Net-MARS

(README under construction...)

Link to the publication at INFOCOM 2023: https://ieeexplore.ieee.org/abstract/document/10228861.

Net-MARS (Network Monitoring and Recovery Simulator) is a simulation framework including libraries to visualize and process real computer network topologies, simulate disruption scenarios, implement damage assessment strategies providing probabilistic and partial knowledge, recovery algorithms, and commonly used routing protocols.

The simulator is the result of the work proposed in the paper "Tomography-based progressive network recovery and critical service restoration after massive failures" published at IEEE INFOCOM 2023 by Viviana Arrigoni, Matteo Prata and Novella Bartolini. To replicate the results on the paper, find the setup_01.py file at src.experimental_setup. Compiled files are at data.infocom_2023.

Figure 1, Minnesota computer network topology, part of the available topologies in the simulator.

Getting Started with Net-MARS

In order to get started, create and activate an environment running Python 3.8. Then run the required libraries in the requirements file

pip install -r requirements.txt

Now you can run your first simulation using an input graph. The simulation can be batched as to run multiple recovery algorithms, in different setups in parallel. The output of a simulation will be a CSV file, showing the repaired elements and the flow restored in time.

To run a parallel simulation, add your setup file at src.experimental_setup. This must contain the following dictionaries.

comparison_dims = {co.IndependentVariable.GRAPH: [
                        co.GraphName.MINNESOTA
                        ],
                   co.IndependentVariable.SEED: range(5),
                   co.IndependentVariable.ALGORITHM: [
                        co.Algorithm.PROTON,
                        co.Algorithm.PROTON_ORACLE,
                        co.Algorithm.CEDAR,
                        co.Algorithm.ST_PATH,
                        co.Algorithm.SHP,
                        co.Algorithm.ISR_SP,
                        co.Algorithm.ISR_MULTICOM
                        ]
                   }

This dictionary represents the dimensions for comparison, the network topologies, the set of seeds, and the set of recovery algorithms described in the paper. The set of available algorithms is expressed in the enumeration at src.constants.Algorithm.

indv_vary = {
    co.IndependentVariable.PROB_BROKEN: [.3, .4, .5, .6, .7, .8],
    co.IndependentVariable.MONITOR_BUDGET: [20, 22, 24, 26, 28, 30],
    co.IndependentVariable.N_DEMAND_EDGES: [4, 5, 6, 7, 8],
    co.IndependentVariable.FLOW_DEMAND: [10, 15, 20, 25, 30],
}

This dictionary represents the independent variables to test, that are enumerated at src.constants.IndependentVariable. PROB_BROKEN is percentage of broken elements in the network, MONITOR_BUDGET the maximum number of monitors that can be placed, N_DEMAND_EDGES the numbe of demand edge, FLOW_DEMAND the demand flow in per demand edge. Each independent variable varies in the specified domain.

indv_fixed = {
    co.IndependentVariable.PROB_BROKEN: .8,
    co.IndependentVariable.MONITOR_BUDGET: 20,
    co.IndependentVariable.N_DEMAND_EDGES: 8,
    co.IndependentVariable.FLOW_DEMAND: 30,
}

This dictionary represents the fixed value for each variable, when another varies.

To run a simulation in parallel to every core, run with -par 1, with the MINNESOTA graph and setup01.

python -m src.main -set setup_01 -par 1

This command will produce a file in data.experiments having the following: format seed={}-g={}-np={}-dc={}-spc={}-alg={}-bud={}-pbro={}-idv={}.csv showing the seed (seed), the name of the topology (g), the number of demand edges (np), the capacity of the demand edges (dc), the capacity of the supply edges (spc), the name of the recovery protocol (alg), the budget of monitors (bud), the percentage of broken elements in the network (pbro), the independent variable according to the experimental batch (idv). The file shows the elements repaired and the flow restored in time.

To plot the results of the simulation use the scripts at src.plotting.stats_plotting.py.

Acknowledgments

Net-MARS was developed by Matteo Prata [email protected] and Viviana Arrigoni [email protected]. We do not do technical support, nor consulting and do not answer personal questions per email.