AWS Kinesis · Rate Limits
Kinesis Rate Limits
Kinesis Data Streams enforces a mix of per-account/per-region control-plane API TPS limits and per-shard data-plane throughput limits (1 MB/s or 1000 records/s write per shard, 2 MB/s or 5 GetRecords/s read per shard). Limits are scoped per AWS account per region; many control-plane limits are adjustable via Service Quotas. Exceeding throughput returns ProvisionedThroughputExceededException.
16 Limits
Throttle: 400
Rate LimitingStreamingReal-Time
Limits
PutRecord (per shard) shard
1000
Each shard supports up to 1000 records/s and 1 MB/s write.
PutRecords (per shard) shard
1048576
500 records max per request; 10 MiB request limit; 10 MiB max per record (burst).
GetRecords (per shard) shard
5
Returns ProvisionedThroughputExceededException for 5 seconds after a 10 MB response.
GetRecords data read rate (per shard) shard
2097152
SubscribeToShard (Enhanced Fan-Out) consumer-shard
1
One call per second per registered consumer per shard.
GetShardIterator shard
5
ListShards stream
1000
DescribeStream account
10
DescribeStreamSummary account
20
CreateStream / DeleteStream / ListStreams account
5
On-Demand Standard default throughput stream
4194304
4 MB/s write / 8 MB/s read default; auto-scales to 200 MB/s write or higher in select regions.
On-Demand max throughput (US-East-1, US-West-2, EU-West-1) stream
10737418240
10 GB/s write, 20 GB/s read with support-ticket increase.
Provisioned default shard quota account/region
see Service Quotas - defaults 20000 / 6000 / 1000 by region
On-demand streams default account
50
Stream retention stream
8760
Default 24 hours; max 365 days (8760 hours).
Registered consumers per stream stream
50
50 on On-Demand Advantage, 20 on Standard / Provisioned.
Policies
Provisioned throughput exceptions
When per-shard write or read throughput is exceeded, the API returns ProvisionedThroughputExceededException; AWS SDKs implement automatic retry with exponential backoff.
Backoff and retry
Use exponential backoff with jitter on throttling. KCL and KPL libraries handle this automatically for typical workloads.
Capacity-mode switching
A stream can switch between On-Demand and Provisioned modes only twice in any 24-hour window.
Service Quotas adjustments
Most control-plane and shard-count limits are adjustable via the AWS Service Quotas console; some (like UpdateShardCount) require an AWS Support case.
Resharding
Use SplitShard / MergeShards (provisioned) or rely on Kinesis auto-scaling (on-demand) to keep per-shard limits ahead of producer/consumer load.