strong data durability guarantees

  • fully fsync'ed writes with group commit
  • data integrity ensured with CRCs at the protocol level
  • synchronous and asynchronous replication
  • online backup

rich semi-structured data model

  • atomic transactions (both read-committed and repeatable-read isolation levels)
  • optimistic and pessimistic concurrency control
  • asynchronous notifications
  • limited support for complex documents (BSON serialized)
  • support for composite keys (REPL and client lib)

performance

  • fast random writes
  • efficient range queries thanks to spatial locality
  • cross-record redundancy reduction at the page level (snappy compression)
  • fast recovery (independent of dataset size)
Read more →