Aamazon SageMaker feature store throughput and latency

There are soft limits on Feature Store TPS, feature number etc. But soft limits can be increased based on your need.

• Maximum number of feature groups per AWS account: Soft limit of 100.
• Maximum number of feature definitions per feature group: 2500.
• Maximum Transactions per second (TPS) per API per AWS account: Soft limit of 10000 TPS per API excluding the BatchGetRecord API call, which has a soft limit of 500 TPS.
• Maximum size of a record: 350KB.
• Maximum size of a feature value: 350KB.
• Maximum number of concurrent feature group creation workflows: 4.
• BatchGetRecord API: Can contain as many as 100 records and can query up to 10 feature groups.

You can check Limits and Quotas in the SageMaker Developer Guide.

Depending on your data ingestion use cases, your requirements and runtime context might be different:

• Experimenting with new ML features
• Streaming feature ingestion
• *Bulk feature ingestion as part of a batch pipeline **

For more details on use cases and Feature ingestion APIs, see this blogpost