← Library / Lakehouse & Databricks
Slide 1Slide 2Slide 3Slide 4Slide 5Slide 6Slide 7Slide 8Slide 9Slide 10
1 / 10

Use the arrows, swipe, or press ← → to move through the slides.

Lakehouse & DatabricksCarousel · 10 slidesSystem Design Series · Day 10

S3/GCS Partitioning Strategies — pruning and the small-file problem

How you lay data out on object storage decides query speed and cost: Hive-style layouts, range vs hash vs composite partitions, pruning and small files.

The trade-off I'd flag

Over-partitioning is the most common mistake on a lakehouse. If a partition holds less than a few hundred MB, you have made the query planner's life harder, not easier.

← PreviousApache Kafka Masterclass Guide (63 pages) Next →Hot / Warm / Cold Tiering — lifecycle policies and cost vs latency