ABSTRACT

Even though there is considerable work in adaptive video streaming, video players still suffer drawbacks which include: inadequate fair share, poor bandwidth utilization, frequent level shifts and high re-buffering ratios. These draw- backs result in poor perceived video quality which degrades a viewers quality of experience (QoE). DASH approaches have been known to deliver higher QoE to viewers by improving the video players segment selection. However, the transport layer protocol is another important aspect of video streaming. Hypertext Transfer Protocol (HTTP) is the most used application layer protocol over the Internet. It utilizes Transmission Control Protocol (TCP) as its transport layer protocol. TCP variants use different mechanisms for congestion control. The DASH standard uses HTTP, thus TCP variant selection becomes very important for effective video streaming. In this paper we test the performance of four linux-based TCP variants using the Conventional, PANDA and ELASTIC client-side DASH players in congested bottleneck link conditions. These experiments illustrate the varying impact on viewer QoE when using these various TCP congestion control mechanisms. The importance of TCP variant selection is exemplified as we observe that Westwood+ and YeAH are the most promising variants.

Keywords: - QoE, DASH, HTTP, TCP, congestion, PANDA, ELASTIC, Westwood, YeAH