High availability with fast failover and retryable writes | Up to 120 second failover, no retryable writes, Global Clusters still route writes to a single primary | Typical failover of 5 seconds with retryable writes |
Support for role-based access control and authentication restrictions | Predefined roles only, no support for LDAP or Active Directory | Granular access control, full auditing, , with integrations to AWS KMS and customers' own key management service |
Multi-statement distributed ACID transactions | Ambiguous commits, poor error handling, small data sizes. Transactions across shards are not available as DocumentDB does not support sharding | Full support |
Native text search, JOINs, geospatial processing, graph traversals | Data must be replicated to multiple adjacent AWS services, driving up cost and complexity | All available from a single API and platform |
Schema governance | Schema controls must be enforced in the app | Optional schema governance and enforcement |
Scale writes and partition data beyond a single node | Single primary only, largest instance supports 30,000 concurrent connections | Largest Atlas instance supports 128,000 concurrent connections |
Replicate and scale beyond a single region | Single primary with up to 15 replicas; Global Clusters still vulnerable to failures of the single primary that can process writes | Global clusters with up to 50 replicas per shard across multiple regions and even multiple cloud providers |
Support for latest MongoDB version | Feature set resembles MongoDB 3.0/3.2, released in 2015 | MongoDB 6.0 |
Fully compatible with MongoDB | Imitation API, fails up to 61% of correctness tests | Yes |
Availability of advanced developer and analysis tools | Limited telemetry: DocumentDB exposes fewer than 50 metrics | MongoDB exports over 100 metrics, in addition to developer aids such as Performance Advisor, which analyzes slow query logs and ranks suggestions by impact to database performance |
Access to MongoDB expertise | DocumentDB is not supported by MongoDB, and there is no escalation path to MongoDB engineers | Full support directly from MongoDB, including escalation to product engineers; deep pool of talent certified on MongoDB |
Native support for time series data | No support | Full support in MongoDB 5.0+ with Time series collections |