The Cloud Data Lake

Chapter 6. Deep Dive on Data Formats

Design is not just what it looks like and feels like. Design is how it works.

Steve Jobs

Traditionally, data warehouses are built on a proprietary data format that they leverage to optimize for the query patterns. Given the increasing number of scenarios that are served by the cloud data lake, especially with the rise of the lakehouse architectural pattern, more and more customers and solution providers are investing in capabilities that enable running warehouse-like queries directly on the cloud data lake. This takes us close to the promise of delivering an architecture that minimizes the need to copy data back and forth across data stores for specific purposes. This promise of a data storage with no silos has resulted in an increasing number of open data formats that enable running warehouse-like queries directly on a cloud data lake storage. In this chapter, we’ll take a look at three such formats: Apache Iceberg, Delta Lake, and Apache Hudi. This chapter is probably the most technical one in the book, where we look at the formats in great detail, including how they serve the scenarios they are designed for. My hope is that this chapter provides you with enough knowledge on why these formats were designed so that when you evaluate one of these formats, you can ask the right questions and find the right data format for your cloud data lake architecture.

Why Do We Need These Open Data Formats?

If I had to summarize the need for open data ...