Collaborative Content Distribution over Wireless Networks with Minimized End-to-End Distribution Delays Amer Mouawad and Sanaa Sharafeddine Department of Computer Science and Mathematics School of Arts and Sciences, Lebanese American University Email: {amer.mouawad, sanaa.sharafeddine}@lau.edu.lb Abstract—Content distribution applications such as video streaming and file sharing are posing strict requirements on the development and enhancement of architectures and protocols for wireless networks. In this work, we address the problem of collaborative content distribution where destination content- requesting nodes collaborate by sharing over mobile-to-mobile links data blocks received from a source node, e.g., a wire- less access point. We propose a dynamic collaborative content distribution protocol that is capable of minimizing the end-to- end distribution delays in the network. Moreover, we present simulation results for three network scenarios to analyze the behavior of the proposed protocol. I. I NTRODUCTION Offering real-time multimedia streaming services to mobile and nomadic users can be critical for end users as well as for network operators and service providers in terms of quality of service (QoS), throughput, fairness, load, delay, and power consumption. Collaboration among end users is expected to play a central role along the evolution path towards the implementation of a multimedia-based application subsystem over an IP-based infrastructure of multiple converged access networks [1], [2]. Mobile-to-mobile (M2M) collaboration has been proposed and investigated to enhance the performance of content distribution over wireless access networks, e.g., see [3], [4], [5], [6], [7]. In this work, we consider a system model consisting of a source node that wants to distribute content to several destination nodes. One applicable scenario would be to have the source node a wireless LAN access point receiving content from a video server with the destination nodes being content- requesting multi-homed mobile devices (e.g., PDAs or laptops) in close proximity with respect to each other. The destination nodes can collaborate among each other by sharing data blocks in order to optimize system performance by increasing throughput, balancing load, reducing delay, and/or reducing energy consumption. The challenge is to determine the most suitable distribution strategies of the data blocks in order to optimize target performance metrics. Several approaches have been proposed for content distribu- tion in wireless access networks with mobile-to-mobile collab- oration. Mobile devices are assumed to be in close proximity so they can exchange data blocks among each other over a short-range wireless interface such as WLAN or Bluetooth. COMBINE [3] uses an opportunistic algorithm for collabora- tive group selection in order to increase system throughput. COSMOS [5] uses a dynamic collaborative broadcast algo- rithm in order to effectively deliver the content to all devices with higher throughput, lower delay, and better fairness in terms of cost sharing. The authors in [8] and [9] demonstrate the advantages of content distribution with mobile-to-mobile collaboration in terms of energy consumption. In this work, we address the problem of dynamic content distribution over collaborative wireless networks with the objective of minimizing the end-to-end distribution delays. We propose a simple, yet efficient, protocol that allows the server (content distributor) to act as a learning agent that is capable of monitoring delay changes in the network in order to dynamically update the content distribution strategy. The efficiency of the proposed protocol in terms of convergence towards the optimal solution is studied via three example network scenarios. This paper is organized as follows. The system model and the general problem description are presented in Section II. The different phases of the proposed novel collaborative content distribution protocol are explained in Section III. Simulation results for various example network scenarios are analyzed in Section IV. Conclusions are drawn in Section V. II. SYSTEM MODEL We assume a content distribution system model composed of a source node and N content-requesting destination nodes. The source node splits the content into K data blocks and distributes each block to one or more destination nodes. The destination nodes take care then of forwarding the received data blocks to a subset of the other destination nodes over mobile-to-mobile links. The data blocks can be different parts of a data file (e.g., software patch or image) or can be multiple descriptions of a video stream. The goal is to deliver all data blocks to all destination nodes. A challenging problem is to determine the distribution strategies of the data blocks among the nodes in order to minimize the end-to-end distribution delays. A distribution strategy for a given data block can be represented as a content distribution m-ary tree with parent nodes forwarding data blocks to their children nodes. One tree structure is selected 2010 17th International Conference on Telecommunications 978-1-4244-5247-7/09/$26.00 ©2009 IEEE 589