In most data architectures, if you want to work with relational tables, JSON documents, and native objects, you often need separate, specialized databases. This leads to data duplication, complex synchronization logic, and performance bottlenecks. Each time data moves between these systems, it must be transformed, a process that adds overhead and points of failure.

InterSystems IRIS is built on a different philosophy: a Common Data Plane (CDP). Instead of juggling multiple databases, IRIS uses a single, unified engine to store data once and make it accessible through different models simultaneously. Whether you query via SQL, interact with it as a native object, or manipulate it as a NoSQL document, you're always working with the same underlying data. This “store once, access many” approach eliminates redundancy and the need for complex data mapping layers like Object-Relational Mapping (ORM).

The core of this architecture is a data structure called a Global—a high-performance, multidimensional sparse array. Think of a Global not as a flat table, but as a vast, hierarchical key-value store. The keys, called subscripts, can be numbers or strings and define a path to a value. Because the arrays are sparse, they only consume storage for the nodes that actually contain data. An empty branch in the hierarchy takes up no space.

Under the hood, Globals are implemented using a highly optimized [{<B+ tree>] architecture. Every node in the multidimensional array is a block in a B+ tree, which allows for incredibly fast traversal and data retrieval, even with billions of records. This structure is the key to IRIS's ability to scale both vertically (on a single powerful machine) and horizontally (across a distributed cluster of machines).

Crucially, this architecture achieves high performance without compromising data integrity. All operations on Globals are fully ACID compliant, whether you're performing a multi-row SQL update or modifying a complex JSON document. Transactions are atomic, consistent, isolated, and durable. This is a significant advantage over many NoSQL databases that often sacrifice full ACID compliance for speed or scalability. With IRIS, you get the flexibility of multi-model access with the reliability of a traditional relational database.