DEV Community

M. Haresh Kumar
M. Haresh Kumar

Posted on

Snowflake Medallion Architecture & Idempotent MERGE INTO Guide

Why Snowflake Medallion Architecture & Idempotent MERGE INTO are Critical for High-Volume Data Engineering:

In high-volume streaming and batch pipelines, 5% of transactional data often arrives 2 to 3 days late due to upstream system delays or network retries.

If your pipeline relies on simple INSERT statements or un-governed batch scripts, late-arriving records cause duplicate rows, corrupted reporting metrics, and inaccurate executive dashboards.

Over 15+ years managing enterprise data operations, SLA delivery, and pipeline reliability, I enforce a 3-tier Medallion Architecture pattern:

  1. Bronze Layer (Raw Landing):
    Ingest raw JSON/CSV files from S3/Azure stages using continuous event-driven Snowpipe or batch COPY INTO.

  2. Silver Layer (Cleaned & Deduplicated CDC Upsert):
    Place a Snowflake Stream on the Bronze table to capture Change Data Capture (CDC) deltas (INSERT/UPDATE).
    Execute an idempotent MERGE INTO SQL transformation keyed on transaction_id:

MERGE INTO silver_transactions target
USING (
SELECT * FROM raw_transactions_stream
QUALIFY ROW_NUMBER() OVER (
PARTITION BY transaction_id
ORDER BY transaction_timestamp DESC
) = 1
) source
ON target.transaction_id = source.transaction_id
WHEN MATCHED THEN UPDATE SET
target.amount = source.amount,
target.status = source.status,
target.updated_at = CURRENT_TIMESTAMP()
WHEN NOT MATCHED THEN INSERT (transaction_id, customer_id, amount, status, created_at)
VALUES (source.transaction_id, source.customer_id, source.amount, source.status, source.transaction_timestamp);

  1. Gold Layer (Analytics Star-Schema): Materialize aggregated business KPIs into Star-Schema dimensions and fact tables for Power BI reporting.

Why Idempotency Matters:
Idempotent MERGE INTO queries guarantee that whether a pipeline runs 1 time or 10 times, the output state remains 100% clean and deduplicated without creating duplicate records.

These production pipeline patterns and data governance standards are covered in detail in my published book:

📚 "The Comprehensive Power BI & Enterprise Business Intelligence Handbook"
(Available worldwide on Amazon: https://www.amazon.com/dp/B0HCR8Q2TY)

Top comments (0)