Skip to Main content Skip to Navigation

Adaptive streaming using Peer-to-Peer and HTTP

Abstract : The increasing growth of video traffic and the number of Internet users, besides the progressing video technologies and device capabilities, have surged the demand for improving the user Quality of Experience (QoE).Today, video traffic accounts for 79% of the global Internet traffic, and this percentage is projected to strike 82% by 2022, with Over The Top (OTT) services accounting for more than 50% of the peak download traffic globally.HTTP Adaptive Streaming (HAS) solutions have shown to be one of the essential techniques to cope with this ever-increasing video traffic, thanks to their embedded Adaptive BitRate (ABR) logic at the client-side which allows adaptation to the bandwidth oscillations and maximizing QoE.In parallel, video distribution over Peer-to-Peer (P2P) networks, along with Content Delivery Networks (CDN), is becoming more important to handle the explosion in the number of video consumers.As a result of P2P and HAS recent improvements, there have been many efforts to bring these two approaches together. However, the deployment of HAS streaming over P2P networks raises many challenges. The P2P nature is problematic due to the heterogeneity of resources and the dynamicity of peers. The layered implementations where HAS and P2P stack are isolated from each other. The P2P prefetching techniques are not aware of the used ABR logic, which leads to inefficient usage of the network resources. This thesis focuses on the layered HAS and P2P stack implementations and aims to analyze the above-mentioned issues and propose methods to solve them, enhancing QoE and P2P efficiency. To achieve this, we build a simulation environment to test HAS solutions in hybrid CDN/P2P systems and analyze the related issues. We propose Response-Delay, a method enabling usage of existing HAS algorithms in the context of prefetching-based P2P networks; Response-Delay is external to the video player and does not require any modification to the implemented ABR algorithm. Besides, we propose ML-based models to predict the quality decisions of HAS algorithms, using only a set of input metrics that the ABR can use to make a bitrate decision. Finally, we combine Response-Delay and the ML-based ABR models towards an ABR-aware prefetching and quality control technique. This technique uses the predicted ABR decision in the prefetching process and controls the ABR externally to make P2P-friendly decisions.
Document type :
Complete list of metadata
Contributor : Abes Star :  Contact
Submitted on : Monday, September 13, 2021 - 2:53:11 PM
Last modification on : Tuesday, October 19, 2021 - 11:14:15 AM


Version validated by the jury (STAR)


  • HAL Id : tel-03342614, version 1



Hiba Yousef. Adaptive streaming using Peer-to-Peer and HTTP. Multimedia [cs.MM]. Institut Polytechnique de Paris, 2021. English. ⟨NNT : 2021IPPAT017⟩. ⟨tel-03342614⟩



Les métriques sont temporairement indisponibles