import csv import statistics as st from pathlib import Path import matplotlib matplotlib.use("Agg") import matplotlib.pyplot as plt ROOT = Path("evidence/track1") OUT = ROOT / "plots" CONFIGS = ["n1_cpu2.0", "n2_cpu1.0", "n4_cpu0.5"] NAMES = {"n1_cpu2.0": "1 реплика", "n2_cpu1.0": "2 реплики", "n4_cpu0.5": "4 реплики"} QUOTA = {"postgres": 100, "redis": 25, "load_balancer": 100, "api": 50} SVC_NAMES = {"postgres": "PostgreSQL", "redis": "Redis", "load_balancer": "Nginx", "api": "API (реплика)"} def load(d): """(label, rate) -> {"target": int, "rps": [...], "p95": [...]}""" data = {} with open(ROOT / d / "bench.csv", newline="") as f: for r in csv.DictReader(f): e = data.setdefault((r["label"], int(r["rate"])), {"target": int(r["target"]), "rps": [], "p95": []}) e["rps"].append(float(r["rps"])) e["p95"].append(float(r["p95_ms"])) return data def scaling(data): """RPS и p95 от нагрузки, линия на конфигурацию.""" fig, axes = plt.subplots(1, 2, figsize=(11, 4.2)) for ax, field, ylabel in zip(axes, ("rps", "p95"), ("успешные RPS", "p95, мс")): for lb in CONFIGS: pts = sorted((e["target"], e[field]) for (l, _), e in data.items() if l == lb) x = [t for t, _ in pts] ax.plot(x, [st.median(v) for _, v in pts], marker="o", label=NAMES[lb]) ax.fill_between(x, [min(v) for _, v in pts], [max(v) for _, v in pts], alpha=0.15) ax.set(xlabel="целевая нагрузка, req/s", ylabel=f"{ylabel} (медиана, min–max)") ax.grid(alpha=0.3) axes[0].legend() fig.suptitle("1/2/4 реплики при бюджете API 2 CPU (Nginx 0,5 CPU)") fig.tight_layout() fig.savefig(OUT / "scaling.png", dpi=130) plt.close(fig) def before_after(before, after, rate=80): """Nginx 0,5 CPU против 1,0 CPU на ступени rate.""" fig, axes = plt.subplots(1, 2, figsize=(11, 4.2)) w = 0.38 for ax, field, ylabel in zip(axes, ("rps", "p95"), ("успешные RPS", "p95, мс")): for i, (data, title) in enumerate(((before, "Nginx 0,5 CPU"), (after, "Nginx 1,0 CPU"))): vals = [data[(lb, rate)][field] for lb in CONFIGS] med = [st.median(v) for v in vals] err = [[m - min(v) for m, v in zip(med, vals)], [max(v) - m for m, v in zip(med, vals)]] ax.bar([j + (i - 0.5) * w for j in range(len(CONFIGS))], med, w, yerr=err, capsize=3, label=title) ax.set_xticks(range(len(CONFIGS))) ax.set_xticklabels([NAMES[c] for c in CONFIGS]) ax.set_ylabel(f"{ylabel} (медиана, min–max)") ax.grid(axis="y", alpha=0.3) axes[0].legend() fig.suptitle(f"Улучшение: квота Nginx 0,5 → 1,0 CPU (ступень {rate} req/s на воркера)") fig.tight_layout() fig.savefig(OUT / "lb_before_after.png", dpi=130) plt.close(fig) def cpu_at_limit(d="generator_check", label="n4_cpu0.5", rate=80): """CPU контейнеров на пределе 4 реплик: медиана замеров docker stats.""" samples = {s: [] for s in QUOTA} for f in sorted((ROOT / d / "stats").glob(f"{label}_r{rate}_run*.csv")): with open(f, newline="") as fh: for row in csv.reader(fh): if len(row) < 3: continue for svc in QUOTA: if svc in row[1]: samples[svc].append(float(row[2].rstrip("%"))) break svcs = [s for s in QUOTA if samples[s]] med = [st.median(samples[s]) for s in svcs] names = [SVC_NAMES[s] for s in svcs] fig, ax = plt.subplots(figsize=(7.5, 4.2)) bars = ax.bar(names, med, 0.5) ax.bar_label(bars, fmt="%.0f%%") ax.plot(names, [QUOTA[s] for s in svcs], "k_", markersize=40, mew=2, label="квота CPU") ax.set_ylabel("CPU, % одного ядра (медиана замеров)") ax.set_title(f"CPU контейнеров на пределе: {NAMES[label]}, 40 воркеров, ступень {rate}") ax.legend() ax.grid(axis="y", alpha=0.3) fig.tight_layout() fig.savefig(OUT / "cpu_at_limit.png", dpi=130) plt.close(fig) def main(): OUT.mkdir(parents=True, exist_ok=True) before, after = load("before"), load("after") scaling(before) before_after(before, after) cpu_at_limit() print(f"Графики записаны в {OUT}") if __name__ == "__main__": main()