality of Experience of HAS in CDN interconnection scenarios. Additionally, several novel inter-CDN request routing policies are proposed that aim to reduce the number of redirects. Our results indicate that redirection latency significantly impacts performance of HAS and more intelligent routing mechanisms are capable of solving this problem.

QoE-driven in-network optimization for Adaptive Video Streaming based on packet sampling measurements

“…Our approach further increases the efficient use of the ISP's network by minimizing the negative impact of competing sessions on QoE. Famaey et al assess the impact of increased latency on QoE caused by the redirection of HAS requests in CDNs and propose updated request routing schemes in order to reduce the number of redirects [25].…”

“…Since the decision variables a c,q are modeled as binary values, multiplying this decision variable with the corresponding µ c,q and σ c,q values allows a linear objective function. To this end, a history H c of previous quality decisions is required for every client c ∈ C. This history allows calculating the average quality µ c,q for every client c ∈ C and every possible quality decision q ∈ Q v for the next timeslot in Equation (25). The variation in quality σ c,q can be calculated for each client using the set of averages µ c,q for each quality decision q ∈ Q v as shown in Equation (26).…”

Collaborative cloud service model for delivering multimedia content in mCloud

“…The service providers that deliver content adaptation on a wide range of devices and networks, allow learners to receive improved user experience, with minimized storage and maintenance requirements on the server side. [8]. Their research has showed that the increased segment duration can have negative effects of on video quality when using the UpstreamRR policy [8].…”

“…[8]. Their research has showed that the increased segment duration can have negative effects of on video quality when using the UpstreamRR policy [8]. It also leads to increased start-up delay, slower convergence to the optimal quality and most importantly a significant increase in lag of streaming sessions compared to the live time [8].…”

“…Their research has showed that the increased segment duration can have negative effects of on video quality when using the UpstreamRR policy [8]. It also leads to increased start-up delay, slower convergence to the optimal quality and most importantly a significant increase in lag of streaming sessions compared to the live time [8]. With the growth of traffic produced by multimedia applications such as web and mobile TV, video on demand, personal video recording and targeted advertising services there is a challenge using different networks to deliver the multimedia content to a variety of user devices.…”