FerretDB 2.0 changes the document database choice

DocumentDB drives new performance

FerretDB 2.0 leverages Microsoft’s DocumentDB extension to accelerate document workloads. This version focuses on PostgreSQL databases through two specific extensions: pg_documentdb_core for BSON optimization and pg_documentdb_api for data operations. I find the 20x speed increase for certain workloads a major reason to reconsider the tool. This update moves the project away from the 1.x versions to focus solely on PostgreSQL-based DocumentDB. This open-source platform uses the MIT license to provide developers with complete freedom to fork the repository. Written in Go, FerretDB functions as a proxy that translates MongoDB 5.0+ wire protocol queries to SQL. The pg_documentdb_core layer parses and manipulates BSON documents at all levels of nesting. It also enables indexing for single field, multi-key, compound, text, and geospatial fields using the PostGIS extension. The engine also supports vector search queries via the pg_vector extension for generative AI and RAG patterns. This technology powers the vCore-based Azure Cosmos DB for MongoDB. You can use existing MongoDB drivers and tools, such as mongosh or Compass, with this setup.

While FerretDB provides a simple interface for developers through the MongoDB wire protocol, the underlying architecture requires translating every incoming query into SQL, which introduces latency during heavy write or aggregation workloads that rely on intense document processing. This translation overhead makes the tool less efficient than native MongoDB for write-heavy tasks.

Comparing managed cloud tiers

FerretDB Cloud targets developers who want an alternative to MongoDB Atlas. While it currently runs on AWS, support for Azure and Google Cloud remains in development. I find the free tier insufficient for production because it lacks TLS and encryption. You should know that managing your own PostgreSQL infrastructure requires patching and scaling that FerretDB Cloud handles for you. The free tier provides 10 Gi of storage and a 99.90% SLA.

The product offers four different subscription plans, including a "Bring Your Own Account" enterprise option. The enterprise tier provides SOC 2 readiness and a 99.99% SLA. Paid tiers offer up to 64 Ti of storage and dedicated tenancy. Subscription for some plans requires approval from the vendor. FerretDB Cloud provides the same feature set as a local FerretDB instance. Free cloud instances are explicitly temporary and may be deleted when inactive.

Migration and legal risks

Migration requires careful planning across three distinct phases. First, you must verify query compatibility in a test environment. Second, you use mongodump or mongoexport to pull data, then use mongorestore or mongoimport to push it to FerretDB. Finally, you update your connection strings to point to the new instance. I recommend testing your application extensively before you touch production data. FerretDB acts as a stateless proxy that targets roughly 80% of MongoDB workloads but lacks support for aggregation pipelines, change streams, and transactions. For those who rely on earlier versions and backends, FerretDB 1.x remains available on GitHub.

The legal conflict between MongoDB and FerretDB remains a headache. Andrew Stephens, general counsel at MongoDB, claims FerretDB uses "imitation, theft, and misappropriation masquerading as compatibility" to attract users. This conflict makes the tool a risky choice for companies sensitive to intellectual property disputes. FerretDB co-founder Peter Farkas argues that the SSPL license creates market restrictions that prevent real competition. Do you trust a tool that faces ongoing patent litigation from its largest competitor?