forked from mtusi2026/mtusi2026lab3
118 lines
3.8 KiB
Python
118 lines
3.8 KiB
Python
import argparse
|
|
import csv
|
|
import json
|
|
import statistics
|
|
from pathlib import Path
|
|
|
|
import matplotlib
|
|
|
|
matplotlib.use("Agg")
|
|
import matplotlib.pyplot as plt # noqa: E402
|
|
|
|
|
|
def main():
|
|
parser = argparse.ArgumentParser()
|
|
parser.add_argument("directory")
|
|
args = parser.parse_args()
|
|
directory = Path(args.directory)
|
|
meta = json.loads((directory / "run.json").read_text())
|
|
assert meta["complete"]
|
|
rows = [
|
|
json.loads((directory / name).read_text()) for name in meta["cells"]
|
|
]
|
|
keys = [
|
|
"series",
|
|
"replicas",
|
|
"concurrency",
|
|
"repeat",
|
|
"pool_size",
|
|
"successful_rps",
|
|
"errors",
|
|
"error_ratio",
|
|
"p50_ms",
|
|
"p95_ms",
|
|
"p99_ms",
|
|
]
|
|
with (directory / "summary.csv").open("w", newline="") as stream:
|
|
writer = csv.DictWriter(stream, fieldnames=keys)
|
|
writer.writeheader()
|
|
writer.writerows({k: r[k] for k in keys} for r in rows)
|
|
fig, axes = plt.subplots(1, 2, figsize=(11, 4), constrained_layout=True)
|
|
for ax, series in zip(axes, ["fixed", "growing"], strict=True):
|
|
for concurrency in [1, 8, 32]:
|
|
values = [
|
|
statistics.median(
|
|
r["successful_rps"]
|
|
for r in rows
|
|
if r["series"] == series
|
|
and r["replicas"] == n
|
|
and r["concurrency"] == concurrency
|
|
)
|
|
for n in [1, 2, 4]
|
|
]
|
|
ax.plot([1, 2, 4], values, marker="o", label=f"C={concurrency}")
|
|
ax.set(
|
|
title=series,
|
|
xlabel="API replicas",
|
|
ylabel="Successful HTTP requests/s",
|
|
xticks=[1, 2, 4],
|
|
)
|
|
ax.legend()
|
|
ax.grid(alpha=0.3)
|
|
fig.savefig(directory / "scaling.png", dpi=160)
|
|
plt.close(fig)
|
|
pool = {
|
|
n: statistics.median(
|
|
r["successful_rps"]
|
|
for r in rows
|
|
if r["series"] == "pool" and r["pool_size"] == n
|
|
)
|
|
for n in [1, 10]
|
|
}
|
|
lines = [
|
|
"# Измерения масштабирования",
|
|
"",
|
|
f"Source commit: `{meta['source_commit']}`.",
|
|
"",
|
|
"Медианы повторов; latency percentile внутри каждого прогона относится только к успешным ответам.",
|
|
"",
|
|
"| Серия | API | C | RPS | p95, ms | Ошибки |",
|
|
"|---|---:|---:|---:|---:|---:|",
|
|
]
|
|
for series in ["fixed", "growing"]:
|
|
for n in [1, 2, 4]:
|
|
for c in [1, 8, 32]:
|
|
group = [
|
|
r
|
|
for r in rows
|
|
if r["series"] == series
|
|
and r["replicas"] == n
|
|
and r["concurrency"] == c
|
|
]
|
|
lines.append(
|
|
f"| {series} | {n} | {c} | {statistics.median(r['successful_rps'] for r in group):.1f} | {statistics.median(r['p95_ms'] for r in group):.2f} | {sum(r['errors'] for r in group)} |"
|
|
)
|
|
lines.extend(
|
|
[
|
|
"",
|
|
f"Пул1 → пул10: {pool[1]:.1f} → {pool[10]:.1f} успешных RPS (N2/C32, три повтора).",
|
|
"",
|
|
"Настройка считается подтверждённым улучшением только при наблюдаемом выигрыше; это ограниченная локальная серия.",
|
|
]
|
|
)
|
|
(directory / "analysis.md").write_text("\n".join(lines) + "\n")
|
|
print(
|
|
json.dumps(
|
|
{
|
|
"cells": len(rows),
|
|
"success": sum(r["success"] for r in rows),
|
|
"errors": sum(r["errors"] for r in rows),
|
|
"pool_rps": pool,
|
|
}
|
|
)
|
|
)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
main()
|