Skip to content

Latest commit

 

History

History
144 lines (103 loc) · 6.22 KB

README.md

File metadata and controls

144 lines (103 loc) · 6.22 KB
Raw

ERRANT: EmulatoR of Radio Access NeTworks

ERRANT is an advanced emulator of radio access networks, tuned thank to a large-scale measurement campaign on operational mobile networks.

It uses tc-netem to install traffic shaping policies, allowing the user to choose between 32 profiles that differ for emulated operator, RAT (3g or 4g) and signal quality. The exact parameters of the shaping policies are dynamic, in the sense that they may vary at each run based on the values observed on the real network. The available profiles are saved in the model.pickle file. ERRANT can also vary parameters dynamically (every n seconds) to emulate variable networks, and simulate moving scenario using the ApplyScenario script.

An overview of the available profiles is depicted below, while you can find more in our paper. ERRANT Framework

Prerequisites

This tool runs on Linux, and builds on top of the tc-netem tool. It also uses the ifb kernel module to shape incoming traffic.

You need also Python3 with updated versions of pandas==1.2.4 and scipy==1.6.3.

ERRANT is able to emulate profiles available in the profiles.csv file, that describes their average values for latency, upload and download bandwidth. Check it to have the complete list, or see the image above.

Usage

You need to execute it as root.

Usage ERRANT for emulation without a scenario:

errant -o operator -c country -t technology -q quality -i interface [-p period] [-l logfile] [-r] [-d] [-h]

Parameters are:

  • operator: the Mobile Network Operator to emulate.
  • country: the country network to emulate.
  • technology: whether to emulate 3g or 4g.
  • quality: signal quality to emulate: bad, medium or good.
  • interface: the name of the interface where to apply shaping.
  • period: change network conditions periodically after period seconds.
  • logfile: write on a logfile the employed values. Useful with -p option.
  • -r: stop doing traffic shaping. Remove all the shaping policies.
  • -d: dry run (only print all commands that would execute).
  • -h: print help.

Note that to use the operator-agnostic models, you must specify universal for operator and country.

Usage ERRANT for emulation with a scenario:

ApplyScenario.py -s scenario.csv -i interface

Parameters are:

  • scenario.csv: the name of the csv file containing the list of profiles to be emulated.
  • interface: the name of the interface where to apply shaping.

The scenario.csv is a csv file where each row describes:

  • operator: the Mobile Network Operator to emulate.
  • country: the country network to emulate.
  • technology: whether to emulate 3g or 4g.
  • quality: signal quality to emulate: bad, medium or good.
  • duration: move to the next network profile after duration seconds.
  • period: change network conditions periodically after period seconds.

Log File

If the -l LOGFILE is set, ERRANT writes on LOGFILE a CSV log reporting the sampled values. It is useful to understand what was going on when analyzing the experimental results. The output file is in CSV format, and has the form:

time,operator,country,technology,quality,download,upload,rtt
1601642003,telenor,norway,4g,good,17580.07864961684,14188.136689046294,82.01200449624072
1601642006,telenor,norway,4g,good,22459.411614246997,16390.649528907834,64.16737469308717
1601642009,telenor,norway,4g,good,16158.579500704867,20363.285880299878,99.9752608514997
1601642012,telenor,norway,4g,good,22858.805018745887,9214.935379313456,129.7896828007226

Running a trace from file

You can use our convenience script trace_run to run a trace-based emulation, in which different models can be applied as time passes. The syntax of the command is:

trace_run [-h] [-f SCENARIO_FILE] -i INTERFACE [-d] [-n]

Arguments:
  -h, --help        show this help message and exit
  -f SCENARIO_FILE  File with the trace to run
  -i INTERFACE      Destination interface
  -d                Make a dry run
  -n                Leave last shaping policy on, when ending

The scenario file must be in csv format and contain for each line the specific of the model to apply (operator, country, rat and quality). It must also contain how long to keep the model on (duration). It can finally contain an eventual period to change network conditions periodically. Leave it blank or - to avoid doing this. See the file trace_example.csv for an example.

Examples

Run simulation with Norway Telenor 4G Good profile. Impose the profile to eth0 interface:

errant -o telenor -c norway -t 4g -q good -i eth0

Run simulation with universal 4G Good profile. Impose the profile to eth0 interface:

errant -o universal -c universal -t 4g -q good -i eth0

Run simulation with Norway Telenor 4G Good profile and periodically change network condition every 10s. Impose the profile to eth0 interface:

errant -o telenor -c norway -t 4g -q good -p 10 -i eth0

Usage ERRANT for emulation with a scenario:

ApplyScenario.py -s scenario.csv -i eth0

Limitations

Due to the use of the ifb kernel module, you can impose shaping to one interface at a time.

We still do not have profile describing 5G networks. We are working for running experiments and create some profiles.

Creation of new models

If you want to create new models based on other measurements, you can use the scripts provided in the model_creation directory, where we provide the code and instructions to generate a new model file compatible with ERRANT.

The ERRANT Paper

ERRANT has been proposed in our paper: Trevisan, Martino, Ali Safari Khatouni, and Danilo Giordano. "ERRANT: Realistic emulation of radio access networks." Computer Networks 176 (2020): 107289. It is avaiable on ScienceDirect and on ArXiv. To cite it in your work, use this BibTex entry:

@article{errant,
  title={ERRANT: Realistic emulation of radio access networks},
  author={Trevisan, Martino and Khatouni, Ali Safari and Giordano, Danilo},
  journal={Computer Networks},
  volume={176},
  pages={107289},
  year={2020},
  publisher={Elsevier},
  issn = {1389-1286},
}