Universal Scalability Law, USL
Updated 2026-08-10
INTRODUCTION
English translation pending.
CORE DEFINITION
Formulated by Neil Gunther, the Universal Scalability Law extends Amdahl's law by adding a second penalty beyond the serial fraction: the cost of maintaining coherence among nodes. Throughput rises with added capacity only until contention and coherence overhead offset the gain, after which it declines, a pattern called retrograde scaling. The core claim is that the coherence term grows with the square of the node count, so beyond a point more resources make the system slower. Key qualifications are that the model applies to systems with shared state and that its two coefficients must be fitted to measurements.
SCAFFOLDING EFFECT
Reduce cognitive load
- Expect the peak: find the node count where throughput stops rising before adding more. - Cut coherence: split shared state and reduce global synchronization between parts. - Prefer small units: divide work into teams or services that need little mutual coordination.
Anchor fast decisions
Two costs compound as nodes are added. Contention, the serial fraction, is the Amdahl penalty and sets a ceiling. Coherence is worse: the effort of keeping every node consistent grows roughly with the square of the node count, so its cost rises faster than the capacity being added. Once coherence growth outpaces the gain, total throughput falls, and the system is in retrograde scaling. Splitting shared state removes the quadratic term, which is why small independent units scale further.
MINIMUM ACTION
In progress 0/1Practice this model in one real situation:
account_treeGenealogyexpand_more
menu_bookReferencesexpand_more
Source support: Explicit
- en.wikipedia.orghttps://en.wikipedia.org/wiki/Neil_J._Guntherverified
PRIVATE NOTES · Only visible to you
SAVED Q&A
ENTRY Q&A · Private saving available
Ask with a clear boundary
thinkingmodels answers from published entry context only.
Your question is sent to thinkingmodels. The answer uses public entry context only.
RELATED MODELS