Принципы построения дата пайплайнов из презентации:

- Understand the data consumer
- Understand the data
- Keep data in its raw form
- Do not delete or move your raw data
- Validate your extracted data before saving
- Transform your data over all time
- Separate you E-TL
- Minimize the number of data and compute nodes
- Store all you data of practically feasible to do so
- Make your ETL acyclical
- Validate your data before its given to consumers
- Join your data at the database level
- Monitor your data