Amazon DynamoDB · Rate Limits

Dynamodb Rate Limits

DynamoDB throttling is governed by per-table and per-account capacity quotas (RCU/WCU or request units), not a flat per-second API rate. Quotas apply per AWS Region and are adjustable via AWS Service Quotas. Throttled requests return ProvisionedThroughputExceededException or RequestLimitExceeded.

11 Limits Throttle: 400 Quota: 400
Rate LimitingDatabaseNoSQL

Limits

Per-table read throughput (provisioned) account/region/table
read_capacity_units · second
40000
Default per-table provisioned RCU; adjustable via AWS Service Quotas.
Per-table write throughput (provisioned) account/region/table
write_capacity_units · second
40000
Default per-table provisioned WCU; adjustable via AWS Service Quotas.
Per-account read throughput (provisioned) account/region
read_capacity_units · second
80000
Sum across all tables and GSIs in a Region; adjustable.
Per-account write throughput (provisioned) account/region
write_capacity_units · second
80000
Sum across all tables and GSIs in a Region; adjustable.
Per-table on-demand read throughput account/region/table
read_request_units · second
40000
Initial default for on-demand tables; auto-scales further with traffic.
Per-table on-demand write throughput account/region/table
write_request_units · second
40000
Initial default for on-demand tables; auto-scales further with traffic.
Tables per account per Region account/region
tables · usage
2500
Initial default; can be increased to 10,000 via AWS Support.
Global secondary indexes per table table
indexes
20
Local secondary indexes per table table
indexes
5
Provisioned-throughput decreases per table per day account/region/table
operations · day
27
4 in first hour plus 1 per hour, capped at 4 available at any time.
Concurrent restores account
concurrent_operations · usage
50
Up to 50 TB total across in-flight restores.

Policies

Adaptive capacity
DynamoDB redistributes throughput across hot partitions automatically; hot keys can still throttle if they exceed partition-level limits (~3,000 RCU / 1,000 WCU per partition).
Backoff and retry
AWS SDKs retry throttled requests with exponential backoff and jitter; honor RequestLimitExceeded and ProvisionedThroughputExceededException by reducing rate.
Service Quotas
Increase per-table and per-account read/write throughput, table count, and concurrent operations by requesting limit increases through the AWS Service Quotas console or AWS Support.
On-demand auto-scaling
On-demand tables double previous peak capacity within 30 minutes; sustained step changes may briefly throttle until DynamoDB scales.
Burst capacity
Provisioned tables retain up to 5 minutes of unused capacity for burst traffic; use sparingly as it is not guaranteed.

Sources