Skip to main content
The Ocient Hyperscale Data Warehouse is a modern, scalable distributed system capable of enabling complex OLAP-style analysis on the world’s largest structured and semi-structured datasets using SQL. It features a purpose-built storage and I/O layer designed for highly parallel interaction with modern NVMe SSD drives in a Compute Adjacent Storage Architecture (CASA) that leverages erasure coding for cost-efficient reliability. On top of this is a fully integrated execution engine and optimizer that were each built from scratch using modern design principles in a manner that deeply integrates with the storage and I/O layer. The result is a state-of-the-art data warehouse that can deliver outstanding, and even unprecedented, analytical capability on datasets with record sets numbering in the trillions, quadrillions, and beyond.

​
Architecture Topics

Core Elements of an Ocient System

Understanding the fundamental components of an Ocient deployment

Compute Adjacent Storage Architecture

Learn about CASA and how it optimizes performance

Query Performance Optimizations

How Ocient optimizes query execution for large datasets

Distributed Query Execution

Understanding how queries execute across nodes

Storage and Compression

Data storage formats and compression techniques

Fault Tolerance

Erasure coding and system resilience

​
Key Architectural Concepts

​
Compute Adjacent Storage Architecture (CASA)

CASA collocates NVMe drive storage with the system compute resources to optimize performance. This design keeps records near and accessible for computation without a separate storage layer, which avoids many common bottlenecks for database engines.

​
Node Types

Ocient is a distributed system consisting of interconnected nodes:
  • SQL Nodes - Parse incoming SQL statements and administer commands
  • Loader Nodes - Manage ETL ingestion and index data
  • Foundation Nodes - Store data and perform query processing

​
Erasure Coding

The Ocient System uses erasure coding to organize and compute parity blocks so the system is fault-tolerant and can rebuild missing data without redundant copies.