What does Fan-Out Tail Latency conclude, in numbers, about “The Classics East and West: Half the Journey at Ninety, and the Law of the Minimum”?

Answer: The result is an almost literal confirmation. In the light-tailed model with a hundred leaves, the ninetieth leaf returns on average at 8.78 milliseconds and the last at 17.56 milliseconds, a ratio of exactly 0.500: at ninety, only half the journey is done. The last ten leaves take 0.500 of the page's mean waiting time. When the first leaf returns only 0.116 of the waiting is done; at the fiftieth, only 0.231; at the ninety-fifth, 0.610; at the ninety-ninth, 0.829. Waiting time is not spread evenly across leaves; each further leaf is dearer than the one before.

Scope of the numbers: Service times, slow shares and costs are illustrative; all calculations are generated by the book's programs; the book is general business analysis, not individualised business advice. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses.

中文版

Data table

The numbers below are printed verbatim by the chapter's Python program and can be reproduced by running it. standard-library Python onlyprinted values that match the answer: 0.116, 0.231, 0.500, 0.610, 0.829, 17.56

k of 100 answeredE[T_(k)] (ms)share of E[max]
12.030.116
504.060.231
908.780.500
9510.710.610
9914.560.829
10017.561.000
Table 1, share of all-leaves time reached at 95% of leaves:
p50:12 /40 ms = 0.300
p95:32 /87 ms = 0.368
p99:70 / 140 ms = 0.500
last 10 leaves take 0.500 of the mean page time
pre-registered band for the 90-of-100 share: 0.45 to 0.55
Show the full program
# Chapter 7. p71: "half the journey at ninety", in leaf order statistics.
# Model: N = 100 leaves, each 2 ms + Exp(mean 3 ms) (illustrative). By
# Renyi's representation the k-th finishing time has the exact mean
#   E[T_(k)] = 2 + 3 (H_N - H_(N-k)),
# with harmonic sums by math.fsum. The share E[T_(k)] / E[T_(N)] says
# how much of the waiting is done when k of the N leaves have answered.
# Measured shares are from Dean and Barroso (2013, Table 1).
import math

N, A, M = 100, 2.0, 3.0


def harmonic(n):
    return math.fsum(1.0 / j for j in range(1, n + 1))


def mean_kth(k):
    return A + M * (harmonic(N) - harmonic(N - k))


if __name__ == "__main__":
    full = mean_kth(N)
    print("k of 100 answered   E[T_(k)] (ms)   share of E[max]")
    for k in (1, 50, 90, 95, 99, 100):
        print("      {:3d}             {:7.2f}          {:.3f}".format(
            k, mean_kth(k), mean_kth(k) / full))
    print("Table 1, share of all-leaves time reached at 95% of leaves:")
    for lab, a, b in (("p50", 12, 40), ("p95", 32, 87), ("p99", 70, 140)):
        print("  {}: {:3d} / {:3d} ms = {:.3f}".format(lab, a, b, a / b))
    print("last 10 leaves take {:.3f} of the mean page time".format(
        1.0 - mean_kth(90) / full))
    print("pre-registered band for the 90-of-100 share: 0.45 to 0.55")

Single source and citation

Fan, Y. (2026). The Multiplicative Law of Fan-Out Tail Latency and Hedged Requests: How Percentile Latency Amplifies When One Page Makes Many Calls (Version 1.0.0), Chapter 7: The Classics East and West: Half the Journey at Ninety, and the Law of the Minimum. 5D Library. DOI pending

DOI pending · monograph page goes live after publication · 5DL-B-326 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026326,
  author    = {Fan, Yafu},
  title     = {The Multiplicative Law of Fan-Out Tail Latency and Hedged Requests: How Percentile Latency Amplifies When One Page Makes Many Calls},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

Other chapter answers in this book