Andy Clark, Tom Gough and Shailan Chudasama return to a topic they first covered a few years ago, and the answer has changed a lot. Tom explains the basics: a data lake is large-scale, file-based storage, and in Fabric everything sits in OneLake anyway. The real difference between a Lakehouse and a Warehouse is how you work with the data. A Warehouse behaves like a traditional SQL Server and is driven by T-SQL. A Lakehouse is built around Python and PySpark and handles unstructured data much better.
Shailan describes how client conversations usually go. They start with goals, data sources, data types, refresh frequency and sensitivity. More clients now want to bring in media, telemetry and other semi-structured data alongside their ERP and finance systems, which pushes them toward a Lakehouse. The team agrees the gap has narrowed sharply. Lakehouse performance has caught up, Spark SQL is close enough to T-SQL to make the switch easy, and the Warehouse can now be read from Python. The clearest remaining reason to choose a Warehouse is a lift-and-shift migration of an existing on-prem SQL Server warehouse, or a hard requirement to write data with T-SQL.
The hosts also cover:
- Future-proofing and recruitment, given Python’s rise as the default entry language.
- Hybrid medallion designs, such as a Silver Lakehouse with a Gold Warehouse, as a way to transition gradually.
- Their prediction that the two will merge into a single analytics platform within a couple of years.
- Copilot now being available across Fabric SKUs from F2 upward, with a reminder that it’s billed separately from capacity.
Key takeaway: Don’t overthink it. Because everything lives in OneLake, neither choice is a bad one, and switching later isn’t hard. Pick based on your team’s existing skills, whether that’s T-SQL or Python. Or, as Tom puts it, Lakehouse vs Warehouse is the rosé wine of Fabric.
Flere episoder fra "The Clearly Podcast"
Gå ikke glip af nogen episoder af “The Clearly Podcast” - abonnér på podcasten med gratisapp GetPodcast.
