Accordia: The Federated Coordination Layer
Accordia is not a database; it is the coordination fabric that sits above fragmented data silos. It solves the classic distributed data problem: how to treat a collection of disconnected Postgres, Mongo, and S3 buckets as a unified, queryable domain without moving all the data into a central warehouse. Accordia provides the shared contract (schema), the routing logic (ownership), and the execution engine (federated planning) needed to run joins and aggregations across the entire estate.
Layer 1: The Schema Registry
The foundation of Accordia is a unified schema registry. Every data silo registers its domain models here, producing a canonical representation of what each entity looks like. This is the "shared contract"—a single source of truth for the shape of the data, regardless of how it is stored on disk.
- Entity Registration: Silos declare entities, fields, types, and relationships.
- Schema Evolution: Versioned schemas allow the registry to track changes without breaking federated queries.
- Cross-Domain Links: Entities can reference IDs in other silos, which the registry resolves for the query planner.
By centralizing the schema, Accordia allows the query planner to know exactly which fields exist in which silo before a query even runs.
Layer 2: The Ownership Graph
The Ownership Graph is Accordia’s core differentiator. Instead of a flat ACL, it explicitly maps which domain owns which entity and which silos host those entities. This graph is the routing engine for the entire system: it tells the query planner where the authoritative copy of each table lives and who is allowed to query it.
- Entity Ownership: A clear assignment of who owns the lifecycle of each entity.
- Silo Mapping: A many-to-many mapping between owned entities and the physical silos that host them.
- Policy Routing: Query access is determined by ownership—if the requester has permission on the entity, the Ownership Graph routes the subquery to the correct silo.
The graph also enables federated joins by identifying common keys across entities and routing the necessary fragments to the appropriate silos.
Layer 3: Federated Query Planning
When a query enters Accordia, it is decomposed into a distributed execution plan. The query planner uses the Schema Registry to expand the query, the Ownership Graph to locate the data, and a cost-based optimizer to determine how to join and aggregate the results.
- Query Decomposition: A global SQL query is broken into subqueries for each involved silo.
- Join Routing: The planner decides whether to ship a small table to a larger silo's engine or vice-versa.
- Execution: Subqueries run in their native environments, and Accordia merges the results in a lightweight coordinator.
Because execution happens in the native silos, Accordia can leverage the specialized query engines of Postgres, MongoDB, and others directly.
Layer 4: Governance and Observability
Governance is baked into the coordination layer, not bolted on. Because every federated query routes through Accordia, the system can audit every access and maintain a complete lineage of how data flows from the silos to the final result.
- Fine-grained Auditing: Log who accessed what entity, which silo served it, and what subqueries were emitted.
- Data Lineage: Automatically trace the provenance of every field in a federated result.
- Throughput and Latency: Monitor the performance of each silo's subquery and the federation merge.
Accordia turns the fragmented data estate into a coherent, governed, and queryable whole.