Blast radius
How much of a system, and how many customers, a single failure, bad deploy or leaked key can affect.
It is a way to judge designs rather than a pattern: a deploy that goes to all users at once has the largest radius; a canary release to a small share, a feature flag, a cell or a key limited to one service each shrink it. The same question applies to AI agents: what is the worst a single wrong tool call can do?
Related terms
- Cell-based architecture: Running several full, isolated copies of a system (cells), each serving a share of customers, so a failure or bad deploy hits one cell only.
- Canary release: Sending a new version to a small share of traffic first, watching error rates and latency, and widening the rollout only if it holds.
- Feature flags: Switches in code that turn a feature on or off for some or all users without a new deploy.
- Bulkhead pattern: Giving each dependency or workload its own limited pool of connections or workers, so one slow part cannot use up everything.
