Design of network coding functionality for 5g networks

• Autores: Tan Do Duy
• Directores de la Tesis: María de los Ángeles Vázquez Castro (dir. tes.)
• Lectura: En la Universitat Autònoma de Barcelona ( España ) en 2018
• Idioma: español
• Tribunal Calificador de la Tesis: Carlos Jesús Bernardos Cano (presid.), José Antonio López Salcedo (secret.), Ramón Agüero Calvo (voc.)
• Programa de doctorado: Programa de Doctorado en Ingeniería Electrónica y de Telecomunicación por la Universidad Autónoma de Barcelona
• Materias:
  • Ciencias tecnológicas
    • Tecnología de las telecomunicaciones
• Enlaces
  • Tesis en acceso abierto en:  TESEO  TDX 
• Resumen

  • Network coding (NC) has recently emerged as a new solution for improving network performance in terms of throughput and reliability. Coding operations over the network information flows can be performed both at the source and at intermediate nodes where re-encoding operations allow capacity achievability both for noiseless and noisy networks. However, the multi-user nature of NC and its inherent applicability to versatile flow engineering across all layers of the protocol stack, call for novel wireless system design approaches. The goal of this thesis is to study the design of NC as a network functionality offered to the 5G wireless communication service designers. The design would facilitate the control of network throughput, reliability, and connectivity over 5G wireless networks. The main objectives of this thesis are: (i) to develop a packet-level NC architectural design that could work as a traditional network function (NF) or be easily integrated in current proposals of virtualized network architectures, (ii) to develop a matricial model that allows us to analyze the corresponding error probabilities of different NC schemes over multi-hop line networks, a simple but common in practice network model, (iii) to develop a methodology to compute and evaluate the finite-length coding performance of different NC schemes which could be selected as the operational NC scheme for the designed network coding functionality (NCF), and (iv) to apply the proposed NCF design and validate its performance for a complete use case in 5G networks.

    The contributions of this thesis are the following. We first develop a design of Network Coding Functionality as a toolbox of NC design domains and show how it can be integrated in current virtualized infrastructures. Second, we evaluate the finite-length performance of different network codes using random vs Pascal matrices. We model the encoding, re-encoding and decoding process of different coding schemes in matrix notation and derive the theoretical expression of packet loss rate after decoding at the destination node. We then propose a binary searching algorithm to identify optimal coding rate for some specific target packet loss rates given a pre-defined coding block-length. We will focus on capacity-achieving codes and coding schemes with scheduling for representative scenarios and show the achievable rate-delay trade-off between random codes and structured codes with scheduling. In the last part of this thesis, we validate the proposed NCF design for a complete use case to enhance connectivity of Mobile Ad-hoc Network (MANET) devices over converged satellite-cloud networks in emergency applications. The key insight is that in an emergency scenario there may not be direct access to fog or cloud computing, which will then be provided via satellite and the only local computational resources available are the MANET devices. To solve this situation, we define a packet-level NCF with inputs from data service quality targets, local computation constraints and per-path statistics. Outputs are centrally-optimized coding rates balancing per-node computational resources and resulting coverage.