What does Fan-Out Tail Latency conclude, in numbers, about “A Fan-Out Audit of a Paid Web Service, and a Plan for Refutation”?

Answer: Chapter 8 gives an offline fan-out audit for the pages of a paid web service, the distilled Fan-Out Tail Latency Calculator and a pre-registered test. Marked boxes separate published evidence, runnable Python calculations (22 programs), classical sources and the author's hypotheses.

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 only

pagecallsP(miss)top class (share)hedged
landing80.0070profile(0.572)0.0030
chart370.0621store(0.563)0.0277
synthesis990.1833store(0.445)0.1065
report410.1221store(0.923)0.0104
rule: a page is flagged when P(miss) > 0.05
Show the full program
# Chapter 8. p81: an offline fan-out audit sheet for a paid web service.
# Inputs are illustrative inventories, not measurements of any server:
# for each page, call classes with a count n and the share p of calls
# that miss the page deadline. Under independence the page miss share
# is 1 - prod (1 - p)^n; each class's share of the "miss budget" is
# n(-ln(1 - p)) over the total. "hedged" replaces the top class's p by
# p^2 (independent replicas, hedge well before the deadline).
import math

PAGES = (
    ("landing", (("static", 6, 0.0005), ("profile", 2, 0.002))),
    ("chart", (("engine", 1, 0.004), ("cache", 24, 0.001),
               ("store", 12, 0.003))),
    ("synthesis", (("engine", 8, 0.004), ("cache", 60, 0.001),
                   ("store", 30, 0.003), ("partner", 1, 0.02))),
    ("report", (("store", 40, 0.003), ("render", 1, 0.01))),
)


def hazard(n, p):
    return -n * math.log1p(-p)


if __name__ == "__main__":
    print("page        calls  P(miss)   top class (share)    hedged")
    for name, classes in PAGES:
        hs = [hazard(n, p) for _, n, p in classes]
        total = math.fsum(hs)
        calls = math.fsum(n for _, n, _ in classes)
        top = max(range(len(classes)), key=lambda i: hs[i])
        tn, tp = classes[top][1], classes[top][2]
        rest = total - hs[top]
        hedged = -math.expm1(-(rest + hazard(tn, tp * tp)))
        print("  {:<9} {:5d}   {:.4f}   {:<8} ({:.3f})    {:.4f}".format(
            name, int(calls), -math.expm1(-total), classes[top][0],
            hs[top] / total, hedged))
    print("rule: a page is flagged when P(miss) > 0.05")

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 8: A Fan-Out Audit of a Paid Web Service, and a Plan for Refutation. 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