On-Prem Warehouse to Snowflake Program Playbook
A wave-based program for migrating a legacy on-prem data warehouse to Snowflake. It covers workload discovery, an ELT redesign on a medallion architecture, validated migration, and disciplined cost tuning before decommissioning.
Legacy on-prem data warehouses such as Teradata, Netezza, or Oracle Exadata are expensive to operate, hard to scale, and increasingly disconnected from cloud-native analytics. Snowflake offers elastic compute, separation of storage and compute, and a consumption model that aligns cost with use. This playbook runs the migration as a wave-based program so analytics keeps running while you move.
The key shift is architectural: most legacy warehouses are built around ETL, where transformation happens before load. Cloud warehouses favor ELT, where raw data lands first and transformations run inside the warehouse. Redesigning for ELT, often with a medallion (bronze/silver/gold) layering, is where most of the value and most of the work lives.
Phase-by-Phase
Workload Discovery. Catalog tables, jobs, reports, and the consumers that depend on them. Profile query patterns to understand concurrency and peak load, then group workloads into migration waves ordered by value and risk.
Platform and ELT Foundation. Provision Snowflake with infrastructure as code, define virtual warehouse sizing and auto-suspend policies, and stand up a standard ELT framework so every team builds transformations the same way.
Data and Pipeline Migration. Move historical data, rebuild transformations as version-controlled models, and validate outputs against the legacy system row-by-row and metric-by-metric before any consumer switches over.
Cost Tuning and Decommission. Cloud warehouses make it easy to overspend. Right-size warehouses, set resource monitors, migrate consumers wave by wave, and retire the legacy platform only after parity is confirmed.
Team and Roles
A data architect owns the target model and ELT standards. Data engineers migrate pipelines and data. A DBA supports performance and access. DevOps manages provisioning and CI. A product owner sequences waves around business priorities and consumer readiness.
Risks and Mitigations
Cost overrun is the signature risk of consumption-based platforms; control it with resource monitors and tagging from day one. Guard against transformation drift with automated output validation, and avoid consumer disruption by migrating in waves with overlap.
Success Criteria
Target improved query latency for analytics, predictable cost per query, verified data parity, and high consumer adoption with the legacy warehouse fully retired.
Tooling
Use Terraform for provisioning, a transformation framework for ELT models in version control, and observability through Datadog. Drive deployments and tests through CI so transformations are tested like application code.