Tensions in Practice: Early first completion in tension with steady output flow¶
Staged work · three ready items
Three ready items each need two ordered tasks, each lasting one invented time unit. One worker can perform both tasks without a handoff. Alternatively, two stage workers can overlap different items, but each item then pays a one-unit transfer delay between tasks. The first item finishes later in the pipeline, while subsequent items emerge more closely spaced.
Finish one item promptly
Avoid adding a transfer delay to an item’s own path.
Keep successive outputs close
Overlap different items without violating either item’s task order.
Why these aims pull against each other
Parallel stages shorten the interval between completed items, not necessarily the elapsed time from an item’s start to its finish. The extra stage capacity and handoff protocol have costs.
Choose an arrangement to see what changes and what remains difficult.
Finite illustrative comparisons. Text states carry the meaning; color is not a measured score or universal preference.
What this choice protects
What it costs
When it fits
Compare the arrangements
Complete one at a time
One worker finishes both tasks on A, then B, then C. All three are ready at time zero.
| Task 1 | Task 2 | Done at | |
|---|---|---|---|
| Item A | 0–1 | 1–2 | 2 |
| Item B | 2–3 | 3–4 | 4 |
| Item C | 4–5 | 5–6 | 6 |
- What it protects
- A finishes at time 2 with no inter-stage transfer or second worker.
- What it costs
- B and C wait; completion times are 2, 4 and 6.
- When it fits
- Fits a small or sparse workload when first completion and minimal staffing matter more than sustained output spacing.
Illustration note: The finite setting and values are editorial assumptions, not measured effects or recommended operating settings.
Overlap the stages
One worker runs task 1; another runs task 2. Transfers impose a one-unit delay and can overlap processing without using either worker.
| Task 1 | Task 2 | Done at | |
|---|---|---|---|
| Item A | 0–1 | 2–3 | 3Later first |
| Item B | 1–2 | 3–4 | 4 |
| Item C | 2–3 | 4–5 | 5Earlier last |
- What it protects
- After filling, outputs are one unit apart: 3, 4 and 5 for the displayed items.
- What it costs
- Each started item takes three units instead of two; two workers and in-flight handoff state are required.
- When it fits
- Fits sustained ready work when separate stage resources and correctly tracked transfers justify the narrower output spacing.
Illustration note: The finite setting and values are editorial assumptions, not measured effects or recommended operating settings. Each stage handles at most one item at a time; the displayed intervals satisfy that constraint.
What this illustration does—and does not—establish
Pipeline: Latency vs Throughput supplies the latency/throughput distinction. The finite schedule adds explicit transfer and staffing assumptions so both effects can be checked.
- Intervals are a declared deterministic schedule, not measured throughput or a queueing prediction.
- The alternatives differ in provisioned workers and transfer overhead; no free capacity gain is claimed.
- Every item still completes task 1 before task 2. Buffer bounds, errors, variable task lengths and return transport are outside the toy.
Source entries
Pipeline
Pipeline: Latency vs Throughput supplies the conflict examined here.
Latency vs Throughput
Longer pipelines (more fine-grained stages) can increase throughput (more opportunity for parallelism) but increase latency (more stages for each item to traverse).
Core Idea
The essential commitment is *staging*: dividing a workflow into discrete, separable steps such that each stage accepts outputs from the prior stage and produces inputs for the next, enabling overlap and parallelism without requiring true simultaneity within a single stage .