Files
mtusi2026lab3/student-tests/plot_results.py
T

72 lines
4.0 KiB
Python

#!/usr/bin/env python3
"""Recreate publication/export-ready figures from the measured JSON files."""
import json
from pathlib import Path
import statistics
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
ROOT = Path(__file__).resolve().parents[1]
def main():
research = json.loads((ROOT/'evidence'/'research.json').read_text())
out = ROOT/'evidence'/'figures'
out.mkdir(exist_ok=True)
plt.rcParams.update({'font.size': 10, 'axes.spines.top': False, 'axes.spines.right': False})
fig, axes = plt.subplots(1, 3, figsize=(15, 4.6))
for series, title, color in [('fixed', 'Fixed API budget: 1 CPU / 512 MiB', '#2563eb'),
('growing', 'Growing API budget', '#0f766e')]:
groups = [research['groups'][series+'-'+str(n)] for n in (1, 2, 4)]
medians = [g['median_rps'] for g in groups]
errors = [[m-g['min_rps'] for m,g in zip(medians,groups)], [g['max_rps']-m for m,g in zip(medians,groups)]]
axes[0].errorbar([1,2,4], medians, yerr=errors, label=title, color=color, marker='o', capsize=5)
axes[0].set(xlabel='API replicas', ylabel='Successful requests / second', title=f"Scale: median and min/max ({research['repetitions']} runs)", xticks=[1,2,4])
axes[0].legend(fontsize=8)
runs = sorted([r for r in research['runs'] if r['group']=='saturation'], key=lambda r:r['concurrency'])
axes[1].plot([r['concurrency'] for r in runs], [r['rps'] for r in runs], 'o-', color='#2563eb')
right = axes[1].twinx()
right.plot([r['concurrency'] for r in runs], [r['latency']['p95'] for r in runs], 's--', color='#b45309')
axes[1].set(xlabel='Concurrent workers (4 API replicas)', ylabel='Successful requests / second', title='Saturation / increasing queueing')
axes[1].set_xscale('log', base=2)
axes[1].set_xticks([r['concurrency'] for r in runs], labels=[str(r['concurrency']) for r in runs])
right.set_ylabel('p95 of all attempts, ms', color='#b45309')
groups = [research['groups']['cache-'+name] for name in ('off', 'on')]
vals = [g['median_rps'] for g in groups]
axes[2].bar(['Cache off','Cache on'], vals, color=['#64748b','#0f766e'], width=.6)
axes[2].errorbar([0,1],vals,yerr=[[v-g['min_rps'] for v,g in zip(vals,groups)], [g['max_rps']-v for v,g in zip(vals,groups)]],fmt='none',ecolor='#111827',capsize=5)
for i, val in enumerate(vals):
axes[2].text(i,groups[i]['max_rps']*1.04,f'{val:.1f}',ha='center')
axes[2].set(ylabel='Successful requests / second', title=f"Controlled A/B: +{(research['cache_improvement_ratio']-1)*100:.1f}% median RPS")
axes[2].set_ylim(0,max(vals)*1.25)
for ax in axes:
ax.grid(axis='y',alpha=.2)
fig.suptitle('Lab 3 / track 1: 80% history reads, 20% durable message writes', fontsize=13)
fig.tight_layout()
fig.savefig(out/'scaling.png',dpi=180)
fig.savefig(out/'scaling.svg')
plt.close(fig)
# Resource evidence for the saturation series; quota share, not host CPU %.
fig, ax = plt.subplots(figsize=(9,4.6))
for service,color in [('postgres','#b45309'),('lb','#64748b'),('api','#2563eb')]:
values=[]
for run in runs:
raw=json.loads((ROOT/'evidence'/'runs'/(run['label']+'.json')).read_text())
matching=[v['quota_utilization_ratio'] for v in raw['cgroup_deltas'].values() if '-'+service+'-' in v['name']]
values.append(statistics.mean(matching)*100)
ax.plot([r['concurrency'] for r in runs],values,'o-',label=service,color=color)
ax.set(xlabel='Concurrent workers',ylabel='CPU quota used, %',title='Shared dependency saturation (warmup + measured interval)')
ax.set_xscale('log', base=2)
ax.set_xticks([r['concurrency'] for r in runs], labels=[str(r['concurrency']) for r in runs])
ax.set_ylim(0, 105)
ax.axhline(100,color='#111827',linestyle='--',linewidth=.8)
ax.legend(); ax.grid(alpha=.2); fig.tight_layout()
fig.savefig(out/'resources.png',dpi=180); fig.savefig(out/'resources.svg')
plt.close(fig)
print('Saved scaling and resource figures (PNG + SVG).')
if __name__ == '__main__':
main()