Section: New Results
Simulation and Optimization Tools
Participants : David Coudert, Olivier Dalle, Luc Hogie, Juan-Carlos Maureira, Christelle Molle, Julian Monteiro, Napoleão Nepomuceno, Brice Onfroy, Fabrice Peix, Judicael Ribault, Hervé Rivano, Michel Syska, Issam Tahiri.
In order to cope with the constant evolution and ever growing complexity and size of networks, new tools and modeling techniques are regularly developed within Mascotte . These tools are first developed to answer the internal needs of the team, but we also pay attention to the visibility and the dissemination of these tools in the scientific community.
Discrete-Event Simulation
In the domain of discrete-event simulation, our development efforts on the Open Simulation Architecture (OSA) are going on [119] (See Section 5.1 and http://osa.inria.fr/wiki/ .) In particular, we still pay a particular attention to the design and software engineering of our simulation tools[71] . Despite its general purpose, OSA is mainly motivated by our on-going research in the “SPREADS” ANR project, a three years project (with four other french teams) about evaluation and optimization of a peer-to-peer based reliable storage system: for this research, we need to run simulations of very large peer-to-peer systems, which is not possible with the currently existing simulation tools. This simulation challenge motivates in turn a collaboration on within the INRIA ARC “Broccoli” project with Institut TELECOM in Evry and the INRIA ADAM EPI in Lille. This collaboration is about very large scale deployment and instrumentation of OSA distributed simulations on Grid-computing facilities (e.g., on Grid 5000).
Since OSA is still in early ages, we have also been using and contributing to other simulation software. This includes in particular significant contributions to the Omnet++ simulator [76] , [77] . These contributions were mainly motivated by our work on the design and prototyping of a new wireless system, called Spiderman, that provides a continuous and high-speed wireless connection on-board fast moving vehicles like trains [76] .
We also designed and developed a Dynamic Routing Model Simulator (DRMSim) [113] (see Section 5.1 ), which addresses the specific problem of large-scale simulations of (inter-domain) routing models on large networks. DRMSim is a discrete-event simulator that comes with a generic routing models and implementations of BGP as well as of State-of-the-Art compact routing protocols. It relies on the Dipergrafs library, which allows efficient operations on large graphs (see Section 5.1 ). This simulator is designed in particular to address the limitations found in other simulators in terms of the number of nodes they can handle and in the models they propose.
Mobility issues in MANETs have also been studied using simulations. In [117] we have studied the impact of mobility on MANET topology. More precisely, we consider a network composed of a finite number of stations which move into a closed environment. The mobility is defined by the Random Waypoint mobility model. Thus, the aim of these works is to determine the impact of the mobility on the network connectivity. The result is an empirical formulation of two bounds on the number of connected components into the network.
In [118] we have proposed a new mobility model concerning MANETs. This article explains a new mobility paradigm based on tasks execution. This is a scheduled mobility model. We consider that each station composing a MANET has to execute mobility tasks into an environment. This environment is defined by a graph in which each edge is a move axis and each node is a location. With this kind of model, we are able to translate any existing mobility model.
Combinatorial Network Optimization
The Mascopt library has reached maturity and is intensively used inside the team for testing and evaluation of optimization programs (see Section 5.1 and http://www-sop.inria.fr/mascotte/mascopt ). During the last year we have pursued its development and the following research has been validated by implementing the algorithm with Mascopt .
In [15] several algorithms on the optimization of the capacity of wireless mesh networks have been implemented and tested using Mascopt .
In [79] , the study of the effects of the acknowledgment traffic on the capacity of wireless mesh networks has been modeled in a linear programming formulation. The implementation in Mascopt was solved using the column generation process.
