What does Fan-Out Tail Latency conclude, in numbers, about “The Multiplicative Law: How One Slow Call in a Hundred Becomes Sixty-Three Slow Pages in a Hundred”?
Answer: Chapter 1 recomputes the multiplicative law: one slow request in ten thousand across two thousand calls still gives slow pages with probability 0.181, and a page 99th percentile allows each call a slow share of only 0.000100.
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.
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: 0.000100, 99
| Per-call percentile needed for a page percentile | ||||
| N | page p95: call q | nines | page p99: call q | nines |
| 1 | 0.950000 | 1.30 | 0.990000 | 2.00 |
| 10 | 0.994884 | 2.29 | 0.998995 | 3.00 |
| 30 | 0.998292 | 2.77 | 0.999665 | 3.48 |
| 100 | 0.999487 | 3.29 | 0.999900 | 4.00 |
| 300 | 0.999829 | 3.77 | 0.999966 | 4.47 |
| 1000 | 0.999949 | 4.29 | 0.999990 | 5.00 |
| Page p99 at N=100 needs each call at pct 99.990 | ||||
| Allowed slow share per call: 0.000100 | ||||
| Rule of thumb (1 - P)/N: | 0.000100 |
Show the full program
# Chapter 1. p12: which per-call percentile a page percentile demands.
# Model: for N independent parallel calls, the page meets its deadline
# with probability q^N when each call does so with probability q. To
# give the page a target probability P, each call needs q = P^(1/N).
# The "nines" column is -log10(1 - q). Targets are illustrative.
import math
def per_call_q(page_target, n):
return math.exp(math.log(page_target) / n)
if __name__ == "__main__":
print("Per-call percentile needed for a page percentile")
print(" N page p95: call q nines page p99: call q nines")
for n in (1, 10, 30, 100, 300, 1000):
cells = []
for tgt in (0.95, 0.99):
q = per_call_q(tgt, n)
nines = -math.log10(1.0 - q)
cells.append("{:12.6f} {:5.2f}".format(q, nines))
print(" {:5d} {} {}".format(n, cells[0], cells[1]))
q = per_call_q(0.99, 100)
print("Page p99 at N=100 needs each call at pct {:.3f}".format(
100 * q))
print("Allowed slow share per call: {:.6f}".format(1.0 - q))
print("Rule of thumb (1 - P)/N: {:.6f}".format(0.01 / 100))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 1: The Multiplicative Law: How One Slow Call in a Hundred Becomes Sixty-Three Slow Pages in a Hundred. 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
- Chapter 2 · The Page Is a Maximum: Order Statistics and the Last Few Leaves
- Chapter 3 · When Independence Fails: Correlated Slowness, Bad Machines and Synchronised Background Work
- Chapter 4 · Hedged Requests: Sending a Second Copy at the 95th Percentile
- Chapter 5 · Queueing Tails: Utilisation Stretches the Tail, and Hedges Are Load Too
- Chapter 6 · Budget Allocation: Splitting a Page Percentile Across Every Call
- Chapter 7 · The Classics East and West: Half the Journey at Ninety, and the Law of the Minimum
- Chapter 8 · A Fan-Out Audit of a Paid Web Service, and a Plan for Refutation