performance

moNNT.py Performance Test Suite

This test suite evaluates the performance of moNNT.py <-> dtnd intercommunication. Due to the architecture of the middleware, there are several potential bottlenecks that are examined here.

Build the container

$ docker build . --tag monntpy-perf

Running the Evaluation

There are several ways to run the evaluation. Essentially, there is a single-run mode and an experiment mode.

A single run can be executed with

$ ./run.sh

Generated logs are kept in logs/.

To run a series of experiments with different sized batches of articles:

$ ./run.sh --run-mode experiment

All the details are contained in the usage string::

$ ./run.sh --help                              
Run performance tests with moNNT.py and dtnd

USAGE:
    ./run.sh [OPTIONS]

OPTIONS:
    -a, --num-articles   Number of articles to send (single run mode only) [default: 100]
    -d, --db-path        Path to database [default: /app/moNNT.py]
    -m, --run-mode       Run mode: single or experiment [default: single]
    -h, --help           Show this message and exit

Generating New Test Data

For the evaluation of data ingestion, CBOR payload data can be created by creating ingest.json to contain the desired data. Then run

$ poetry run python convert-json-to-cbor.py

This will produce a ingest.cbor with uncompressed body data and an ingest_zlib.cbor with zlib body compression.

Sample articles can be found in here (used for spool and sequential sender testing). In the current version, only the med_text article is used.