Version: current [25.0.x]

Dremio's Well-Architected Framework

Dremio’s Well-Architected Framework can help anyone who is designing and/or operating solutions with Dremio. It provides insight from the lessons learned over time helping hundreds of customers be successful. The framework is composed of pillars that provide design principles and concrete best practices that are based on those principles. The best practices are generally applicable across on-premise and public cloud deployments.

The Well-Architected Framework is considered complementary to Dremio’s Shared Responsibility Model. Where the Shared Responsibility Model lays out Dremio's responsibilities and those of customers when it comes to maintaining and operating an optimal Dremio environment, the Well-Architected Framework provides customers with some of the details for how to carry out their responsibilities.

Dremio’s Well-Architected Framework follows five common pillars from cloud providers AWS, Microsoft, and Google, and includes a sixth pillar, the self-service semantic layer, that is specific to Dremio:

Each pillar is a grouping of principles, best practices, and how-to articles around that specific pillar.

Dremio is deployed in production as a cluster. The diagram below outlines the key components in the Dremio architecture.

Dremio Architecture

Grant access to Dremio with the Dremio console, Dremio API, Arrow Flight, or Dremio ODBC/JDBC drivers. For high availability, configure one or more coordinator nodes with the master-coordinator role, though only one coordinator can actively assume the master-coordinator role at any one time. Add one or more scale-out coordinators to help with planning queries that come from JDBC and ODBC clients. You can also configure one or more executor nodes—the number of executor nodes to configure depends on your workload. Executor nodes are organized into logical engines to provide workload isolation. The Distributed Store is set up locally and configured on all Dremio nodes.