Part 2 — Modeling  ::  Chapter 10

2.6 Measurement Models: How Much, and Against What Bound?

Every system emits numbers, and most of them are just telemetry: observations with no engineering meaning attached. A measurement model gives selected observations meaning by relating them: usage to cost, cost to a budget, concurrency to a capacity. It answers the quantitative question the earlier families set aside — how much, and against what bound?

2.6.1 The Canonical Move: Quantities, Relations, and Bounds

Engineering has many quantitative representations: performance and parametric models, capacity models, cost models, and the budgets and envelopes they are compared against; queueing and resource models appear where contention matters11. Edward D. Lazowska et al., Quantitative System Performance: Computer System Analysis Using Queueing Network Models (Prentice-Hall, 1984).. The move they share is small but decisive. Telemetry is observation; a measurement model relates quantities so that a bound becomes statable. A node is a quantity; a relation is an accounting or functional link between quantities — usage times price is cost, concurrent jobs are capacity demand; a bound is the envelope a quantity is compared against (Figure 2.6-1).

Canonical measurement model: two chains, two bounds A measurement opener. A cost chain relates usage and pricing to an attributable cost, compared against a bracketed BUDGET envelope. A capacity chain relates concurrency to a capacity demand, compared against a bracketed CAPACITY ENVELOPE. The bracketed envelope is the reusable quantity-versus-bound glyph. Caption: Question, Semantics, Typical analysis. Neutral ground marks the measurement family. Measurement model — how much, against what bound? COST CHAIN usage + pricing attributable cost BUDGET compare CAPACITY CHAIN concurrency capacity demand CAPACITY ENVELOPE compare Question: How much, and against what bound? Semantics: relate quantities (usage × price = cost; concurrency = demand), then compare each against a bracketed bound. Typical analysis: aggregation, comparison against a bound, capacity planning.
Figure 2.6-1. The measurement move: quantities against a bound. Telemetry becomes a model when quantities are related and a value is read against a declared envelope; the analyses aggregate and compare rather than watch a stream call by call.

Once quantities are related to a bound, the claims a designer wants become expressible: upper and lower bounds, budgets, capacity constraints, rates, conservation relations, tolerances, and margins. The analyses follow — aggregation, comparison against a bound, sensitivity and capacity analysis, forecasting. Figure 2.6-2 draws the core relation: a current value read against a separately declared reference. The payoff is that "are we within budget?" becomes a comparison against a modeled bound rather than a guess over a dashboard.

A current value is compared against a separately declared reference bound Two peer inputs — a current value and a reference bound — converge into a comparison. The reference bound is declared separately; it is not derived from the measured quantity. The model says what is compared, not what to do about the result. CURRENT VALUE what is now REFERENCE BOUND declared separately COMPARISON The bound is declared apart from the quantity. The model compares; it does not decide the consequence.
Figure 2.6-2. Core measurement relation. A current value is interpreted against a separately declared reference bound. The model defines what is compared; it does not determine the consequence of the comparison.

2.6.2 DocAble: GenAI Cost and Capacity

DocAble's consequential quantities are the cost of its GenAI usage and the capacity its concurrent work demands. The model keeps two chains deliberately separate, because they answer different questions and are compared against different bounds.

MODEL CARD

GenAI cost and capacity · Measurement

Figure 2.6-3 draws the two chains. Note the deliberate wording of the property: the model says cost can be compared against a budget, not that cost must not exceed one. The cost chain is observational — a computed ratio, surfaced and reported, that never blocks work as it is spent. The capacity chain is a runtime-tunable cap, adjustable rather than proven. A cost figure is evidence; whether crossing a budget should block work, raise an alarm, or simply be recorded is a separate question of enforcement.

DocAble's GenAI cost-and-capacity measurement model: two separate envelopes Generation work splits into two independent quantity chains. One chain: usage relates to cost, which is compared against a cost budget. A separate chain: concurrency relates to capacity, which is compared against a capacity envelope. The two chains are drawn apart because they are distinct envelopes, not one shared budget. TWO ENVELOPES, NOT ONE generation work usage cost cost budget concurrency capacity capacity envelope Two chains, two envelopes — usage feeds a cost budget, concurrency feeds a capacity envelope; they are not one relation.
Figure 2.6-3. GenAI cost-and-capacity model. Generation work fans into usage and concurrency; usage and pricing determine cost against a budget, and concurrency contributes to capacity demand against a capacity envelope. The model names the quantities worth relating — the interpretation placed over raw requests and charges, not the telemetry itself.

DocAble does not attach a hard invariant to either quantity. Cost is routed to the administrative pane for observation and accounting, while capacity is controlled through a runtime-tunable limit. The model makes both quantities available for engineering use without requiring either to become an enforced obligation.

2.6.3 Tolerance and Margin

A bound is rarely a single line. A budget usually carries a margin; a capacity envelope leaves headroom. Those are the measurement family's version of the tolerance idea from the opening chapter: the model represents not just a target but the band of acceptable variation around it, so that "within tolerance" and "over the line" are both statable. The point is only that a measurement model can carry a margin — not to reopen the degrees-of-freedom discussion the opening chapter already settled.

The cost-and-capacity model makes spending attributable and demand comparable to an envelope. It does not decide the budget, guarantee the numbers are collected correctly, or enforce the bound. Figure 2.6-4 draws that separation: the model defines the quantity and the reference; whether a comparison is merely observed, used to adapt behavior, or enforced by a gate is Alignment's decision. This is the clearest case in the Part of a model that earns its place by exposing a property while leaving every question of enforcement open.

One measurement, different enforcement decisions Part II, the model, names a quantity and a reference bound and compares them, answering what quantity and bound matter. Part III, the enforcement decision, takes the same model and chooses its force: observe, adapt, degrade, or gate. Observe and adapt are shown in gray; degrade and gate, the consequential decisions, in rust. Modeling and enforcement are separate moves over one shared model. One Measurement, Different Enforcement Decisions PART II · MODEL quantity bound comparison “What quantity and bound matter?” PART III · ENFORCEMENT DECISION same model observe adapt degrade gate “What enforcement is justified?” The model defines the quantity and bound; Alignment decides whether to observe, adapt, degrade, or gate.
Figure 2.6-4. Modeling and enforcement are separate decisions. The measurement model defines the quantity and reference bound; Alignment determines whether the result is observed, used to adapt behavior, used to trigger graceful degradation, or enforced by a gate.

Measurement counts what is happening now. The last family asks what already happened, and what records it.

Works Cited

  1. Lazowska, Edward D., John Zahorjan, G. Scott Graham, and Kenneth C. Sevcik. Quantitative System Performance: Computer System Analysis Using Queueing Network Models. Prentice-Hall, 1984.
© James C. Davis, 2026–present