Content aware optimization for video delivery over WCDMA

  • PDF / 575,205 Bytes
  • 14 Pages / 595 x 794 pts Page_size
  • 4 Downloads / 192 Views

DOWNLOAD

REPORT


RESEARCH

Open Access

Content aware optimization for video delivery over WCDMA Kartik Pandit1 , Amitabha Ghosh2* , Dipak Ghosal1 and Mung Chiang2

Abstract A content-aware networking framework for transmitting video on the uplink of a Wideband Code Division Multiple Access (WCDMA) cellular network is studied in this article. We consider a multi-user scenario and exploit the underlying structure of the video content to optimally schedule frames in order to minimize the total distortion of video quality across all users. In particular, we propose a novel Content-Aware Distortion Fair (CADF) scheme that determines, between multiple flows, the best set of frames to schedule using optimal transmit power, while meeting interference and delay constraints. The optimization problem is considered on the time scale of a single Group of Pictures, and is formulated as a restricted Multiple Knapsack Problem. A key contribution of this study is evaluating the CADF scheme on a Qualcomm 3G emulator, called “High data rate System Emulator 2” (HSE 2), and conducting rate control experiments with different types of emulated channel conditions. Our experimental results show that the CADF scheme significantly reduces video distortion, compared to the existing Foschini-Miljanic closed-loop distributed power control algorithm implemented on the WCDMA uplink. In the Appendix section, we also use the emulator to analyze the Traffic-to-Pilot resource allocation algorithm implemented on the reverse link of Evolution-Data Optimized (EV-DO) Revision-A by profiling over different traffic classes. Keywords: Content-aware networking, Rate distortion fair, Qualcomm 3G emulator, Video delivery, Power control, Scheduling, WCDMA

Introduction Motivation for content-aware networking

It is projected that mobile video traffic will comprise more than 70% of all mobile data traffic by 2016 [1]. This challenges many of our basic assumptions about designing future computer communications networks. Traditionally, network protocols have been designed to be contentoblivious, i.e., each bit is assumed to be equally important when it is transported over the network. This has given rise to the problem of content-pipe divide [2]. On the one hand, pipe-owners such as Internet Service Providers (ISPs), network infrastructure vendors, and municipalities treat all content equally as simply bits of information to be transported between given nodes. On the other hand, content-providers such as media companies, end-users who post videos online, and operators of both peer-topeer (P2P) systems and content distribution networks *Correspondence: [email protected] 2 Department of Electrical Engineering, Princeton University, Princeton, NJ, USA Full list of author information is available at the end of the article

(CDNs) generate content treating the network as simply a means of transportation. This gap between the pipeowners and content-providers is exacerbated especially for video content, where different frames can have different degrees of importance and can contribute differently

Data Loading...