Lakehouse basics

1. Lakehouse basics

Previously, we compared a lakehouse to a flexible workshop. Now let's step inside. Every lakehouse is laid out the same way, with a place for raw material and a place for the parts you build from it.

2. The two main lakehouse areas

Let's begin with the two storage areas we mentioned earlier: Files and Tables. Files holds the raw material, and Tables holds what we build from it. In our workshop example, Files is the crates of raw material, and Tables is the shelf of finished parts.

3. The Files area

The Files area works much like the file system on a computer. In business terms, it's like holding raw materials in a workshop. Business systems export data as files. For example, a sales system might drop a daily CSV of transactions. The Files area holds those exports so they can be cleaned and transformed, and the original stays untouched.

4. Files in File view

When we open a dataset file in a lakehouse, it opens in File view, which shows the raw content. For a text file such as a CSV, that's the same as opening it in Notepad. We see every row, but not the shape of the data. Working out which value belongs to which column means tracing the separator characters ourselves.

5. Files in Table view

Switching from File view to Table view lays the same file out as a grid. Each column is now distinctly visible, which makes the data much easier to read and navigate.

6. The Tables area

Under the hood, every lakehouse table is stored in the open Delta format. A Delta table is a set of Parquet data files plus a transaction log that records every change. Because the format is open, the same table can be read by SQL, Spark, semantic models, and reports without being copied. That shared format is one reason OneLake can serve every workload in Fabric.

7. Creating a table from a file

We can create tables in a lakehouse directly from dataset files. This works for CSV and Parquet files. This is done by right-clicking a file and choosing the Load to Tables option. The original file remains, but its data is now also in a table, where it's much easier to access.

8. Configuring the new table

When we create a new table, we have a few choices. We can set the column separator used in the original data, name the target table, and indicate whether the first row contains column headers. That last one matters, because not all datasets have headers. If we select the Use headers for column names option, the text in the first row becomes our column names.

9. Let's practice!

Now let's explore the Files and Tables areas of a lakehouse.

Create Your Free Account

By continuing, you accept our Terms of Use, our Privacy Policy and that your data is stored in the USA.