What does One Commander and the Serial Bottleneck conclude, in numbers, about “Does the Problem Grow? Gustafson's Law, Per-Item Review and the Measured Serial Share”?
Answer: Gustafson's law says that when the problem grows with the number of processors, the share of the serial part is diluted and the speedup can approach proportional growth; program p21 shows that on 1,024 processors a serial share of 0.004 on the parallel machine corresponds to a scaled speedup of 1019.9. Program p22 shows that when every item takes 12 minutes of the commander's time, a growing problem does not help, and the speedup stops at 5.000 from four agents onwards. Program p23 shows that the Karp and Flatt metric separates a fixed serial share from overhead that grows with scale. Yet \(\min(N/a,1/r)\) is only an optimistic bound: items from agents do not arrive evenly, and the commander's review times are sometimes short and sometimes long. How long a queue does randomness build at the commander's door? That is the question of the next chapter.
Scope of the numbers: With illustrative parameters, cutting the serial share from a fifth to a twentieth raises the speedup by 3.325, while raising the agents from ten to a hundred adds only 1.236. Its parameters are illustrative; all calculations are generated by its programs.
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: 0.004, 1019.9, 1024
| N | 1,024 processors |
| s' S_G Amdahl s S_Amdahl | |
| 0.004 | 1019.9 0.00000392 1019.9 |
| 0.006 | 1017.9 0.00000589 1017.9 |
| 0.008 | 1015.8 0.00000788 1015.8 |
| reported speedups (Gustafson 1988): | |
| beam stress | 1,021 implied s' = 0.0029 |
| surface wave | 1,020 implied s' = 0.0039 |
| fluid flow | 1,016 implied s' = 0.0078 |
| the fixed-size ceiling for s' read as Amdahl's s: | |
| s | 0.004 S(1,024) = 201.1 1/s = 250.0 |
| s | 0.008 S(1,024) = 111.5 1/s = 125.0 |
Show the full program
# Chapter 2. p21: Gustafson's scaled speedup on 1024 processors.
# Model: Gustafson (1988) measures the serial share s' on the parallel
# machine itself and lets the problem grow with N, so the scaled
# speedup is S_G = s' + (1 - s') N. He reported s' of 0.4 to 0.8
# percent and measured speedups of 1021, 1020 and 1016 on N = 1024.
# The same run seen through Amdahl's fixed-size lens has the much
# smaller share s = s' / (s' + (1 - s') N), giving the same number.
# Source values from Gustafson (1988); the rest is arithmetic.
from fractions import Fraction as F
N = 1024
REPORTED = [("beam stress", 1021), ("surface wave", 1020),
("fluid flow", 1016)]
print("N = {:,} processors".format(N))
print("%8s %10s %14s %10s" % ("s'", "S_G", "Amdahl s", "S_Amdahl"))
for sp in [F(4, 1000), F(6, 1000), F(8, 1000)]:
sg = sp + (1 - sp) * N
s = sp / (sp + (1 - sp) * N)
sa = 1 / (s + (1 - s) / N)
print("%8.3f %10.1f %14.8f %10.1f" % (float(sp), float(sg),
float(s), float(sa)))
print("reported speedups (Gustafson 1988):")
for name, v in REPORTED:
sp = F(N - v, N - 1)
print(" %-12s %6s implied s' = %.4f" % (
name, "{:,}".format(v), float(sp)))
print("the fixed-size ceiling for s' read as Amdahl's s:")
for sp in [F(4, 1000), F(8, 1000)]:
sa = 1 / (sp + (1 - sp) / N)
print(" s = %.3f S(1,024) = %6.1f 1/s = %6.1f" % (
float(sp), float(sa), float(1 / sp)))Single source and citation
Fan, Y. (2026). Constitutional Engineering for a One-Commander Silicon Empire: Serial Decision Bandwidth and Breaking Amdahl's Bottleneck (Version 1.0.0), Chapter 2: Does the Problem Grow? Gustafson's Law, Per-Item Review and the Measured Serial Share. 5D Library. DOI pending
DOI pending · monograph page goes live after publication · 5DL-B-253 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026253,
author = {Fan, Yafu},
title = {Constitutional Engineering for a One-Commander Silicon Empire: Serial Decision Bandwidth and Breaking Amdahl's Bottleneck},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}Other chapter answers in this book
- Chapter 1 · The Ceiling of the Serial Share: Amdahl's Law and the Exact Account of Twenty Percent against Five
- Chapter 3 · The Commander Is a Single Server: Queueing Theory, the Finite-Source Model and Little's Law
- Chapter 4 · Batch Approval Windows: Harris's Square-Root Formula and the Price of Waiting
- Chapter 5 · Send Up Only the Exceptions: Automatic Gates, Random Audits and Coordination Overhead
- Chapter 6 · Decision Bandwidth: a Written Constitution as a Cache, a Faster Serial Part and the Span of Control
- Chapter 7 · Eastern and Western Sources Examined: Han Fei's Tally of Name and Form and Smith's Pin Factory
- Chapter 8 · Closing the Loop: a Weekly Decision Log, Batch Approval Windows and a Pre-Registered Test