Skip to main contentSkip to user menuSkip to navigation
Interactive design lab

Architecture Decision Workbench

What is an architecture decision?

An architecture decision is a consequential technical choice with explicit context, viable alternatives, and accepted downsides. A useful decision record explains why the choice is defensible and what evidence could overturn it.

Step 1

Model

Model traffic, latency, consistency, data shape, geography, and operating capacity. Compare candidates before selecting the path your team is prepared to own.

Step 2

Observe

Leave with a ranked alternative, a visible serving path, explicit contradictions, required mitigations, and a shareable architecture decision record.

Step 3

Challenge

Choose a candidate, then inject a traffic spike, region loss, hot partition, or replication lag. Increase severity until the recommendation changes or an accepted downside becomes a user-visible failure.

Architecture decision workbench

Turn constraints into a defensible decision

Compare viable data paths, choose a candidate, then inject a failure to expose its accepted consequences.

Decision state
Defensible starting point
0 exposed conflicts
Best fit
76/100
Relational core
Stressed load
5K RPS
2.8x capacity headroom
Projected p95
42 ms
150 ms target
2. Compare alternatives

Choose what you are willing to operate

Fitness scores explain the current inputs. Select any candidate to stress it, even when it is not the top recommendation.

3. Trace the consequence

Follow the selected serving path

Evaluating: Relational core
Clients
Multiple regions
Edge and routing
Regional routing
Service boundary
5K RPS under test
Data ownership
Relational core

Decision logic

  • fits relational data and joins.
  • keeps the operating model familiar.

Accepted downside

Shard deliberately before one writer becomes the permanent bottleneck.

4. Challenge the decision

Inject pressure before production does

This loop is independent of the requirement model.

Measure the design against the stated peak without an injected failure.

Major
Observed outcome
Defensible starting point

The selected path retains 2.8x headroom at the tested load.

Capacity used
36%
14,000 RPS plan
Availability
99.95%
Normal operation
Consistency
Low
bounded

Exposed contradictions

No direct contradiction is visible. That is a testable hypothesis, not proof that the design is complete.

Required mitigations

  1. 1Prefer managed infrastructure and one clear operational ownership boundary.
Take the decision into design

The whiteboard export includes the active architecture path. The ADR includes its constraints, consequences, and challenge result.

No quiz questions available
Could not load questions file
Spotted an issue or have a better explanation? This page is open source.Edit on GitHub·Suggest an improvement