"""Reproduksi hasil utama artikel di Lab Reproduksi Peervie (Pyodide/WebAssembly). Berkas ini berdiri sendiri (tanpa impor modul lain) karena Lab Reproduksi hanya memasang berkas Dataset di /data. Rumus model identik dengan kode/model_saas.py. Yang direproduksi: Tabel 2 validasi model terhadap simulasi kejadian diskrit (seed tetap), Tabel 3 ukuran kinerja per jumlah tenant (dicocokkan dengan sapuan_tenant.csv), Tabel 4 perbandingan biaya (perbandingan_biaya.csv), Tabel 5 ambang konsolidasi (sensitivitas.csv), Gambar 1 latensi p95 terhadap jumlah tenant. Keluaran teks bersifat deterministik sehingga sidik SHA-256-nya dapat dibandingkan. """ import csv, heapq, math, os, random, statistics JAM_PER_BULAN, SLO_MS = 730.0, 200.0 LAM, S_MS, C, HARGA = 2.0, 40.0, 8, 6000.0 # ── Model M/M/c (sama dengan model_saas.py) ───────────────────────────── def erlang_c(c, a): if a >= c: return 1.0 b = 1.0 for k in range(1, c + 1): b = a * b / (k + a * b) rho = a / c return b / (1.0 - rho + rho * b) def sojourn_tail(t, c, lam, mu): pw = erlang_c(c, lam / mu) theta = c * mu - lam if theta <= 0: return 1.0 if abs(theta - mu) < 1e-9: conv = math.exp(-mu * t) * (1.0 + mu * t) else: conv = (theta * math.exp(-mu * t) - mu * math.exp(-theta * t)) / (theta - mu) return (1.0 - pw) * math.exp(-mu * t) + pw * conv def quantile(q, c, lam, mu): lo, hi = 0.0, 1.0 while sojourn_tail(hi, c, lam, mu) > 1.0 - q: hi *= 2.0 for _ in range(80): mid = (lo + hi) / 2.0 if sojourn_tail(mid, c, lam, mu) > 1.0 - q: lo = mid else: hi = mid return (lo + hi) / 2.0 def metrics(N, lam, s_ms, c): mu = 1000.0 / s_ms L = N * lam rho = L / (c * mu) if rho >= 1.0: return None pw = erlang_c(c, L / mu) return {"rho": rho, "pw": pw, "mean": (pw / (c * mu - L) + 1.0 / mu) * 1000.0, "p95": quantile(0.95, c, L, mu) * 1000.0, "p99": quantile(0.99, c, L, mu) * 1000.0} def n_max(lam, s_ms, c): m1 = metrics(1, lam, s_ms, c) if m1 is None or m1["p95"] > SLO_MS: return 0 lo, hi = 1, 5000 while lo < hi: m = (lo + hi + 1) // 2 r = metrics(m, lam, s_ms, c) if r is not None and r["p95"] <= SLO_MS: lo = m else: hi = m - 1 return lo # ── Simulasi kejadian diskrit (sama dengan validasi_des.py) ───────────── def des(N, seed, permintaan=120_000, pemanasan=10_000): rng = random.Random(seed) lam_total, mu = N * LAM, 1000.0 / S_MS bebas = [0.0] * C t, tinggal = 0.0, [] 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))] # ── Pembanding dengan berkas data artikel ─────────────────────────────── hasil_cek = [] def baca(nama): for d in (".", "/data"): p = os.path.join(d, nama) if os.path.exists(p): with open(p, newline="") as f: return list(csv.DictReader(f)) return None def cek(label, hitung, tercatat, toleransi): ok = abs(hitung - tercatat) <= toleransi hasil_cek.append(ok) return "cocok" if ok else "BEDA (tercatat %.3f)" % tercatat def main(): print("Reproduksi: Ambang Konsolidasi Tenant pada Arsitektur SaaS Multi-Tenant") print("Parameter dasar: lambda=%.1f req/dtk per tenant, S=%.0f ms, c=%d worker, harga Rp%.0f/jam, SLO p95 %.0f ms" % (LAM, S_MS, C, HARGA, SLO_MS)) print("\n[Tabel 2] Validasi model vs simulasi kejadian diskrit (10 replikasi, seed 20261115)") data = {r["N"]: r for r in (baca("validasi_des.csv") or [])} for N in (30, 45, 60, 75, 90): rep = [des(N, 20261115 + 1000 * N + r) for r in range(10)] p95d = statistics.fmean(x[1] for x in rep) m = metrics(N, LAM, S_MS, C) galat = abs(p95d - m["p95"]) / m["p95"] * 100 status = cek("t2", p95d, float(data[str(N)]["p95_des_ms"]), 0.05) if str(N) in data else "data tidak ada" print("N=%3d p95 model %7.2f ms p95 DES %7.2f ms galat %.2f%% %s" % (N, m["p95"], p95d, galat, status)) print("\n[Tabel 3] Ukuran kinerja per jumlah tenant") data = {r["N"]: r for r in (baca("sapuan_tenant.csv") or [])} for N in (30, 60, 70, 80, 85, 90, 95): m = metrics(N, LAM, S_MS, C) status = cek("t3", m["p95"], float(data[str(N)]["p95_ms"]), 0.01) if str(N) in data else "data tidak ada" print("N=%3d utilisasi %5.1f%% tunggu %5.1f%% rerata %6.1f ms p95 %6.1f ms p99 %6.1f ms SLO %-5s %s" % (N, m["rho"] * 100, m["pw"] * 100, m["mean"], m["p95"], m["p99"], "ya" if m["p95"] <= SLO_MS else "tidak", status)) print("\n[Tabel 4] Perbandingan biaya") nm = n_max(LAM, S_MS, C) iso = metrics(1, LAM, S_MS, 1) biaya_iso = HARGA / C * JAM_PER_BULAN biaya_kon = HARGA * JAM_PER_BULAN / nm data = {r["skema"]: r for r in (baca("perbandingan_biaya.csv") or [])} s1 = cek("t4", biaya_iso, float(data["isolasi_penuh"]["biaya_tenant_rp_bulan"]), 1) if data else "data tidak ada" s2 = cek("t4", biaya_kon, float(data["konsolidasi_n_maks"]["biaya_tenant_rp_bulan"]), 1) if data else "data tidak ada" print("Isolasi penuh : p95 %.1f ms, biaya Rp%.0f/tenant/bulan %s" % (iso["p95"], biaya_iso, s1)) print("Konsolidasi N=%d : p95 %.1f ms, biaya Rp%.0f/tenant/bulan %s" % (nm, metrics(nm, LAM, S_MS, C)["p95"], biaya_kon, s2)) print("Penghematan : %.1f%%" % ((1 - biaya_kon / biaya_iso) * 100)) print("\n[Tabel 5] Ambang konsolidasi (N maks) per worker dan waktu layanan") data = {(r["worker_c"], r["layanan_ms"]): r for r in (baca("sensitivitas.csv") or [])} for c in (4, 8, 16, 32): baris = [] for s in (20.0, 40.0, 80.0): nmx = n_max(LAM, s, c) key = (str(c), str(s)) if key in data: cek("t5", nmx, float(data[key]["n_maks"]), 0) baris.append("S=%2.0f ms: %s" % (s, nmx if nmx else "SLO tidak tercapai")) print("c=%2d " % c + " | ".join(baris)) total, cocok = len(hasil_cek), sum(hasil_cek) print("\nRingkasan: %d dari %d angka cocok dengan berkas data artikel." % (cocok, total)) if total == 0: print("Berkas data tidak ditemukan; hanya hasil perhitungan yang ditampilkan.") try: # Gambar 1 (opsional, butuh matplotlib) import matplotlib.pyplot as plt xs = list(range(10, 96)) fig, ax = plt.subplots(figsize=(6.4, 3.6)) ax.plot(xs, [metrics(n, LAM, S_MS, C)["p95"] for n in xs], color="#0f766e", lw=2, label="p95 model") ax.plot(xs, [metrics(n, LAM, S_MS, C)["mean"] for n in xs], color="#9ca3af", ls="--", label="Rerata model") ax.axhline(SLO_MS, color="#dc2626", ls=":", label="SLO 200 ms") ax.set_xlabel("Jumlah tenant per node (N)") ax.set_ylabel("Waktu tanggap (ms)") ax.set_ylim(0, 340) ax.legend(frameon=False) ax.set_title("Gambar 1 (reproduksi)") fig.tight_layout() except ImportError: pass main()