bb014f47d2
Some checks are pending
Sync Graph Data to teleo-app / sync (push) Waiting to run
Pentagon-Agent: Ganymede <F99EBFA6-547B-4096-BEEA-1D59C3E4028A>
36 lines
2.6 KiB
Markdown
36 lines
2.6 KiB
Markdown
---
|
||
type: claim
|
||
domain: internet-finance
|
||
description: "Bursty arrival processes require more safety capacity than Poisson models predict, scaled by variance-to-mean ratio"
|
||
confidence: proven
|
||
source: "Whitt et al., 'Staffing a Service System with Non-Poisson Non-Stationary Arrivals', Cambridge Core, 2016"
|
||
created: 2026-03-11
|
||
---
|
||
|
||
# Square-root staffing formula requires peakedness adjustment for non-Poisson arrivals because bursty processes need proportionally more safety capacity than the Poisson baseline predicts
|
||
|
||
The standard square-root staffing formula (workers = mean load + safety factor × √mean) assumes Poisson arrivals where variance equals mean. Real-world arrival processes violate this assumption through burstiness (arrivals clustered in time) or smoothness (arrivals more evenly distributed than random).
|
||
|
||
Whitt et al. extend the square-root staffing rule by introducing **peakedness** — the variance-to-mean ratio of the arrival process — as the key adjustment parameter. For bursty arrivals (peakedness > 1), systems require MORE safety capacity than Poisson models suggest. For smooth arrivals (peakedness < 1), systems need LESS.
|
||
|
||
The modified staffing formula adjusts the square-root safety margin by multiplying by the square root of peakedness. This correction is critical for non-stationary systems where arrival rates vary over time (daily cycles, seasonal patterns, or event-driven spikes).
|
||
|
||
## Evidence
|
||
|
||
- Whitt et al. (2016) prove that peakedness — the variance-to-mean ratio — captures the essential non-Poisson behavior for staffing calculations
|
||
- Standard Poisson assumption (variance = mean) fails empirically for bursty workloads like research paper dumps, product launches, or customer service spikes
|
||
- Using constant staffing (fixed MAX_WORKERS) regardless of queue state creates dual failure: over-provisioning during quiet periods (wasted compute) and under-provisioning during bursts (queue explosion)
|
||
|
||
## Relevance to Pipeline Architecture
|
||
|
||
Teleo's research pipeline exhibits textbook non-Poisson non-stationary arrivals: research dumps arrive in bursts of 15+ sources, futardio launches come in waves of 20+ proposals, while other days see minimal activity. The peakedness parameter quantifies exactly how much extra capacity is needed beyond naive square-root staffing.
|
||
|
||
This directly informs dynamic worker scaling: measure empirical peakedness from historical arrival data, adjust safety capacity accordingly, and scale workers based on current queue depth rather than using fixed limits.
|
||
|
||
---
|
||
|
||
Relevant Notes:
|
||
- domains/internet-finance/_map
|
||
|
||
Topics:
|
||
- core/mechanisms/_map
|