ELT Pipeline in Datalake

ELT - Extract Load Transform pipeline ingestion from Database into a single datalake location to do transform processing.

Diagram

Introduction & Goals

• Ingest Data from RDBMS to a Data lake bucket.
• Process transaction files to a single location
  • Data set of customer list of sales person
  • We use AWS Glue
  • Create Data Catalog, transform, load to Data lake
  • Data will be transform from CSV to Parquet file
  • Parquet stores the file schema in the file metadata. It is easier to work with because they are supported by so many different projects.
  • Transformed data will be ready for query in Data Warehouse

The Data Set

• We use transaction data set customer.csv.
• Data set of sales person's customer list
• appropiate size 197 KB for Glue ETL demo

Used Tools

• Amazon RDS postgreSQL, AWS Glue, Glue crawler, Athena

  • Load .csv file RDS into input folder inside a bucket.
  • Create output folder as target inside the same bucket.
  
  • Add Glue crawler to craw Data catalog of .csv file ( in input folder)
  
  • run Glue job to transform to parquet file
  
  • Set target to load into output folder
  

• In Athena, check view or query of transformed file.

• Add Glue crawler to craw Data catalog of .parquet file in (output folder).

Pipelines

• Batch processing pipeline for bulk import.
• Use source code of AWS Glue

Conclusion

• Data set has transform inside a single datalake bucket
• File can be loading later to Data Warehouse for analysis purpose.
• Glue job can be assign for aggregate , join, filter tables

Follow Me On

https://www.linkedin.com/in/jirasak-pakdeeto-900665214/