Cloud Computing Study Guide - 01
Study Guide
Beginner → Exam Ready
By Levi J.
2026 Edition
ETL vs ELT: Data Processing Approaches
Overview
When extracting data from a source and ingesting it into a cloud data model system, there are two main
approaches depending on how the storage and transformation engines are connected:
ETL (Extract, Transform, Load)
ELT (Extract, Load, Transform)
ETL (Extract, Transform, Load)
Definition
ETL is a data processing workflow that processes data in a strict sequence:
1. Extract - Pulls data from various sources
2. Transform - Changes data according to business rules
3. Load - Loads it into a destination database
Why We Use ETL
By separating each phase into a distinct step, properly designed ETL pipelines can:
Process multiple sources in parallel
Save time and improve efficiency
Page 1 of 6
, Cloud Computing Study Guide - 01
ETL Process Phases
Phase 1: Extract
Goal: Pull raw data from disparate source systems and consolidate it
Staging Layer: A consolidated storage area where extracted data is placed to make it easier for
transformation tools to access without straining the source systems
How to Extract:
Incremental extractions: Pulling only the source data that has been recently inserted or updated,
which minimizes processing time
Full data extractions: Taking a complete copy of the source data, which takes longer but is necessary
if system records have changed
Phase 2: Transform
Goal: Clean the extracted data, remove any formatting inconsistencies, and shape it to the specific structure
required by the destination data model
Stage Transformations: To keep pipelines running efficiently, transformations are often divided into three
intermediate stages:
Bronze layer - Minimal transformations
Silver layer - Refined data where data is scrubbed of bad records and merged together
Gold layer - Final curated data for consumption
Phase 3: Load
Goal: Take the final curated data and load it into the destination tables
Schema-on-Write: The strict requirement that data must be shaped to perfectly match the destination tables'
preferred schema before it is loaded
How to Load:
Data can be loaded incrementally by adding new values to existing tables
If a full swap-and-replace is needed, the system will load the new data into a temporary staging table
first, then instantly swap it with the production table to ensure end-users never experience downtime or
any errors
Page 2 of 6
Study Guide
Beginner → Exam Ready
By Levi J.
2026 Edition
ETL vs ELT: Data Processing Approaches
Overview
When extracting data from a source and ingesting it into a cloud data model system, there are two main
approaches depending on how the storage and transformation engines are connected:
ETL (Extract, Transform, Load)
ELT (Extract, Load, Transform)
ETL (Extract, Transform, Load)
Definition
ETL is a data processing workflow that processes data in a strict sequence:
1. Extract - Pulls data from various sources
2. Transform - Changes data according to business rules
3. Load - Loads it into a destination database
Why We Use ETL
By separating each phase into a distinct step, properly designed ETL pipelines can:
Process multiple sources in parallel
Save time and improve efficiency
Page 1 of 6
, Cloud Computing Study Guide - 01
ETL Process Phases
Phase 1: Extract
Goal: Pull raw data from disparate source systems and consolidate it
Staging Layer: A consolidated storage area where extracted data is placed to make it easier for
transformation tools to access without straining the source systems
How to Extract:
Incremental extractions: Pulling only the source data that has been recently inserted or updated,
which minimizes processing time
Full data extractions: Taking a complete copy of the source data, which takes longer but is necessary
if system records have changed
Phase 2: Transform
Goal: Clean the extracted data, remove any formatting inconsistencies, and shape it to the specific structure
required by the destination data model
Stage Transformations: To keep pipelines running efficiently, transformations are often divided into three
intermediate stages:
Bronze layer - Minimal transformations
Silver layer - Refined data where data is scrubbed of bad records and merged together
Gold layer - Final curated data for consumption
Phase 3: Load
Goal: Take the final curated data and load it into the destination tables
Schema-on-Write: The strict requirement that data must be shaped to perfectly match the destination tables'
preferred schema before it is loaded
How to Load:
Data can be loaded incrementally by adding new values to existing tables
If a full swap-and-replace is needed, the system will load the new data into a temporary staging table
first, then instantly swap it with the production table to ensure end-users never experience downtime or
any errors
Page 2 of 6
