Migration
Informatica PowerCenter to Snowflake migration: options and a practical plan
How to move Informatica PowerCenter ETL to Snowflake: keep Informatica and push down, or rebuild in dbt, Snowflake SQL and Snowpark. Transformation mapping, orchestration, CDC and reconciliation.
Hemant Walchale
· 2 min read
When a company moves its warehouse to Snowflake, the Informatica PowerCenter estate that feeds it has to go somewhere. There are two sensible paths, and choosing between them early saves months.
Option A: keep Informatica, push the work into Snowflake
Informatica’s cloud platform can push transformation logic down into Snowflake as SQL, so the compute runs in Snowflake while the mappings stay in Informatica. This is the faster path: skills, mappings and operating procedures stay familiar. The trade-off is that you keep an ETL licence and split logic across two tools.
Option B: rebuild the logic in Snowflake
Here the mappings become SQL models (most teams use dbt), with Snowflake features doing the work PowerCenter did:
| PowerCenter | On Snowflake |
|---|---|
| Source Qualifier | Data landed by Snowpipe, COPY INTO or a CDC tool; the SQL override becomes the first model |
| Expression, Filter, Router | SELECT expressions, WHERE and CASE |
| Joiner, Lookup | Joins |
| Aggregator, Sorter, Rank | GROUP BY, window functions, QUALIFY |
| Update Strategy | MERGE, or incremental models in dbt |
| Sequence Generator | Sequences or IDENTITY columns |
| Slowly changing dimensions | dbt snapshots or MERGE logic |
| Mapplets | Shared dbt models or macros |
| Java, SQL, Stored Procedure transformations | SQL, Snowpark or user-defined functions; rewrite and test each |
| PowerExchange CDC | A CDC tool landing changes, applied with Streams and Tasks or MERGE |
| Workflows, sessions, scheduler | Snowflake Tasks, dbt jobs or an orchestrator such as Airflow |
Option B usually costs less to run and puts transformation, testing and documentation in one version-controlled place.
The plan, whichever option you choose
- Export and inventory. Export the repository as XML with
pmrepor Repository Manager, count transformation types, and use workflow run history to find what still runs. Retire the rest. - Convert in domains. Move one business domain at a time (finance, risk, customer), so owners can sign off their own numbers.
- Watch the usual differences. Implicit type conversions in Expression transformations, null handling in lookups, trailing spaces, decimal precision, and time zone and date defaults set at session level.
- Parallel run and reconcile. Run old and new against the same sources and compare every target table before cutting over.
- Retire workflows on the day of cutover, and plan back from the licence renewal date.
The short version
Decide early between pushdown (faster, keeps the licence) and rebuilding in dbt and Snowflake (cheaper to run), inventory and retire unused mappings, convert by domain, reconcile every table and switch off workflows as each domain moves.
Moving to Databricks instead? See our Informatica to Databricks guide, or browse all migration guides. For an estimate on your own estate, see the Data Platform Assessment.
Not sure where your own platform stands? Take the free 3-minute data platform scorecard, or see how we work on Snowflake. Unfamiliar terms are explained in the data engineering glossary.