Section: New Results

Open Network Architecture

Controller load in SDN networks

Participant: Damien Saucez.

In OpenFlow, a centralized programmable controller installs forwarding rules into switches to implement policies. However, this flexibility comes at the expense of extra overhead in signalling and number of rules to install. The community considered that it was essential to install all rules and strictly respect routing requirements, hence working on making extra fast and large memory switches and controllers. Instead we took an opposite direction and came with a new vision that leverages the SDN concept and considers the network as a black box where tailored rules should be used only for network traffic that really matters while for the rest a good-enough (sub-optimal but cheap) default behaviour should be enough. In the past, we applied this vision to limit the needed memory on network switches in [7]. Lately, we proposed solutions to limit the number of exchanged messages between the switches and the controller. More precisely, in [19], we developed a distributed sampling adaptive algorithm that allows switches to locally decide if they can contact the controller or if instead they should make their own decision locally. Numerical evaluation and emulation in Mininet demonstrate the benefit of the approach. The results were published in IEEE INFOCOM 2018, April 2018.

Resilient Service Function Chains in virtual networks

Participants: Ghada Moualla, Damien Saucez, Thierry Turletti.

Virtualization of network functions has led to the whole new concept of Service Function Chaining (SFC) that aims at building on the fly network services by deploying them in the Cloud. A vast literature proposes techniques to build virtual service chains and map them into physical infrastructure to maximize performance while reducing costs. However, the resiliency of chains is not investigated. However, such service chains are used for critical services like e-health or autonomous transportation systems and thus require high availability. Respecting some availability level is hard in general, but it becomes even harder if the operator of the service is not aware of the physical infrastructure that will support the service, which is the case when SFCs are deployed in multi-tenant data centers. With this work, we propose algorithms to solve the placement of topology-oblivious SFC demands such that placed SFCs respect availability constraints imposed by the tenant. In order to be practically usable, i.e., without knowledge on future demands, we leverage the structural properties of multi-tier data-center topologies such as Fat-Tree or Sine and Leaf topologies to build fast yet efficient online algorithms. We explored two radically different approaches: a deterministic one and a stochastic one and results show that both can be used in very large scale data-centers (i.e., 40k nodes or more) and our simulation results show that the algorithms are able to satisfy as many demands as possible by spreading the load between the replicas and enhancing the network resources utilization [23].

Initial results were published in IEEE International Conference on Cloud Networking 2018, October 2018.

Privacy preserving distributed services

Participants: Damien Saucez, Yevhenii Semenko, Alberto Zirondelli.

Blockchains are expected to help in reducing dependency on centralized platforms (e.g., Uber, Airbnb). With this internship, we have designed a protocol to make a fully distributed, secured, and privacy protecting taxi service – a distributed version of Uber. The analytical study shows that in such system the privacy protection comes with an important overhead in network communications which raises reasonable doubt on the feasibility of actually using fully distributed platforms in an “internet-scale environment” even though our implementation on Android phones shows that it is technically possible to build such systems. This work is done in collaboration with the GREDEG (Groupe de Recherche en Droit, Economie, Gestion, a research center related to both the CNRS and the University of Nice-Sophia Antipolis and dealing with economic, managerial and legal aspects. See http://unice.fr/laboratoires/gredeg in French.), that is evaluating the incentives for users to move to fully distributed platforms that are privacy preserving but that require the users to play an active role in the system.

P4Bricks: Enabling multiprocessing using Linker-based network data plane architecture

Participants: Hardik Soni, Thierry Turletti, Walid Dabbous.

Packet-level programming languages such as P4 usually require to describe all packet processing functionalities for a given programmable network device within a single program. However, this approach monopolizes the device by a single large network application program, which prevents possible addition of new functionalities by other independently written network applications. We propose P4Bricks, a system which aims to deploy and execute multiple independently developed and compiled P4 programs on the same reconfigurable hardware device. P4Bricks is based on a Linker component that merges the programmable parsers/deparsers and restructures the logical pipeline of P4 programs by refactoring, decomposing and scheduling the pipelines' tables. It merges P4 programs according to packet processing semantics (parallel or sequential) specified by the network operator and runs the programs on the stages of the same hardware pipeline, thereby enabling multiprocessing. We present the initial design of our system with an ongoing implementation and study P4 language's fundamental constructs facilitating merging of independently written programs [34], [12].

Applications in ITS Message Dissemination

Participants: Thierry Turletti.

We build upon our prior work on D2-ITS, a flexible and extensible framework to dynamically distribute network control to enable message dissemination in Intelligent Transport Systems (ITS), and extend it with handover and load balancing capabilities. More specifically, D2-ITS’ new handover feature allows a controller to automatically “delegate” control of a vehicle to another controller as the vehicle moves. Control delegation can also be used as a way to balance load among controllers and ensure that required application quality of service is maintained. We showcase D2-ITS’ handover and load-balancing features using the Mininet-Wifi network simulator/emulator. Our preliminary experiments show D2-ITS’ ability to seamlessly handover control of vehicles as they move. This work has been presented at the 27th International Conference on Computer Communications and Networks (ICCCN 2018), Jul 2018, Hangzhou, China [17].

Low Cost Video Streaming through Mobile Edge Caching: Modelling and Optimization

Participants: Luigi Vigneri, Chadi Barakat.

Caching content at the edge of mobile networks is considered as a promising way to deal with the data tsunami. In addition to caching at fixed base stations or user devices, it has been recently proposed that an architecture with public or private transportation acting as mobile relays and caches might be a promising middle ground. While such mobile caches have mostly been considered in the context of delay tolerant networks, in this work done in collaboration with Eurecom with the support of the UCN@Sophia Labex, we argue that they could be used for low cost video streaming without the need to impose any delay on the user. Users can prefetch video chunks into their playout buffer from encountered vehicle caches (at low cost) or stream from the cellular infrastructure (at higher cost) when their playout buffer empties while watching the content. Our main contributions are: (i) to model the playout buffer in the user device and analyze its idle periods which correspond to bytes downloaded from the infrastructure; (ii) to optimize the content allocation to mobile caches, to minimize the expected number of non-offloaded bytes. We perform trace-based simulations to support our findings showing that up to 60 percent of the original traffic could be offloaded from the main infrastructure. These contributions were published in IEEE Transactions on Mobile Computing [16]. The part specifying the framework to a chunk-based scenario by accounting for partial storage of videos in vehicles was published in [25].

Cost Optimization of Cloud-RAN Planning and Provisioning for 5G Networks

Participants: Osama Arouk, Thierry Turletti.

We propose a network planning and provisioning framework that optimizes the deployment cost in C-RAN based 5G networks. Our framework is based on a Mixed Integer Quadratically Constrained Programming (MIQCP) model that optimizes “virtualized” 5G service chain deployment cost while performing adequate provisioning to address user demand and performance requirements. We use two realistic scenarios to showcase that our framework can be applied to different types of deployments and discuss the computational cost and scalability of our solution. This work has been presented at the IEEE International Conference on Communications, in May 2018, at Kansas City, MO, United States [18].

Slice Orchestration for Multi-Service Disaggregated Ultra Dense RANs

Participants: Osama Arouk, Thierry Turletti.

Ultra Dense Networks (UDNs) are a natural deployment evolution for handling the tremendous traffic increase related to the emerging 5G services, especially in urban environments. However, the associated infrastructure cost may become prohibitive. The evolving paradigm of network slicing can tackle such a challenge while optimizing the network resource usage, enabling multi-tenancy and facilitating resource sharing and efficient service-oriented communications. Indeed, network slicing in UDN deployments can offer the desired degree of customization in both vanilla Radio Access Network (RAN) designs, but also in the case of disaggregated multi-service RANs. We propose a novel multi-service RAN environment, i.e., RAN runtime, capable to support slice orchestration procedures and to enable flexible customization of slices as per tenant needs. Each network slice can exploit a number of services, which can either be dedicated or shared between multiple slices over a common RAN. The novel architecture we present concentrates on the orchestration and management systems. It interacts with the RAN modules, through the RAN runtime, via a number of new interfaces enabling a customized dedicated orchestration logic for each slice. We present results for a disaggregated UDN deployment where the RAN runtime is used to support slice-based multi-service chain creation and chain placement, with an auto-scaling mechanism to increase the performance. This work has been published in IEEE Communications Magazine [13].