This document discusses migrating data from MongoDB to Cassandra for the Atlas metadata platform. Some key points:
- MongoDB was becoming too simple and lacked features like control and sharding needed for Atlas' large scale. Cassandra offers better scalability, performance, and consistency controls.
- The data model is being adapted to Cassandra's column-based model with secondary indexes to support annotations. Unique IDs will be generated during ingest rather than using MongoDB's findAndModify().
- Ops processes like deployment, monitoring, and disaster recovery require adjustments for Cassandra's distributed architecture and use of tokens but integrate well with tools like Puppet, OpsCenter, and AWS.
- Overall the migration is progress