Section: New Results

Experimental Environment for Future Internet Architecture

Participants : Walid Dabbous, Thierry Parmentelat, Baris Metin, Fréderic Urbani, Daniel Camara, Alina Quereilhac, Shafqat Ur-Rehman, Thierry Turletti, Julien Tribino.

• SFA Federation of experimental testbeds

  The OneLab2 project has come to its end in spring 2010. We are now involved in the NOVI (E.U. STREP) project, the F-Lab (French A.N.R.) project, and have the lead of the “Federation” WorkPackage of OpenLab (E.U. IP) project. Within these frameworks, we are codevelopping with Princeton University a reference implementation for the Testbed-Federation architecture known as SFA for Slice-based Federation Architecture. As a sequel of former activities we also keep a low-noise maintenance activity of the PlanetLab software, which has been running in particular on the PlanetLab global testbed since 2004, with an ad-hoc federated model in place between PlanetLab Central (hosted by Princeton University) and PlanetLab Europe (hosted at INRIA) since 2007.

  During 2011 we have focused on the maturation of the SFA codebase, with several objectives in mind. Firstly we have contributed to a major overhaul of the specification as defined essentially within the GENI (N.S.F.) Project, with participations from all over the world. These changes, that affected both the core API and the schema used to expose and manage resource specifications, aimed at reaching a mature level of interoperabity between the PlanetLab world and the EmuLab a.k.a. ProtoGeni world that has its own implementation, and are now available in SFA-2.0 issued late 2011.

  Secondly, the SFA codebase has been redesigned to provide a more generic shelter that other testbeds can easily leverage in order to come up with their own SFA-compliant wrapper. This is perceived as a powerful means to foster further adoption of the architecture, and the Planète team has been intrumental in bringing two entirely different testbeds to the federation, namely Senslab - developped in other INRIA Project-teams - and FEDERICA, the outcome of another E.U.-funded Project. Along the same lines we are working, although more remotely, with NICTA in Australia that publishes the O.M.F. testbed for running wireless testbeds, and who are interested in adopting this federation paradigm.

  Finally, as part of the pure PlanetLab development, we have added a feature for running nodes in a 'reservable' mode, which breaks the usual best-effort PlanetLab model, but turned out very helpful both for making experiments possible, that needed a more reproducible behaviour of experiments, and also in a federation perspective, for closing the usage gap with, notably wireless testbeds, that typically have a reservable-only provisioning mechanisms.

   

• Content Centric Networks Simulation

   

  We worked this year on the extension of the DCE framework for ns-3 in order to run CCN implementation under the ns-3 simulator. DCE stands for Direct Code Execution, its goal is to execute unmodified C/C++ binaries under ns-3 network simulator. With this tool researchers and developpers can use the same code to do simulation and real experiments. DCE operation principle is to catch the standard systems calls done by the real application in the experiment and to emulate them within the ns-3 virtual network topology. Concerning CCN we use the PARC implementation named CCNx which is a well working open source software reference implementation of Content Centric Network protocol. As promised by DCE this integration of CCNx requires no modification of its code, it requires 'only' working on adding the system calls used by CCN that are not already supported by DCE. The advantage of this approach is that the integration work of CCN advanced DCE and will be useful in others completly different experiments. Another great advantage is that every evolution of the CCNx implementation is very easy to integrate, all what is needed is to compile the new source code. The next steps will be naturally to use DCE/ns-3 to evaluation CCN protocols in specific scenarii, to improve the coverage of systems calls supported by DCE, and to improve the DCE scheduler to be more realistic and to take into account CPU time spent in router queues. This work is done in the context of the ANR CONNECT project.

   

• ns-3 Module store

   

  Bake is an integration tool which is used by software developers to automate the reproducible build of a number of projects which depend on each other and which might be developed, and hosted by unrelated parties. This software is being developed with the participation of the Planète group and is intended to be the automatic building tool adopted by the ns-3 project.

  The client version of Bake is already working and the Planète group had a significant participation in its development. The contributions were in the context the addition of new functionalities, bug fixing and in the development of the regression tests. We are now starting the development of the ns-3 modules repository, which is a web portal to store the meta-information of the available modules. In the present state we have already designed and implemented the portal data basis and the main interface. It is already possible to register new modules and browse among the already registered ones.

  The web portal has to be finished, notably the part that will create the xml file that will be used to feed the bake's client. We also need to add new functionalities to the client part, to enable incremental build over partially deployed environments. As it is today, bake does not enable the user to add just one new module to an already deployed version of the ns-3 simulator. This work is done in the context of the ADT MobSim in collaboration with Hipercom and Swing Inria project-teams.

   

• The ns-3 consortium

   

  We have founded last year a consortium between INRIA and University of Washington. The goals of this consortium are to (1) provide a point of contact between industrial members and the ns-3 project, to enable them to provide suggestions and feedback about technical aspects, (2) guarantee maintenance of ns-3's core, organize public events in relation to ns-3, such as users' day and workshops and (3) provide a public face that is not directly a part of INRIA or NSF by managing the http://www.nsnam.org web site. This web site is now finalized. However, activities related to developing the consortium have slowed down during 2011 due to the leave of Mathieu Lacage. We plan to put more resources on this aspect in 2012.

   

• Using Independent Simulators, Emulators, and Testbeds for Easy Experimentation

   

  Evaluating new network protocols, applications, and architectures uses many kinds of experimentation environments: simulators, emulators, testbeds, and sometimes, combinations of these. As the functionality and complexity of these tools increases, mastering and efficiently using each of them is becoming increasingly difficult.

  We designed the preliminary prototype of the Network Experiment Programming Interface (NEPI) whose goal is to make easier the use of different experimentation environments, and switch among them easily. NEPI intends to make it possible to write a single script to control every aspect of a potentially mixed experiment, including a hierarchical network topology description, application-level setup, deployment, monitoring, trace setup, and trace collection. We showed how a single object model which encompasses every aspect of a typical experimentation workflow can be used to completely describe experiments to be run within very different experimentation environments. The development of NEPI started in 2009 with the implementation of the core API, an address allocator, a routing table configurator, but also a prototype ns-3 backend driven by a simple graphical user interface based on QT. Last year, we validated and evolved the core API with the addition of a new backend based on linux network namespace containers and stabilized the existing ns-3 backend. This year we have enhanced the design of NEPI and provided experiment validation, distributed experiment control, and failure recovery functionalities. In particular, we enforced separation between experiment design and execution stages, with off-line experiment validation. We also introduced a hierarchical distributed monitoring scheme to control experiment execution. We implemented a stateless message-based communication scheme, and added failure recovery mechanisms to improve robustness. The NEPI approach has been validated by implementing support for three complementary environments: a physical testbed, a network emulator, and a network simulator. Furthermore, we showed with a concrete experiment use case, available online for reproduction, how easy it is with NEPI to integrate these environments for hybrid-experimentation [48] .

   

• Guidelines for the accurate design of empirical studies in wireless networks

   

  Traditionally, wireless protocol proposals have been often tested and validated using only analytical and simulation models [73] . However, as the wireless environment is very complex to model accurately, and since the cost of wireless cards has decreased in an exponential way, today more and more research papers include evaluation of new proposals using experimentation on real devices. Indeed, experimentation is a mandatory step before possible deployment of new network protocols with real users. However, wireless experimentation is much more complex to set up and run than simulation, and it is important to avoid many pitfalls that can occur during experimentation. The objectives of this work are twofold. First, we described typical problems currently encountered in wireless-based experimentation, and we presentd simple guidelines to avoid them [50] . Second, we proposed an experimental methodology where the detection of anomalies, calibration of the measurement setup, and clear definition of the scenario (among others) make easier the repeatability of results [55] . This work has been done in collaboration with Cristian Tala, Luciano Ahumada and Diego Dujovne from the Universidad Diego Portales of Chili, in the context of the WELCOME STIC AMSud 2011.

   

• Multicast Video Streaming over WiFi Networks: Impact of Multipath Fading and Interference

   

  We conducted an experimental study in order to analyse the impact of interference, multipath fading and path loss on multicast video streaming (i.e., goodput, packet loss and ultimately on the video quality) using off-the-shelf fixed WiFi equipment in a wireless (802.11 b/g) local area network (WLAN) environment. We used the ricean K-factor as a measure of multipath fading, spectrum analyzer to estimate channel interference and received signal strength indicator (RSSI) as indication of signal power and attenuation. In order to realistically measure forementioned metrics, we conducted extensive wireless experiments against six test cases representing common real-world situations using off-the-shelf wireless equipment.

  We showed that interference has more impact on performance than multipath fading. Multipath fading can result in considerable performance degradation in environments where moving objects cause perturbance. On the contrary, channel interference is more frequent and more prominent cause of performance degradation in wireless networks because ISM 2.4 GHz band is increasingly being utilized in homes and work places. Being able to quantify the impact of multipath fading and interference is crucial in planning, troubleshooting, managing as well as benchmarking and optimizing wireless networks. This study has been pusblished in MediaWin 2011 [51] .

   

• Making easier Experimentation

   

  Wireless experimentations are challenging to evaluate due to the high variability of the channel characteristics and its sensitivity to interferences.

  Merging traces represents a complex problem especially in wireless experimentations, due to packet redundancy in multiple probes. Merging traces solutions need to be efficient in order to process the large amount of generated traces. These solutions should provide an output data structure that allows easy and fast analysis and must be scalable in order to be used in large and various experimental settings. We have designed an algorithm that performs trace synchronization and merging in a scalable way. The algorithm output is stored in a configured MYSQL database allowing for smart packet trace storage. This solution reduces processing time by 400% and storage space by 200% with regard to raw trace file solutions. It has been implemented in an open source software called CrunchXML, available under the GNU General Public License v2 at http://twiki-sop.inria.fr/twiki/bin/view/Projets/Planete/CrunchXML .

   

• An Integration Framework for Network Experimentation

   

  Many different experimentation environments address complementary aspects of network protocol evaluation, but because of their disparities and complexities it is often hard to use them to reproduce the same experiment scenario.

  Simulation is often used for the evaluation of new network protocols and architectures. In order to perform more realistic simulations, modern simulators such as ns-3 integrate more detailed models and even support direct execution of real protocol code. However, such complex models require more computational and memory requirements. We have studied the feasibility of a hybrid approach based on distributing a complex simulation scenario on several nodes in a grid network. We showed that by exploiting the real time operation of the ns-3 simulator, it is possible to map such complex scenarios on grid nodes. We also proposed a basic mapping algorithm to distribute a simulation scenario in several nodes [42] .