Real-time inference needs a customer's latest features in milliseconds. Which store do you use?
Why A is the answer
The online store of SageMaker Feature Store
The online store serves low-latency reads. The offline store holds history for training.
In detail
- The online store gives low-latency access to the latest values.
- Training reads history from the offline store.
Use case
An illustrative scenario, not a real incident.
A fraud check needs a customer's latest features within milliseconds, so it reads the online store.
See it in different ways
flowchart TD A["Feature pipeline"] --> B["SageMaker Feature Store"] B --> C["Offline store: training"] B --> D["Online store: low-latency inference"]
sequenceDiagram participant P as Feature pipeline participant O as Offline store participant N as Online store participant T as Training participant I as Inference P->>O: Historical features P->>N: Latest features O->>T: Training data N->>I: Values in milliseconds
flowchart TD
Q{"What is the read for?"}
Q -->|"Real-time inference"| A0["Online store"]
Q -->|"Training history"| A1["Offline store"]Key point: The online store serves low-latency reads.
Why the other options are wrong
- B. The offline store, queried with Amazon Athena on each request
- The offline store is for training.
- C. A CSV file downloaded from Amazon S3 for each prediction
- Downloading a file is too slow.
- D. An Amazon S3 Glacier archive restored before every prediction
- Restoring an archive is far too slow.