PyDES

A pythonic discrete-event simulation suite

Coursework in Performance Modeling of Computer Systems and Networks

Requirements

• Python 3.6+

Build

Install all required packages with PIP, running:

$> pip3 install -r requirements.txt

Simulations

PyDES provides the user with the following simulation models:

• cloud: a simulation about Cloud computing

Launch a simulation, running:

$> python simulation.py [MY_SIMULATION] --config [MY_CONFIGURATION]

where [MY_SIMULATION] is the name of the simulation to launch, i.e. the package name contained in pydes.exp.simulation, and [MY_CONFIGURATION] is the relative path to the YAML configuration file for the simulation.

For example, to launch the cloud simulation, run:

$> python simulation.py cloud --config simulations/cloud/sample.yaml

Configuration

We state here a sample configuration, that is the one specified by experiments/cloud/simulation.yaml:

general:
  t_stop: 50000
  replica: 3
  random:
    generator: "MarcianiMultiStream"
    seed: 123456789
cloudlet:
  n_servers: 20
cloud:
  t_service_rate_1: 0.75
  t_service_rate_2: 0.85
  t_setup: 95

Experiments

The package provides experiment on randomness and simulations. In package 'pydes.exp.rnd' you can find experiments on multi-stream Lehmer pseudo-random generator. In package 'pydes.exp.simulation' you can find experiments on the simulated system.

Run the experiments and visualize results through the MATLAB Live Script pmcsn.mlx.

• pydes.exp.rnd.modulus: Find a suitable modulus for a multi-stream Lehmer pseudo-random generator, given the number of bits.
• pydes.exp.rnd.mulfind: Find suitable FP, MC, FP/MC multipliers for a multi-stream Lehmer pseudo-random generator, given a modulus.
• pydes.exp.rnd.mulcheck: Check FP, MC, FP/MC constraints for multipliers for a multi-stream Lehmer pseudo-random generator, given a modulus and a multiplier.
• pydes.exp.rnd.jmpfind: Find a suitable jumper for a multi-stream Lehmer pseudo-random generator, given a modulus, a multiplier and a number of streams.
• pydes.exp.rnd.extremes: Find a suitable jumper for a multi-stream Lehmer pseudo-random generator, given a modulus, a multiplier and a number of streams.
• pydes.exp.rnd.kolmogorov-smirnov: Find a suitable jumper for a multi-stream Lehmer pseudo-random generator, given a modulus, a multiplier and a number of streams.

To run an experiment:

$> python3 pydes.py [EXPERIMENT_NAME] [EXPERIMENT_OPTIONS]

Results

Bits	Streams	Modulus	Multiplier	Jumper	Jump Size	Spectral Test	Test of Extremes	Test of Kolmogorov-Smirnov
32	128	2147483647	16807	188756	16776028	Failed	Failed (91.406% confidence)	Succeeded
32	128	2147483647	48271	40509	16775552	Succeeded	Succeeded (96.875% confidence)	Succeeded
32	128	2147483647	50812	15707	16769483	Succeeded	Succeeded (97.656% confidence)	Failed
32	256	2147483647	16807	36563	8335476	Failed	Succeeded (95.312% confidence)	Succeeded
32	256	2147483647	48271	22925	8367782	Succeeded	Failed (92.969% confidence)	Succeeded
32	256	2147483647	50812	29872	8362647	Succeeded	Failed (94.531% confidence)	Succeeded

Sample Simulations

Performance Analysis (batchdim)

ALGORITHMS=(1 2)
BATCHDIMS=(10 20 30 40 50)

for algorithm in $ALGORITHMS; do
    for batchdim in $BATCHDIMS; do
        CONFIG="config/performance_analysis_${algorithm}.yaml"
        OUTDIR="out/performance_analysis/algorithm_${algorithm}/batchdim_${batchdim}"
        PARAMETERS="'{\"general\":{\"batchdim\": ${batchdim}}}'"
        ./pydes.py simulation-performance --config $CONFIG --outdir $OUTDIR --parameters '{"general":{"batchdim": ${batchdim}}}'
    done;
done;

Performance Analysis (thresholds)

ALGORITHMS=("2")
THRESHOLDS=(1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20)

for algorithm in $ALGORITHMS; do
    for threshold in THRESHOLDS; do
        CONFIG="config/performance_analysis_${algorithm}.yaml"
        OUTDIR="out/performance_analysis/algorithm_${algorithm}/threshold_${threshold}"
        PARAMETERS='{"system":{"cloudlet":{"threshold": ${threshold}}}}'
        echo "./pydes.py simulation-performance --config $CONFIG --outdir $OUTDIR --parameters $PARAMETERS"
    done;
done;

Algorithm 1

./pydes.py simulate-performance --config config/performance_analysis_1.yaml --outdir out/performance_analysis/algorithm_1 --parameters '{"general": {"batches": 64, "batchdim": 128}, "system":{"cloudlet": {"n_servers": 20}}}'

Algorithm 2

./pydes.py simulate-performance --config config/performance_analysis_2.yaml --outdir out/performance_analysis/algorithm_2/threshold_20 --parameters '{"general": {"batches": 64, "batchdim": 128}, "system":{"cloudlet": {"n_servers": 20, "threshold": 20}}}'

Analytical Solution

Algorithm 1

./pydes.py solve-cloud-cloudlet --config config/analytical_solution_1.yaml --outdir out/analytical_solution/algorithm_1

Algorithm 2

./pydes.py solve-cloud-cloudlet --config config/analytical_solution_2.yaml --outdir out/analytical_solution/algorithm_2/threshold_20
./pydes.py solve-cloud-cloudlet --config config/analytical_solution_2.yaml --outdir out/analytical_solution/algorithm_2/threshold_2 --parameters '{"system":{"cloudlet":{"n_servers": 2, "threshold": 2}}}'

Validation

Algorithm 1

./pydes.py validate-cloud-cloudlet --analytical-result out/analytical_solution/algorithm_1/result.csv --simulation-result out/performance_analysis/algorithm_1/result.csv --outdir out/validation/algorithm_1

Algorithm 2

./pydes.py validate-cloud-cloudlet --analytical-result out/analytical_solution/algorithm_2/threshold_20/result.csv --simulation-result out/performance_analysis/algorithm_2/threshold_20/result.csv --outdir out/validation/algorithm_2/threshold_20

Contributing

Install pre-commit

pip install pre-commit

Install pre-commit hook for the local repo:

pre-commit install

For the first time, run pre-commit checks on the whole codebase:

pre-commit run --all-files

Authors

Giacomo Marciani, mgiacomo@amazon.com

References

• "Discrete-Event Simulation", 2006, L.M. Leemis, S.K. Park
• "Performance Modeling and Design of Computer Systems, 2013, M. Harchol-Balter

License

The project is released under the MIT License.