0

OrientDB documentation here:

http://orientdb.com/docs/last/Concurrency.html#atomic-operations http://orientdb.com/docs/last/Distributed-Architecture.html#distributed-transactions

States that OrientDB implements MVCC. But does OrientDB actually keep multiple versions of records around? I don't see any documentation about how timestamps are chosen, how and when past versions of records are cleaned up, how any of this works in a distributed context, and whether or not consistent scans of data are supported across the cluster.

Could someone please clarify this?

Jonathan Ellithorpe
  • 564
  • 1
  • 4
  • 17

1 Answers1

4

I think what we call MVCC in OrientDB (I work for OrientDB) is widely known as optimistic locking. We work on a new transactional model which will support MVCC in the same way as InnoDB and probably PostgreSQL do. We will keep versions of data to allow writes do not block reads but they will be discarded once transactions which read them will be committed. But we have no plans to keep long-term versions of data.

Andrey Lomakin
  • 637
  • 3
  • 12