AWS Lambda · Rate Limits
Aws Lambda Rate Limits
AWS Lambda enforces concurrency-based and API-based limits per account per Region. Concurrent executions default to 1,000 (increasable). Each execution environment serves up to 10 synchronous requests per second, so per-function steady-state RPS scales as 10x concurrency. Burst scaling adds 1,000 environments every 10 seconds. Control-plane APIs have separate fixed RPS quotas.
13 Limits
Throttle: 429
ServerlessFunctionsRate LimitingConcurrency
Limits
Concurrent executions account/region
1000
Default per Region; new accounts may start lower. Increasable to tens of thousands via Service Quotas.
Burst concurrency scaling function
1000
Each function can add 1,000 execution environments every 10 seconds (i.e. ~100 environments per second of burst headroom).
Synchronous invocation per env execution-environment
10
Each execution environment serves up to 10 synchronous requests per second; total synchronous RPS is 10 x concurrency.
Asynchronous invocation per env execution-environment
-1
Asynchronous invocation is unlimited per environment; total bounded only by available concurrency.
Provisioned concurrency invocation function-version-or-alias
-1
10 x allocated provisioned concurrency. Applies only to functions with provisioned concurrency configured.
GetFunction API account
100
Fixed; not increasable.
GetPolicy API account
15
Fixed; not increasable.
Other control-plane APIs account
15
15 RPS aggregate across all other control-plane APIs (excludes invocation, GetFunction, GetPolicy). Not increasable.
Storage for uploaded functions / layers account/region
75
Increasable to terabytes via Service Quotas.
Function timeout function
900
15 minutes max execution time. Fixed.
Sync invocation payload invocation
6
6 MB request and 6 MB response (sync). 200 MB for streamed responses. 1 MB async.
Deployment package (zip) function-version
50
50 MB via API/console; 250 MB unzipped including layers; container images up to 10 GB.
ENIs per VPC vpc
500
Shared with other services such as EFS.
Policies
Backoff Strategy
AWS SDKs implement exponential backoff with jitter on 429 / TooManyRequestsException / 5xx automatically. Raw HTTP clients should mirror this and honor Retry-After.
Reserved & Provisioned Concurrency
Use reserved concurrency to cap a function's slice of the account pool; use provisioned concurrency to pre-warm and eliminate cold starts (billed separately).
Service Quotas Increases
Concurrent executions, storage, ENIs, and durable executions are increasable via Service Quotas; control-plane RPS, invocation payload, timeout, and runtime quotas are fixed.
Burst Headroom
Cold-start bursts beyond the 1,000-environments-per-10-seconds rate will throttle even if total concurrency is well below the account ceiling — pre-warm with provisioned concurrency for spiky traffic.
Idempotency
Asynchronous and event-source invocations may retry on failure (default 2 retries asynchronous). Design handler logic to be idempotent on the event identifier.