Latency Planning
Timeout Budget Planner
Allocate a user-facing deadline across downstream calls, connection and processing time, retries, and safety margin.
BUDGET SUMMARY
Results
Use the defaults or enter the request path you need to evaluate.
Enter values to see the critical-path equation.
READ THE BUDGET
A deadline must cover the whole critical path.
A dependency timeout is only one slice of the user-facing deadline. Queueing, connections, processing, retries, serialization, and application work all compete for the same elapsed time.
- 1Reserve safety time
The safety margin stays outside the dependency budget so the application can finish its own work and return a response.
- 2Build the critical path
Sequential dependency budgets add together. Independent concurrent calls are modeled by the longest call.
- 3Fund complete attempts
Each retry needs enough remaining time for both its connection and processing budget. A partial attempt is not useful.
- 4Stop at the deadline
Pass the remaining deadline downstream and avoid starting work that cannot finish before it expires.
SUGGESTED CHANGES
Ways to improve this plan
- Calculate a plan to see suggestions.
Planning model: This calculator uses equal per-dependency budgets and assumes concurrent calls are independent. It does not measure your network or service. Validate the plan with production latency distributions, connection-pool limits, provider quotas, cancellation behavior, and end-to-end tests. Read Choosing Reasonable API Timeouts, then use the concurrency simulation and retry planner for related decisions.