Use S3 and data cloud (snowflake) to simplify. Object store and data cloud
Store in teh data cloud in a well defioned data model: structured and unstructured. Partition correctly. Use external functions to clean data ready for ML. Use external functions to run ML models (all this eliminates additional tools)
Have a variety of runtime engines to run ML models:
Results to Object store Share directly from here to visualisation tools
Use variant column type, gets structured and semi structured data in a single data model. Then can query both with single SQL. Reduced storage by huge factor, as e.g. 100 chanels of data in one row.
to organise data in the way that it is queried.
Data cloud manages scaling of teh external functions.
Traditional is Python notebooks or e.g. Tableau Can now use native query in snowflake. Input data and output data in the same platform with SQL Build templates that you can use directly, don't need to do so much coding to visualise data
Time sensitive models doing predict for real time vs Batch models
Add snowflake metrics into cloudwatch to make it easier for ops teams
monitoring for DevOops
