Segment-based teletraffic model for MPEG-DASH

Abstract

This paper proposes a novel segment-based teletraffic model for dynamic adaptive video streaming over HTTP (DASH), based on the MPEG-DASH standard. The functionality of the standard's framework is mapped on the proposed model for which the probabilities of buffer overflow, empty buffer and active buffer are adopted as relevant performance metrics. These metrics describe the MPEG-DASH streaming process from the teletraffic viewpoint and are observed for different segment sizes of the encoded video at the server and different buffer sizes at the client. The results show that small segment sizes of the encoded video increase the probabilities of empty buffer and active buffer, while decreasing the probability of buffer overflow, whereas large segment sizes are suitable for decreasing the probabilities of empty buffer and active buffer, while increasing the probability of buffer overflow. Regardless of the segment sizes, the performance metrics are further improved when the buffer size at the client is increased.