In this scheme, Trino is responsible for data processing — how does the technology stack allow separating storage and computation?

Trino is a distributed SQL engine for data analysis that allows executing queries across different data sources without the need to move the data itself. In such an architecture, the separation of storage and computation is achieved as follows:

• Data storage (e.g., HDFS, S3, databases) is responsible for long-term data storage.
• Trino acts as a computation layer that processes queries, performs aggregations, filtering, and joins, accessing data in the storage where it resides.

This allows scaling computational resources independently of storage, optimizing costs, and increasing flexibility. For example, you can increase the number of Trino nodes to speed up processing without changing the storage.

Example: data is stored in S3, Trino connects to it via a connector, executes SQL queries, and returns results without copying data locally.