Choose Lambda reserved concurrency separately from preinitialized execution capacity

A scaling ceiling and a latency preparation setting solve different operational problems.

Source facts

Reserved concurrency sets an exclusive allocation and maximum concurrency for a Lambda function, reducing the pool available to other functions. Provisioned concurrency prepares execution environments and has separate charges. AWS documents reserved concurrency as a way to limit pressure on downstream resources; setting it to zero intentionally stops event processing.

Source: AWS Documentation.

Applicability

This concerns the concurrency controls described for Lambda functions, with event-source-specific limits checked separately.

DSE recommendation

DSE recommends identifying whether the problem is downstream saturation, shared account capacity or initialization latency before changing a setting. Measure representative request duration and traffic demand. Review the effect on other functions with the account owner and define acceptable throttling behavior. Keep a rollback value and an owner for queue or retry consequences. Do not treat reserved capacity as evidence that execution environments are already initialized for latency-sensitive work. Ask the downstream owner to approve the maximum simultaneous demand the function should generate. Record the desired outcome for throttled work and the observation that would justify a later increase, keeping a latency objective separate from the downstream-capacity decision.

Verification

Proposed verification: inspect the current concurrency configuration, run an approved bounded load against a test function and observe concurrency, throttles and downstream demand. Compare the result with the intended ceiling. Review provisioned-capacity needs separately and record the cost decision before enabling it.

Official reference

AWS Documentation. Reviewed October 20, 2026.