#!/usr/bin/env python3 """Generate standalone figures and CSV/Markdown tables from raw research JSON.""" from collections import defaultdict import csv import json import os import re from pathlib import Path import statistics os.environ.setdefault('MPLCONFIGDIR', '/tmp/mtusi-lab3-matplotlib') import matplotlib matplotlib.use('Agg') import matplotlib.pyplot as plt ROOT = Path(__file__).resolve().parents[1] DIRECTORY = ROOT / 'evidence/scaling' def memory_mib(value): number, unit = re.fullmatch(r"([0-9.]+)([A-Za-z]+)", value.split(" / ")[0]).groups() return float(number) * {"B":1/1048576,"KiB":1/1024,"MiB":1,"GiB":1024,"kB":1000/1048576,"MB":1000000/1048576,"GB":1000000000/1048576}[unit] def main(): groups = defaultdict(list) for path in DIRECTORY.glob('*.json'): data = json.loads(path.read_text()) if 'case' in data: groups[data['case']['name']].append(data) rows = [] for name, values in sorted(groups.items()): rps = [d['benchmark']['successful_rps'] for d in values] row = {'case': name, 'repeats':len(values), 'replicas':values[0]['case']['replicas'], 'rps_mean':round(statistics.mean(rps),3), 'rps_min':min(rps), 'rps_max':max(rps), 'rps_stdev':round(statistics.stdev(rps),3) if len(rps)>1 else 0, 'p50_ms_mean':round(statistics.mean(d['benchmark']['successful_latency_ms']['p50'] for d in values),3), 'p95_ms_mean':round(statistics.mean(d['benchmark']['successful_latency_ms']['p95'] for d in values),3), 'p99_ms_mean':round(statistics.mean(d['benchmark']['successful_latency_ms']['p99'] for d in values),3), 'errors':sum(sum(d['benchmark']['errors'].values()) for d in values), 'db_cpu_percent_mean':round(statistics.mean(d['cpu_percent_mean']['mtusi-lab3-db-1'] for d in values),3), 'lb_cpu_percent_mean':round(statistics.mean(d['cpu_percent_mean']['mtusi-lab3-lb-1'] for d in values),3), 'client_cpu_percent_mean':round(statistics.mean(d['cpu_percent_mean'].get('mtusi-lab3-research-load',0) for d in values),3), 'api_memory_peak_mib':round(max(memory_mib(r['MemUsage']) for d in values for r in d['stats_samples'] if r['Name'].startswith('mtusi-lab3-api-')),3) if any(r['Name'].startswith('mtusi-lab3-api-') for d in values for r in d['stats_samples']) else None, 'db_memory_peak_mib':round(max(memory_mib(r['MemUsage']) for d in values for r in d['stats_samples'] if r['Name']=='mtusi-lab3-db-1'),3), 'cache_new_connections_mean':round(statistics.mean(d['dependency_after']['valkey']['total_connections_received']-d['dependency_before']['valkey']['total_connections_received'] for d in values),3)} rows.append(row) with (DIRECTORY/'summary.csv').open('w') as handle: writer = csv.DictWriter(handle,fieldnames=list(rows[0])) writer.writeheader() writer.writerows(rows) (DIRECTORY/'summary.json').write_text(json.dumps(rows,indent=2)+'\n') by_name={row['case']:row for row in rows} plt.rcParams.update({'font.family':'DejaVu Sans','font.size':10,'axes.grid':True,'grid.alpha':.25}) fig,axes=plt.subplots(1,2,figsize=(11,4),layout='constrained') for family,label in [('fixed','Fixed API: 1 CPU / 1024 MiB total'),('growing','Growing API: 0.5 CPU / 256 MiB each')]: data=[by_name[f'{family}-{n}'] for n in (1,2,4) if f'{family}-{n}' in by_name] if data: for ax,metric in zip(axes,('rps_mean','p95_ms_mean')): ax.plot([d['replicas'] for d in data],[d[metric] for d in data],marker='o',label=label) axes[0].errorbar([d['replicas'] for d in data],[d['rps_mean'] for d in data],yerr=[d['rps_stdev'] for d in data],fmt='none',capsize=4) for ax in axes: ax.set_xlabel('API replicas') ax.set_xticks([1,2,4]) axes[0].set_ylabel('Successful POST/s (mean ± sample SD)') axes[1].set_ylabel('Successful request p95, ms (mean across runs)') axes[0].legend(fontsize=8) fig.suptitle('Replica scaling, concurrency 16; PostgreSQL limited to 0.25 CPU') fig.savefig(DIRECTORY/'scaling.png',dpi=180) fig.savefig(DIRECTORY/'scaling.svg') plt.close(fig) if all(f'saturation-c{c}' in by_name for c in (1,4,16,64)): fig,axes=plt.subplots(1,2,figsize=(11,4),layout='constrained') data=[by_name[f'saturation-c{c}'] for c in (1,4,16,64)] axes[0].plot([1,4,16,64],[d['rps_mean'] for d in data],marker='o') axes[0].set(xscale='log',xlabel='Concurrency',ylabel='Successful POST/s',title='Saturation at fixed resources') ax2=axes[0].twinx() ax2.plot([1,4,16,64],[d['p95_ms_mean'] for d in data],marker='s',color='tab:red') ax2.set_ylabel('p95 ms',color='tab:red') names=['saturation-c16','db-0.5','db-1'] if all(n in by_name for n in names): axes[1].bar(['0.25 CPU','0.5 CPU','1 CPU'],[by_name[n]['rps_mean'] for n in names]) axes[1].set(xlabel='PostgreSQL CPU quota',ylabel='Successful POST/s',title='Bottleneck intervention (4 API, c=16)') fig.savefig(DIRECTORY/'saturation.png',dpi=180) fig.savefig(DIRECTORY/'saturation.svg') plt.close(fig) if all('cache-'+mode in by_name for mode in ('per_request','shared')): fig,axes=plt.subplots(1,2,figsize=(10,4),layout='constrained') names=['cache-per_request','cache-shared'] axes[0].bar(['Per-request client','Shared pool'],[by_name[n]['rps_mean'] for n in names], yerr=[by_name[n]['rps_stdev'] for n in names], capsize=5) axes[0].set_ylabel('Successful POST/s') axes[1].bar(['Per-request client','Shared pool'],[by_name[n]['cache_new_connections_mean'] for n in names]) axes[1].set_ylabel('New Valkey connections per full run') fig.suptitle('Connection reuse: 2 API × 0.5 CPU, PostgreSQL 1 CPU, c=16') fig.savefig(DIRECTORY/'cache-improvement.png',dpi=180) fig.savefig(DIRECTORY/'cache-improvement.svg') plt.close(fig) print(json.dumps(rows,indent=2)) if __name__=='__main__': main()