README.md

• Notebooks to generate figures
• Probabilistic distribution figures
  • Generate models
  • Generate plots
• Utility timeline and other figures
• Breakdown figures
• Solver figures

Notebooks to generate figures

• plot_utility_timeline.ipynb: Figure 1, 2, 4b, 11
• roadmap.drawio: Figure 3
• plot_utility.ipynb: Figure 4a, 6
• plot_solver.ipynb: Figure 5
• plot_prob_pred.ipynb: Figure 8
• plot_hierarchical_utility.ipynb: Figure 7
• figure.drawio: Figure 9
• plot_stats.ipynb: Figure 10, 12, 13, 15 and Table 7
• plot_stats_mixed.ipynb: Figure 14
• plot_breakdown.ipynb: Figure 16
• plot_stats_large.ipynb: Table 8 (100 jobs)
• plot_stats_32vms.ipynb: Table 8 (20 jobs)

Probabilistic distribution figures

Generate models

$ python pred/train.py data/azure/2019/scaled_by_day8_max40_sample8_day8-12.pkl --tool=darts --context-len=60 --pred-len=40 --lr=1e-4 --epochs=200 --blocks=2 --stacks=3 --layers=4 --model-name=nhits --layer-width=256 --batch-size=32 --dropout=0.1
$ python pred/train.py data/azure/2019/scaled_by_day8_max40_sample8_day8-12.pkl --tool=darts --context-len=60 --pred-len=40 --lr=1e-4 --epochs=200 --blocks=2 --stacks=3 --layers=4 --model-name=nhits --layer-width=256 --batch-size=32 --dropout=0.1 --likelihood=gaussian

This will output trained models at results/pred/scaled_by_day8_max40_sample8_day8-12.

Generate plots

Run plot_prob_pred.ipynb

Utility timeline and other figures

See plot_utility_timeline.ipynb, plot_stats.ipynb, plot_stats_mixed.ipynb

Breakdown figures

See plot_breakdown.ipynb

Solver figures

See plot_solver.ipynb