forked from mtusi2026/mtusi2026lab4
60 lines
3.0 KiB
Python
60 lines
3.0 KiB
Python
#!/usr/bin/env python3
|
|
"""Rebuild CSV summaries and standalone plots from sanitized research artifacts."""
|
|
import csv
|
|
import json
|
|
import os
|
|
from pathlib import Path
|
|
import statistics
|
|
|
|
os.environ.setdefault('MPLCONFIGDIR','/tmp/mtusi-lab4-matplotlib')
|
|
import matplotlib
|
|
matplotlib.use('Agg')
|
|
import matplotlib.pyplot as plt
|
|
|
|
ROOT=Path(__file__).resolve().parents[1]
|
|
EVIDENCE=ROOT/'evidence'
|
|
|
|
|
|
def main():
|
|
research=json.loads((EVIDENCE/'reconciliation.json').read_text())
|
|
rows=[]
|
|
for n in sorted({r['extra_orphans'] for r in research['scan_cost_runs']}):
|
|
group=[r for r in research['scan_cost_runs'] if r['extra_orphans']==n]
|
|
times=[1000*r['elapsed_seconds'] for r in group]
|
|
rows.append({'extra_orphans':n,'objects_scanned':group[0]['objects_scanned'],
|
|
'metadata_scanned':group[0]['metadata_scanned'],'repeats':len(group),
|
|
'mean_ms':round(statistics.mean(times),3),'min_ms':round(min(times),3),
|
|
'max_ms':round(max(times),3),'stdev_ms':round(statistics.stdev(times),3),
|
|
'list_calls':group[0]['s3_list_calls'],'head_calls':group[0]['s3_head_calls']})
|
|
with (EVIDENCE/'scan-cost.csv').open('w') as stream:
|
|
writer=csv.DictWriter(stream,fieldnames=list(rows[0]));writer.writeheader();writer.writerows(rows)
|
|
(EVIDENCE/'scan-cost-summary.json').write_text(json.dumps(rows,indent=2)+'\n')
|
|
plt.rcParams.update({'font.size':10,'axes.grid':True,'grid.alpha':.25})
|
|
fig,ax=plt.subplots(figsize=(7,4),layout='constrained')
|
|
ax.errorbar([r['objects_scanned'] for r in rows],[r['mean_ms'] for r in rows],
|
|
yerr=[r['stdev_ms'] for r in rows],marker='o',capsize=5)
|
|
ax.set(xlabel='S3 objects scanned',ylabel='Dry-run time, ms (mean ± sample SD)',
|
|
title='Reconciler scan cost: fixed metadata baseline, three repeats')
|
|
for suffix in ('png','svg'):
|
|
fig.savefig(EVIDENCE/('scan-cost.'+suffix),dpi=180)
|
|
plt.close(fig)
|
|
memory=json.loads((EVIDENCE/'upload-memory.json').read_text())
|
|
phases=memory['workload']['phases']
|
|
fig,axes=plt.subplots(1,2,figsize=(10,4),layout='constrained')
|
|
labels=[f"{r['size_bytes']/1048576:.2f} MiB" for r in phases]
|
|
for api in range(len(phases[0]['api_memory'])):
|
|
x=[i+(api-.5)*.3 for i in range(len(phases))]
|
|
axes[0].bar(x,[r['api_memory'][api]['sampled_peak_mib'] for r in phases],width=.3,label='API '+str(api+1))
|
|
axes[0].axhline(512,ls='--',color='tab:red',label='Per-API memory limit')
|
|
axes[0].set(xticks=range(len(labels)),xticklabels=labels,ylabel='Sampled cgroup memory peak, MiB',title='10 concurrent uploads, two rounds')
|
|
axes[0].legend(fontsize=8)
|
|
axes[1].bar(labels,[r['latency_ms']['p95'] for r in phases])
|
|
axes[1].set(ylabel='Successful upload p95, ms',title='20 successful uploads per size')
|
|
for suffix in ('png','svg'):
|
|
fig.savefig(EVIDENCE/('upload-memory.'+suffix),dpi=180)
|
|
plt.close(fig)
|
|
print(json.dumps({'scan_cost':rows,'upload_phases':[{k:v for k,v in p.items() if k!='individual_requests'} for p in phases]},indent=2))
|
|
|
|
if __name__=='__main__':
|
|
main()
|