tyurek/hbACSS

Running and benchmarking hbACSS

Setup

First, docker-compose will need to be installed if it has not been previously:

1. Install Docker. (For Linux, see Manage Docker as a non-root user) to run docker without sudo.)

2. Install docker-compose.

Next, the image will need to be built (this will likely take a while)

$ docker-compose build honeybadgermpc

Running benchmarks and generating data points for hbACSS

You need to start a shell session in a container. The first run will take longer if the docker image hasn't already been built:

$ docker-compose run --rm honeybadgermpc bash

Then, to rerun our benchmarks, you can use:

$ pytest --benchmark-save=hbavss_dummy_pcl --benchmark-min-rounds=3 --benchmark-warmup-iterations=0 benchmark/test_benchmark_hbavss_loglin.py
$ pytest --benchmark-save=pcl_detailed --benchmark-min-rounds=3 --benchmark-warmup-iterations=0 benchmark/test_benchmark_poly_commit_log.py
$ pytest --benchmark-save=hbacss2_dummy_pcl --benchmark-min-rounds=3 --benchmark-warmup-iterations=0 benchmark/test_benchmark_hbacss2_dummy_pcl.py

This will save the results under .benmarks in the same format as DataWinterfell. The last benchmark may crash if your machine does not have sufficient memory. If so, try removing 22 and 42 from short_param_list_t in test_benchmark_hbacss2_dummy_pcl.py

Generating graphs

We've included the benchmarks that were used in our paper in the DataWinterfell folder. Within Datawinterfell, the amt_benchmarks folder contains the benchmarking results for AMT, which is obtained with the described modifications in our paper. The Linux-CPython-3.7-64bit folder contains the benchmarking results we obtained from our own code.

Calling the following graphing script will recreate the plots from our paper:

$ python polycommit_loglin_gengraphs.py

Similarly, the following will generate graphs from the local benchmarks generated above and place them in the gen_graphs folder:

$ python gengraphs.py

this script will look for .benchmarks/amt_benchmarks/vssresults.csv to plot AMT results alongside ours. If this file is not found, it will only plot our results.