"""Validasi model analitis dengan simulasi kejadian diskrit (DES) M/M/c FCFS. Menjalankan R replikasi ber-seed untuk beberapa jumlah tenant N, mengukur latensi rata-rata dan p95 waktu tinggal, lalu membandingkannya dengan model_saas.metrics(). Hasil: data/validasi_des.csv. Hanya pustaka standar. """ import csv, heapq, math, os, random, statistics, sys sys.path.insert(0, os.path.dirname(__file__)) import model_saas as model # noqa: E402 LAM, S_MS, C = 2.0, 40.0, 8 NS = [30, 45, 60, 75, 90] REPLIKASI = 10 PERMINTAAN = 120_000 PEMANASAN = 10_000 SEED = 20261115 def jalankan(N, seed): rng = random.Random(seed) lam_total = N * LAM mu = 1000.0 / S_MS bebas = [0.0] * C # waktu worker kembali bebas (heap) heapq.heapify(bebas) t = 0.0 tinggal = [] for i in range(PERMINTAAN): t += rng.expovariate(lam_total) mulai = max(t, heapq.heappop(bebas)) selesai = mulai + rng.expovariate(mu) heapq.heappush(bebas, selesai) if i >= PEMANASAN: tinggal.append((selesai - t) * 1000.0) tinggal.sort() return statistics.fmean(tinggal), tinggal[int(0.95 * len(tinggal))] def main(): keluar = os.path.join(os.path.dirname(__file__), "..", "data", "validasi_des.csv") with open(keluar, "w", newline="") as f: w = csv.writer(f) w.writerow(["N", "utilisasi", "rerata_model_ms", "rerata_des_ms", "rerata_des_ik95", "p95_model_ms", "p95_des_ms", "p95_des_ik95", "galat_p95_persen"]) for N in NS: hasil = [jalankan(N, SEED + 1000 * N + r) for r in range(REPLIKASI)] rm = [h[0] for h in hasil] rp = [h[1] for h in hasil] m = model.metrics(N, LAM, S_MS, C) ik = lambda xs: 2.262 * statistics.stdev(xs) / math.sqrt(len(xs)) # t(0,975; 9) p95d = statistics.fmean(rp) w.writerow([N, round(m["rho"], 4), round(m["mean_ms"], 3), round(statistics.fmean(rm), 3), round(ik(rm), 3), round(m["p95_ms"], 3), round(p95d, 3), round(ik(rp), 3), round(abs(p95d - m["p95_ms"]) / m["p95_ms"] * 100, 3)]) print(N, round(m["p95_ms"], 2), round(p95d, 2), "±", round(ik(rp), 2)) if __name__ == "__main__": main()