Lakehouse Architecture — Merging Data Lake and Warehouse¶

Lakehouse unifies data lake and warehouse into a single layer. Open table formats, medallion architecture and unified data access.

From Warehouse and Lake to Lakehouse¶

Medallion Architecture¶

• Bronze — raw data, append-only
• Silver — cleansed, validated
• Gold — business aggregations

# Bronze: ingestion from Kafka
bronze.writeStream.format("delta")
    .start("/lakehouse/bronze/orders")

# Silver: cleansing
silver = spark.read.format("delta")
    .load("/lakehouse/bronze/orders")
    .dropDuplicates(["order_id"])
silver.write.format("delta").save("/lakehouse/silver/orders")

# Gold: aggregation
gold = spark.read.format("delta")
    .load("/lakehouse/silver/orders")
    .groupBy("order_date").agg(sum("total_czk").alias("revenue"))
gold.write.format("delta").save("/lakehouse/gold/revenue")

Advantages¶

• Single storage — no duplication
• Open formats — no vendor lock-in
• Cost efficiency — inexpensive object storage

Summary¶

Lakehouse with the medallion pattern is the preferred approach. Bronze-Silver-Gold ensures progressive quality improvement.