RESCOM 2017

"If computers of the kind I have advocated become the computers of the future, then computing may someday be organized as a public utility just as the telephone system is a public utility... The computer utility could become the basis of a new and important industry”, John McCarthy, Speaking at the MIT centennial in 1961. Because it enables Utility Computing providers to go beyond the capacity of a single machine, distributed computing became the prevalent infrastructure for delivering IT as a service in the middle of the 1990’s. However, due to the complexity to operate and use distributed computing platforms, it has taken almost two decades of research activities and technology transfers to popularize them in all domains of science and industry. Among the key enablers of this democratization are virtualization technologies. Proposed as early as Mainframes but used for about a decade in distributed computing, system virtualization had a huge impact on distributed computing. In this talk, I will give an overview of the different generations of the Utility Computing paradigm, starting from the centralized mainframe model to the cloud computing one. For each architecture kind, I will illustrate what have been the major challenges and how the distributed computing community has been addressing them. I will conclude this introduction by highlighting intrinsic limitations of current Cloud Computing infrastructures, which have been designed in somehow by levering centralized models.
These limitations as well as the requirements of new usages strongly push for a massively geo-distributed Utility Computing model.

As defined in Wikipedia, Software Defined Networking (SDN) is about “decoupling the system that makes decisions about where trafficis sent (the control plane) from the underlying systems that forward traffic tothe selected destination (the data plane)”. In this introductive talk, we will overview the motivation that has lead, less than a decade ago, to the emergenceof SDN as well as the reasons and the pragmatic compromises (most notably inthe OpenFlow Design choices) which have made it viable. The talk will then discuss recent evolution, deployment perspectives, relation with emerging network function virtualization technologies, as well as current limitations and open research issues, with specific focus to the programmability challenges of the data plane.

• Constant Colombo, René Kopp, Francis Lepage, Eric Gnaedinger (CRAN, TDF).
  Elimination of network resilience impact in a video stream transport by implementing an SDN architecture constraint by existing network - BEST POSTER AWARD
• Montida Pattaranantakul, Yuchia Tseng, Ruan He, Zonghua Zhang, Ahmed Meddahi (IMT Lille, Paris Descartes, Orange).
  A First Step Towards Security Extension for NFV Orchestrator
• Arif Ahmed, Guillaume Pierre (IRISA).
  Efficient Container Deployment in Edge Computing Platforms
• Nicolas Herbaut, Daniel Négru (LaBRI).
  vCDN Deployment and Optimization: the Case for a NFV and SDN-Ready ISP Network
• Florian Leman, Salima Hamma, Benoît Parrein (LS2N).
  A-TLSA : An Adaptation of the Two-Level Scheduling Algorithm from WiMAX to LTE Networks
• Chi Dung Phung, Stefano Secci (LIP6)
  The LISP-LAB platform for network virtualization
• Karyna Gogunska, Chadi Barakat, Guillaume Urvoy-Keller, Dino Lopez-Pacheco (i3S, Inria).
  Empowering Virtualized Networks with Measurement As a Service

Cloud computing data centers are composed of very powerful computing nodes connected by reliable backbone networks. However, these resources are concentrated in a small number of a data centers. The latency between an end user and the closest available cloud data center comes in the range of 20-150 ms. A number of latency-sensitive applications (e.g., augmented reality) require extremely low end-to-end latencies and therefore cannot make use of traditional cloud platforms. Fog computing therefore aims to complement traditional cloud infrastructures with additional resources located extremely close to the user, within a couple of network hops. This requires one to distribute machines in a very large number of geographical locations so computation capacity is always available in immediate proximity of any end user. In this presentation I will discuss the application scenarios where fog computing is or isn't useful, and the architectural challenges one needs to face when designing the next-generation fog computing architectures.

OpenStack has become the de-facto solution to operate compute, network
and storage resources in public and private clouds. However, developers and scientists are facing challenges when it comes to deploy and benchmark such a complex software stack. This lab aims at exploring EnOS, a holistic framework leveraging containers to conduct easy and reproducible evaluations of different OpenStack deployments. In particular, we will use EnOS to deploy, operate, benchmark and analyse a basic OpenStack instance.

More info here.

This is a hands-on lab showing how two major SDN controllers, Open Network Operating Sytem (ONOS) and OpenDayLight, work, that is, how network topology is discovered, how switching rules are installed and used, with some simple applications.

More info here.

The VM scheduler continuously deploys VMs to appropriate physical servers according to the established Service Level Objectives. The VM scheduler is the cornerstone of the good functioning of an IaaS cloud. The provider bases his offering and the clients base their requirements on its features. Developing a VM scheduler is however a challenging task. Above all, the developer must overcome the theoretical foundations of the problem and their consequences over the problem scalability. He must also provide an architecture that is flexible enough to support the evolving requirements of the users. The quality of a scheduler might then always be the consequence of a good match between its reasoning capabilities, the infrastructure particularities and the workload profile. It might then be vain to look for the perfect pre-packaged VM scheduler and start to package its own.
In this talk, I will discuss about VM schedulers, from their basic principles and capabilities to their implementations. I will also talk more extensively about the notion of flexibility. This will be illustrated by through BtrPlace. the VM scheduler developed by the University of Nice Sophia Antipolis since 2011 and a complete rewrite of Entropy, the project initiated in 2006 at the Mines de Nantes. BtrPlace allows third party developers to enhance its reasoning capabilities and make it support new, possibly unique concerns. Despite being initially a research prototype, BtrPlace is used in production, because of its flexibility. Nutanix, a provider of entreprise clusters uses it continuously in thousand of private clouds worldwide to mitigate local load spikes.

NFV orchestration is key in the emergence of virtualized networks: it comes as a complement to the Software Defined Network paradigm and represents the next step towards automating network management. In a virtualized network setting the network functions (routing, name resolution, firewalling, network acceleration, …) are provided as software packages, Virtual Network Functions (VNFs). This changes dramatically how network management must be addressed, as it must adopt the standard practice of software development and operations. This also opens new opportunities to achieve in practice the long standing goal of automating network management. In this talk we review the main challenges of NFV orchestration, VNF chaining and illustrate the benefits that this approach can bring to network management. The talk draws on example work done in the lab on on-line and off-line VNF placement and chaining using game theoretic approaches. It concludes with future perspectives to further improve network elasticity with adaptive NFV orchestration.

Data volumes are ever growing, for a large application spectrum going from traditional database applications, scientific simulations to emerging applications including Web 2.0 and online social networks. To cope with this added weight of Big Data, we have recently witnessed a paradigm shift in the way data is processed through the MapReduce model. First promoted by Google, MapReduce has become, due to the popularity of its open-source implementation Hadoop, the de facto programming paradigm for Big Data processing in large-scale data-centers and clouds.

The goal of this talk is to serve as a first step towards exploring the Hadoop platform and also to provide a short introduction into working with big data in Hadoop. An overview on Big Data including definitions, the source of Big Data, and the main challenges introduced by Big Data, will be presented. We will then present the MapReduce programming model as an important programming model for Big Data processing in the Cloud. Hadoop ecosystem and some of major Hadoop features will then be discussed. Finally, we will discuss several approaches and methods used to optimise the performance of Hadoop in the Cloud. Finally, we will discuss the limitations of Hadoop and introduce new Big Data systems including Spark.

Home made VM schedulers.

More info here.

In this hands-on lab, we will explore forwarding state propagation and route computation in SDN by first implementing a reactive and a proactive forwarding information diffusion mechanism. Then we will implement various routing decisions.

To understand the impact of each solution, we will compare them with the same benchmark distributed application and discover that depending on the scenario the performance of the application relying on the network may be really different.

More info here.

• Aboozar Rajabi, Claude Duvallet (LITIS).
  The AIS (Automatic Identification System) messages as big data and its applications in maritime field
• Fetia Bannour, Sami Souihi, Abdelhamid Mellouk (LiSSI).
  Software-Defined Networking: A Self-Adaptive Consistency Model for Distributed SDN Controllers
• Florian Greff, Ye-Qiong Song, Laurent Ciarletta, Arnaud Samama (LORIA, Thales).
  Software-Defined Real-Time Mesh Networking: Protocol and Experimentation Method
• Thomas Bonald, Céline Comte (Telecom Paristech, Nokia).
  A Round-Robin Scheduling for Computer Clusters with Compatibility Constraints
• Hardik Soni, Walid Dabbous, Thierry Turletti, Hitoshi Asaeda (Inria, NICT).
  Scalable Guaranteed-Bandwidth Multicast Service in Software Defined ISP networks
• Bastien Confais, Benoît Parrein, Adrien Lebre (LS2N).
  An object store for Fog infrastructures based on IPFS and a Scale-Out NAS
• Kevin Phemius, Marco Casazza, Mathieu Bouet, Stefano Secci (Thales, LIP6).
  Availability-driven Virtual Network Functions Orchestration on OpenStack