UHD-on-5G Associated Team: Ultra High Definition video streaming on future 5G networks



The UHD-on-5G project is an associated team between researchers at INRIA and NICT; it is financed by INRIA and the Japanese Society for the Promotion of Science. The UHD-on-5G associated team started formally for Inria in January 2016 for a duration of three years ending in December 2018. However, from the japanese side, the associated team started in April 2016 for a duration of three years.

Main objectives

The aim of this collaboration is to design and develop efficient mechanisms for streaming UHD video on 5G networks and to evaluate them in a realistic and reproducible way by using novel experimental testbeds. In a nutshell, our approach leverages and extends when necessary ICN and SDN technologies to allow very high quality video streaming at large scale. We also plan to use Virtual Network Functions (VNF) in order to place easily and dynamically different functions (e.g. transcoding, caching) at strategic locations within the network. Specifically, the placement of these functions will be decided by SDN controllers to optimize the quality of experience (QoE) of users. Moreover, we plan to integrate ICN functionalities (e.g., name-based forwarding and multipath transport using in-network caching) with SDN/NFV to provide better QoE and mobility services support to users than traditional IP architectures. Monitoring mechanisms such as the Contrace tool we developed in the SIMULBED associated team will be helpful to provide an accurate view of the network at the SDN controllers side. In addition, we will build a large-scale testbed to evaluate our solutions through reproducible experimentations based on two testbeds: the ICN wired CUTEi testbed developed by NICT and the wireless FIT R2lab testbed developed by Inria.


With the tremendous bandwidth improvement made in access links and backbone networks, end users are now expecting very high-quality multimedia applications to work anywhere, and on a variety of heterogeneous devices, e.g., Ultra-high-definition (UHD)  video streaming on smartphones and 4K or 8K TV. Recently, as one of the new evolutionary network architectures, Information-Centric Networking (ICN) has been proposed to facilitate applications such as content dissemination and high-quality streaming services. However, several challenges including cooperation with in-network caching, multipath forwarding, and quality or congestion control, are still to be addressed before service deployment in the Internet. Indeed, most of related works are based on various assumptions in which, for example, target applications are only non real-time applications, the target network must be a small management network or a wired (or wireless) only network, and so on. In parallel, Software-Defined Networking and Network Functions Virtualization (SDN/NFV) are remarkable technologies to control not only communication flows but also service functions in the networks. They have been recently developed by international vendors and deployed to the market. However, as they are basically designed to work with the traditional point-to-point IP networks usually limited to a single autonomous system, they require some extensions to enable very high-quality video streaming to a large number of heterogeneous users on the wide Internet. Furthermore, the design of such communication mechanisms requires rigorous evaluation on realistic testbeds before deployment. The collaboration leverages the two platforms developed by each partner: the CUTEi ICN testbed at NICT and the FIT R2lab wireless testbed at Inria that allows to reproduce wireless experiments in an anechoic chamber.

Scientific Progress

Traceroute facility for Content-Centric Networks:

Concerning the activity on Name Based Routing, we have published an Internet-Draft at the IETF ICN Research Group, describing the  specification of Contrace, an active network measurement tool for investigating the path and caching condition in Content Centric Networks. In CCN/NDN, while consumers do not generally need to know which content forwarder is transmitting the content to them, operators and developers may want to identify the content forwarder and observe the forwarding path information per name prefix for troubleshooting or investigating the network conditions. The well-known Traceroute tool does not help for this as IP-based network tool cannot trace the name prefix paths used in CCN/NDN.  Moreover, given a source-rooted forwarding path per name prefix, specifying a forwarding source (i.e., router or publisher) for any content is difficult, because we do not always know which branch of the source tree the consumer is on.  Additionally, it is not feasible to flood the entire source-rooted tree to find the path from a source to a consumer. Furthermore, such IP-based network tool does not allow the states of the in-network cache to be discovered. The Contrace tool is designed to overcome these difficulties and investigates in particular: 1) the forwarding path information per name prefix, device name, and function/application, 2) the Round-Trip Time (RTT) between content forwarder and consumer, and 3) the states of in-network cache per name prefix. An Internet-Draft describing Contrace is in discussion at the IRTF ICN Research Group.

Streaming using In-Network Coding and Caching

Another important task in the same activity domain is to enhance streaming quality. Owing to the rapid growth in high-quality video streaming over the Internet, preserving high-level robustness against data loss and low latency, while maintaining higher data transmission rates, is becoming an increasingly important issue for high-quality real-time delay-sensitive streaming. We have proposed a low latency, low loss streaming mechanism, L4C2, specialized for high-quality delay-sensitive streaming. With L4C2, nodes in a network estimate the acceptable delay and packet loss probability in their uplinks, aiming at retrieving lost data packets from in-network cache and/or coded data packets using in-network coding within an acceptable delay, by extending the Content-Centric Networking (CCN) approach. Further, L4C2 naturally provides multiple path and multicast technologies to efficiently utilize network resources while sharing network resources fairly with competing data flows by adjusting the video quality as necessary. We validate through comprehensive simulations that L4C2 achieves a high success probability of data transmission considering the acceptable one-way delay and outperforms the existing solution. This work has been presented at the IEEE Infocom conference in May 2017.

Scalable Multicast Service in Software Defined ISP networks

In the context of the SDN-based multicast mechanisms activity, we have proposed an architectural solution to provide scalable multicast service in ISP networks. In fact, new applications where anyone can broadcast video are becoming very popular on smartphones. With the advent of high definition video, ISP providers may take the opportunity to propose new high quality broadcast services to their clients. Because of its centralized control plane, Software Defined Networking (SDN) seems an ideal way to deploy such a service in a flexible and bandwidth-efficient way. But deploying large scale multicast services on SDN requires smart group membership management and a bandwidth reservation mechanism to support QoS guarantees that should neither waste bandwidth nor impact too severely best effort traffic. We have proposed a Network Function Virtualization based solution for Software Defined ISP networks to implement scalable multicast group management. We also propose in the same paper a routing algorithm called Lazy Load balancing Multicast (L2BM) for sharing the network capacity in a friendly way between guaranteed-bandwidth multicast traffic and best-effort traffic. Our implementation of the framework made on Floodlight controllers and Open vSwitches is used to study the performance of L2BM. This work has been presented at the IEEE ICC conference in May 2017.

Towards unifying content level and network level operations

The programmability of the network to provide content level operations is highly desirable.  With the advent of virtualization and network function softwarization, the networking world shifts to Software Defined Networking (SDN) and OpenFlow is one of the most suitable candidates to implement the southbound API. In the meanwhile, the generalization of broadband Internet has led to massive content consumption. However, while content is usually retrieved via layer 7 protocols, OpenFlow operations are performed at lower layers (layer 4 or lower) making the protocol ineffective to deal with contents. To address this issue, we define an abstraction to unify network level and content level operations and present a straw-man logically centralized architecture proposal to support it. Our implementation demonstrates the feasibility of the solution and its advantage over fully centralized approach. This task performed by a Master level student has been published in the CoNext student workshop in December 2016. A demonstration was also presented at IEEE SDN/NFV conference in November 2016.

Impact of Caching on HTTP Adaptive Streaming Decisions

In the context of the Scalable and reliable high-quality video streaming activity, we have studied the impact of caching on HTTP Adaptive streaming. The interplay between caching and HTTP Adaptive Streaming (HAS) is known to be intricate, and possibly detrimental to QoE. We have studied this topic and made the case for caching-aware rate decision algorithms at the client side, which do not require any collaboration with cache or server. To this goal, we introduced an optimization model which allows to compute the optimal rate decisions in the presence of cache, and compare the current main representatives of HAS algorithms (RBA and BBA) to this optimal. This allows us to assess how far from the optimal these versions are, and on which to build a caching-aware rate decision algorithm. The results were published in IEEE Infocom Student Workshop and IEEE Globecom Workshop on QoE for Multimedia Communications in December 2016.

Resiliency in Service Function Chaining

In the context of the dynamic placement of Virtual Network Functions in the network activity, we have studied the importance of resiliency in service functions chaining. When deploying network service function chains the focus is usually given on metrics such as the cost, the latency, or the energy and it is assumed that the underlying cloud infrastructure provides resiliency mechanisms to handle with the disruptions occurring in the physical infrastructure. In this position paper, we advocate that while usual performance metrics are essential to decide on the deployment of network service function chains, the notion of resiliency should not be neglected as the choice of virtual-to-physical placement may dramatically improve the ability of the service chains to handle with failures of the infrastructure without requiring complex resiliency mechanisms. A position paper on this topic has been published in the PROCON Workshop in September 2016.

LTE support on R2lab

We have deployed the Open Air Interface 4G Cellular Stack on commercial off-the-shelf (COTS) hardware in R2lab. It is possible now to run end-to-end scenarios involving 4G operator networks. This will be very useful in the future to test the quality of experience of streaming applications on LTE networks. A reproducible demonstration has been done in the R2lab inauguration ceremony and a demo has been presented at the ACM SIGCOMM conference in August 2017 on how to deploy a 5G network in R2lab in less than 5 minutes.


Work Plan for 2018

  • Streaming using In-Network Coding and Caching : We will make an implementation of the low-latency low-loss streaming (L4C2) mechanism  for real network devices and evaluate its performance on the CUTEi testbed.
  • Scalable Multicast Service in Software Defined ISP networks : We plan to study extensions of the Lazy Load balancing Multicast (L2BM) mechanism for wireless end-users and evaluate the performance on the R2lab testbed.
  • Robust placement of Virtual Network Functions (VNFs) : We will investigate linear optimization mechanisms to chain and place virtualized functions (e.g., caches, video transcoders) and propose algorithms to decide when to migrate them in order to accommodate demand and resource availability of users.
  • Support of SDN and 4G/5G in R2lab : We will extend the R2lab testbed at Inria to support SDN/NFV functionalities by integrating an open source controller and Open vSwitches on R2lab nodes. Concerning the support of 4G/5G scenarios, we will enhance the current support of the open source OpenAirInterface (OAI) software from Eurecom, and in particular, the automatic support of the latest release of the code.
  • CUTEi ICN extensions in R2lab : We will develop CUTEi ver.2 to support connection from external PCs and smartphones. We will also study how to support part of the CUTEi ICN functionalities in the R2lab testbed. For instance, in-network caching could be integrated in R2lab as a VNF, controlled by a SDN controller.


Testbeds and Software


Visits between Partners

  • January 15-17, 2018: meeting at INRIA (visit of Hitoshi Asaeda)
  • December 4-6, 2017: meeting at NICT (visit of Thierry Turletti –Walid’s visit cancelled due to visa issue)
  • June 27-29, 2017: meeting at INRIA (visit of Hitoshi Asaeda and Kazuhisa Matsuzono)
  • February 13-15, 2017: meeting at INRIA (visit of Hitoshi Asaeda and Kazuhisa Matsuzono)
  • December 5-7, 2016: meeting at NICT (visit of Walid Dabbous, Hardik Soni and Thierry Turletti)
  • June 27-29, 2016: meeting at INRIA (visit of Hitoshi Asaeda and Kazuhisa Matsuzono)



At Inria:

  • Walid Dabbous (PI), Senior Researcher, lead of DIANA EPI. Network architectures and Testbeds skills
  • Thierry Turletti, Senior Researcher. Network architectures and Testbeds skills
  • Chadi Barakat, Senior Researcher. Measurement, QoE and network modeling skills
  • Damien Saucez, Researcher. Network architectures and network modeling skills
  • Thierry Parmentelat, Senior Engineer, working on the R2lab testbed
  • Ghada Moualla, PhD student (Nov 2015 – Dec 2018), PhD work focusing on Network Function Virtualization
  • Vitalii Poliakov, PhD student (Oct 2015 – Sep 2018), PhD work focusing on SDN for Quality of Experience-based network optimization
  • Salma Ait Lhaj, MSc student (Nov – Dec 2017), PFE work on L2BM experimentation within R2lab
  • Hardik Soni, PhD student (Sep 2014 – Dec 2017) , PhD work focusing on Software Defined Networking
  • Ahmed Amine Loukili, MSc student (Mar 2016 – Aug 2016), Master work on SDN-based content distribution


  • Hitoshi Asaeda (PI), Research Manager, Network System Research Institute. Permanent position. Network architectures and Testbeds skills
  • Kazuhisa Matsuzono, Researcher, Network architecture laboratory. Network architectures and QoE skills
  • Ruidong Li, Researcher, Network architecture laboratory. Network architectures and security skills
  • Atsushi Ooka (from Jan 2018), Researcher, Network research headquarters. Routing and caching algorithms skills
  • Shao Xun (up to Dec 2017), Researcher, Network research headquarters. Network modeling and theory skills




Comments are closed.