generate_sensitivity.py

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

SRV = 4000
DUR = 192

# Load data
df = pd.read_csv('Paper_Workspace/resources/machine_usage_sample_clean.csv')
df['frac'] = df['cpu_util_percent'] / 100.0

p_idles = [100.0, 150.0, 200.0]
labels = ['Optimistic (100W)', 'Baseline (150W)', 'Pessimistic (200W)']
res = []

for p, l in zip(p_idles, labels):
    w_watt = p * (1 - df['frac'])
    # Apply Power Usage Effectiveness (PUE) of 1.2
    w_mwh = (w_watt.mean() * SRV * DUR * 1.2) / 1000000
    res.append({'Scenario': l, 'Wasted_MWh': w_mwh})

d_res = pd.DataFrame(res)

plt.figure(figsize=(9, 5))
sns.barplot(data=d_res, x='Scenario', y='Wasted_MWh', palette='YlOrRd')
plt.title('Sensitivity Analysis: Wasted Energy vs P_idle Assumption (8 Days)')
plt.ylabel('Wasted Energy (MWh)')
plt.xlabel('Idle Power Scenario')

for p in plt.gca().patches:
    plt.gca().annotate(f'{p.get_height():,.0f} MWh', 
                       (p.get_x() + p.get_width() / 2., p.get_height()), 
                       ha='center', va='center', xytext=(0, 8), textcoords='offset points')

plt.tight_layout()
plt.savefig('Paper_Workspace/resources/figure8_sensitivity_analysis.png', dpi=150)
print("Saved figure8_sensitivity_analysis.png")