关于「实证闭环:收费站排盘结果缓存的只读容量规划」,《缓存命中率与车氏近似》给出的定量结论是什么?

结论:第八章给出收费站排盘结果缓存的只读规划流程、蒸馏小工具「缓存命中率估算器」与预注册检验方案,特征时间达 9.58 天的容量须标出漂移提示。书中以标注框区分已发表的证据、可运行的 Python 计算(22 段程序)、原典与作者的假说。

数据口径:本书用标准库程序推翻这条直觉:在齐普夫请求下,最近最少使用(LRU)缓存的命中率随容量增长得很慢,示意参数下容量放大一百二十八倍,命中率只放大 6.80 倍;而车氏近似只凭请求概率与容量,就能事先算出 LRU 命中率,与逐条模拟的最大差距为 0.0013。 书中参数为示意;全部计算由书中程序生成;本书给出只读的容量建议,不访问任何服务器、不写任何配置;「排盘」仅作一种计算结果的名称,不做任何命理推断;属于非个人化经营建议。

English version

对照数据表

下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库与结论吻合的打印值: 9.58

Chart cache sheet: N50000, a = 0.9, R = 30000/day
C hit misses/day CPU-s/day memory MB T_C days
12500.4174 17477 3495 14.6 0.06
25000.5000 15001 3000 29.3 0.14
50000.5928 12215 2443 58.6 0.33
100000.6983 9052 1810 117.2 0.81
200000.8189 5433 1087 234.4 2.25
400000.9557 1330 266 468.8 9.58
gain per doubling:
12502500: +0.0825 hit, 2475 fewer misses/day
25005000: +0.0929 hit, 2786 fewer misses/day
500010000: +0.1054 hit, 3163 fewer misses/day
1000020000: +0.1207 hit, 3620 fewer misses/day
2000040000: +0.1367 hit, 4102 fewer misses/day
查看完整验算程序
# Chapter 8. p81: a read-only capacity sheet for a chart-result cache.
# Model: the Che approximation turns four planning inputs into a table:
# distinct chart keys N, Zipf exponent a, requests per day R and the
# compute time per recomputed chart. Entry size converts slots into
# memory. Every input is illustrative (N = 50000, a = 0.9, R = 30000,
# 0.2 CPU-seconds per miss, 12 KB per entry); the sheet is advice to be
# read, not a setting written anywhere.
import math

N, A, R, CPU, KB = 50000, 0.9, 30000, 0.2, 12


def zipf(n, a):
    w = [1.0 / (i ** a) for i in range(1, n + 1)]
    s = math.fsum(w)
    return [x / s for x in w]


def occupancy(p, t):
    return math.fsum(-math.expm1(-x * t) for x in p)


def che_t(p, c, tol=1e-9):
    lo, hi = 0.0, 1.0
    while occupancy(p, hi) < c:
        hi = 2.0 * hi
    while hi - lo > tol * hi:
        mid = 0.5 * (lo + hi)
        if occupancy(p, mid) < c:
            lo = mid
        else:
            hi = mid
    return 0.5 * (lo + hi)


def che_hit(p, t):
    return math.fsum(x * -math.expm1(-x * t) for x in p)


if __name__ == "__main__":
    p = zipf(N, A)
    print("Chart cache sheet: N = {}, a = {}, R = {}/day".format(N, A, R))
    print("      C   hit     misses/day   CPU-s/day   memory MB   T_C days")
    rows = []
    for c in (1250, 2500, 5000, 10000, 20000, 40000):
        t = che_t(p, c)
        h = che_hit(p, t)
        rows.append((c, h))
        print("  {:5d}  {:.4f}    {:7.0f}      {:7.0f}     {:7.1f}    {:6.2f}"
              .format(c, h, R * (1 - h), R * (1 - h) * CPU,
                      c * KB / 1024, t / R))
    print("  gain per doubling:")
    for (c0, h0), (c1, h1) in zip(rows, rows[1:]):
        print("    {:5d} -> {:5d}: +{:.4f} hit, {:5.0f} fewer misses/day"
              .format(c0, c1, h1 - h0, R * (h1 - h0)))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《缓存命中率与车氏近似、排盘结果缓存容量的最优边界母典》第 8 章「实证闭环:收费站排盘结果缓存的只读容量规划」(版本 1.0.0). 5D Library. DOI 待发布

DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-327 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026327,
  author    = {Fan, Yafu},
  title     = {Cache Hit Ratios and the Che Approximation: Optimal Capacity for Caching Computed Charts},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论