Snowflake alongside Postgres

Snowflake is not a replacement for Postgres. They serve different workloads. Postgres is your transactional serving database: small queries, low latency, concurrent reads. Snowflake is your analytical warehouse: multi-gigabyte scans, big joins, pay-per-compute elasticity. A serious platform often runs both.

Postgres vs Snowflake

Serving versus analytics. Latency versus scale. Row versus column.

Two warehouses, two workloads, one sync to keep them in step.
snowflake/ddl.sql
sql
USE WAREHOUSE PIPELINE_WH;
USE DATABASE PIPELINE_DB;
USE SCHEMA ANALYTICS;

CREATE OR REPLACE TABLE DIM_CUSTOMERS (
  DIM_CUSTOMER_SK NUMBER AUTOINCREMENT,
  CUSTOMER_NATURAL_KEY VARCHAR(64),
  CUSTOMER_EMAIL VARCHAR(256),
  CUSTOMER_COUNTRY VARCHAR(2)
);

CREATE OR REPLACE TABLE FACT_ORDERS (
  ORDER_NATURAL_KEY VARCHAR(64),
  DIM_CUSTOMER_SK NUMBER,
  ORDER_TS TIMESTAMP_NTZ,
  AMOUNT NUMBER(12,2),
  STATUS VARCHAR(16)
);

CREATE OR REPLACE TABLE AGG_DAILY_ORDERS (
  ORDER_DAY DATE,
  TOTAL_AMOUNT NUMBER(18,2),
  ORDER_COUNT NUMBER
);

A Snowflake star schema. Notice the WAREHOUSE clause: Snowflake compute is elastic and paid per minute. Your DDL is close to Postgres DDL but the runtime model is different.

You can, and many teams do. The friction is cost and latency for transactional patterns. A dashboard that makes a thousand little queries an hour will saturate Snowflake credits fast. Postgres handles that workload at a fraction of the cost. The common pattern is: serving on Postgres, analytics on Snowflake, periodic sync between them.

Ordering exercise: Order the criteria that tip you toward Snowflake

Loading practice…

Quiz: Quiz

Loading practice…

Checkpoint: Checkpoint: you can defend every service in the platform

Loading practice…