This repo is divided into

• client-scripts/, which contains scripts I wrote for client probing of the network on a Windows machine. If not on Windows, you are free to still use the JSON files and/or map_mcs_table.py, but if you are on Linux just prefer to use iw.
  • If on Windows, py measure_mcs.py to get live MCS inference for current network.
• CRUSP/, which houses the main codebase cloned from https://gitlab.com/gitlabwolf/infrastructure-for-mobile. The code has been minimally adapted (extracted Dockerfile dependencies, replaced HTTP request library, disconnected auxillary services) such that the Rust binaries can be run locally with cargo as detailed in the next section.
  • CRUSP/measurement_server contains the code for spawning the CRUSP server
  • CRUSP/measurement_client_standalone contains the code for spawning a CRUSP client.
  • All other subdirectories either support/link with these two directories, such as measurement_client_rust and measurement_shared_rust, or are non-essential auxillary services, such as settings-service and database-service.
  • If you would like to try the full CRUSP experience, clone directly from https://gitlab.com/gitlabwolf/infrastructure-for-mobile and follow the READMEs to set up the Android application.

First, to get CRUSP running, spawn the server using cargo. You may need to install cargo and run cargo update before this step.

## In root dir ##

cd CRUSP/measurement_server
RUST_LOG=DEBUG cargo run

Will spawn a CRUSP server running at localhost:8099 that you can test with local CRUSP clients. To expose the server to your LAN/WLAN, instead run cargo run --release, which exposes the CRUSP server at your local IP and port 8099. This IP is specified in CRUSP/measurement_server/Rocket.toml. For example, mine looks like:

[development]
address = "localhost"
port = 8099
log = "normal"
#workers = [number of cpus * 2]
#keep_alive = 5
#secret_key = [randomly generated at launch]
#limits = { forms = 32768 }

[production]
address = "192.168.1.156"
port = 8099
log = "normal" # normal, debug or cricitcal
workers = 16 # [number of cpus * 2] - hossman has 8 cores
keep_alive = 0 # indicating the minimum amount of time an idle connection has to be kept opened (in seconds), no multiple access is needed
# secret_key = [randomly generated at launch]
limits = { forms = 32768 } # maximum amount of data rocket will accept for a givewn data type

Simply change the address field under the production block with your local IP, which you can find using ifconfig on Linux.

With the server running, you can test a client connection:

## In root dir ##

cd CRUSP/measurement_client_standalone
RUST_LOG=DEBUG cargo run -- -h 192.168.1.156 -p 8099

You can configure host with the -h flag and port with the -p flag. You can also toggle downlink vs uplink tests with the -u flag. For a full list of options, check CRUSP/measurement_client_standalone/src/main.rs.

You are free to create your own scripts to perform batches of experiments, or you can re-use mine. I provide the following:

• CRUSP/measurement_client_standalone/measure.sh
• CRUSP/measurement_client_standalone/run_experiments.sh
  • These two spawn CRUSP clients to connect to an already running CRUSP server and dump grep'd output of measured/estimated ABW and packets lost. Examine each script to see which suits your use case. run_experiments.sh also includes an iperf dump. The former dumps output in ./output/, the latter in ./output/mcs/
  • The one and only cmd line argument to each script is the name of the output file to dump results in.
• CRUSP/measurement_client_standalone/output
  • Contains raw output of experiment results
  • analysis.py has two methods that aggregate/summarize (and in the case of distance-based, plot) the data in the output folder. Feel free to use it or adapt it.
  • CRUSP Workbook.xlsx is just a working place - see the report in the root of this repo for results discussion.