USER MANUALS

Materialization Decision Guide: Comparison Matrix

This table presents the available options to materialize data. This will help you choose the most appropriate materialization option for your use case.

Summary

Full Materialization

Query Results (pka partial)

Use cases

  • Ad-hoc Reporting and Analytics Acceleration

  • Protect data sources

  • ELT modeling

  • Precalculate complex transformations

  • Protect data sources

  • Manage workloads

  • Enhance performance (for specific scenarios)

  • Reduce Costs (SaaS API)

Usage

Dynamic: Optimizer can decide to leverage a Summary on a query execution

Always. Data is always retrieved from materialization.

From materialization if possible. Otherwise from underlying source.

Linked to the logical model

Full lineage, invalidation on changes

Full lineage, invalidation on changes

Full lineage, invalidation on changes

Data Load mechanisms

  • Automatic (based on queries)

IDU

N/A

N/A

N/A

Target Materialization

Any

Cache data source

Cache data source

Table life duration

Manual drop

  • Manual drop

  • Cache invalidations (No schema evolution selected)

  • Manual drop

  • Time to Live

  • Cache invalidations (No schema evolution selected)

Remote Tables

Temporary Tables

Materialized Tables

Use cases

  • ETL

  • Machine Learning and Data Science

  • Wizards (Data preparation)

  • Enabler for Reporting/ Analytics tools

  • Store data from independent queries sharing the same schema.

  • SQL compatibility with external tools.

Usage

Always from remote table

Always from temporary table. Only available to the user session

Always from materialized table. For specific scenarios.

Linked to the logical model

Only the SQL query used

None

None

Data Load mechanisms

  • Load on creation

  • Refresh with condition

Explicit

Explicit

IDU

Insert, Delete, Update

Insert

Insert

Target Materialization

Any

Cache data source

Cache data source

Table life duration

Permanent (Can be manually dropped)

Session

Permanent (Can be manually dropped)
Add feedback