Amazon SageMaker Feature Store introduces UpdateRecord for feature-level writes

This article announces the UpdateRecord API for Amazon SageMaker Feature Store, enabling efficient partial feature updates without full record rewrites.

• UpdateRecord API allows updating individual feature values atomically without read-modify-write cycles
• Available for both Standard (DynamoDB-backed) and In-Memory (ElastiCache-backed) online store tiers
• Standard tier requires new Standard_V2 storage format; In-Memory tier works with existing feature groups
• Eliminates latency, reduces read capacity costs, and prevents lost updates in multi-pipeline architectures
• Supports EventTime-based temporal ordering to reject stale writes
• Integrates with IAM for fine-grained access control via sagemaker:IsUpdateRecord and sagemaker:UpdatableFeatures conditions
• Automatic offline store replication maintains accurate training datasets
• Two migration strategies available: bulk migration with Feature Processor or in-place UpdateFeatureGroup API switchover

UpdateRecord simplifies feature engineering workflows by enabling efficient partial updates, reducing operational complexity for real-time and multi-producer feature group scenarios.