Algorithm benchmarking

from pydistsim.logging import enable_logger
from pydistsim.benchmark import AlgorithmBenchmark
from pydistsim.demo_algorithms.broadcast import Flood
from pydistsim.network.behavior import NetworkBehaviorModel

enable_logger()

Set-up the benchmark

• Pass algorithm, network sizes to test and the network behavior.

• Then run the benchmark.

benchmark = AlgorithmBenchmark(
    ((Flood, {"initial_information": "Hello Wold!"}),),
    network_sizes=range(1, 20),  # equivalent to [1, 2, 3, ..., 19]
    network_behavior=NetworkBehaviorModel.UnorderedRandomDelayCommunication,  # You can define their own network behavior, check behavior vizualization notebook
)

benchmark.run()

Automatically generated plots with the results

benchmark.plot_analysis()  # Create a plot

<seaborn.axisgrid.PairGrid at 0x7fe2df155bd0>

Filter the results by network type and only show specific graphs:

benchmark.plot_analysis(
    y_vars=["Qty. of messages sent", "Qty. of steps"],
    result_filter=lambda result: result["Network type"] == "homogeneous",
)

<seaborn.axisgrid.PairGrid at 0x7fe2df0b13d0>

Get the data as a pandas DataFrame for in-depth analysis

df = benchmark.get_results_dataframe()  # Get the results as a pandas DataFrame

df  # This will display the results as a table

	Net. node count	Net. edge count	Network type	Qty. of messages sent	Qty. of messages delivered	Qty. of status changes	Qty. of steps
0	1	0	complete	0	1	3	3
1	1	0	hypercube	0	1	3	3
2	1	0	ring	0	1	3	3
3	1	0	square mesh	0	1	3	3
4	1	1	square torus	1	2	3	5
...	...	...	...	...	...	...	...
349	16	24	rectangular mesh	33	34	33	63
350	18	36	rectangular torus	55	56	37	44
351	19	56	homogeneous	94	95	39	50
352	18	27	rectangular mesh	37	38	37	74
353	15	30	rectangular torus	46	47	31	39

354 rows × 7 columns