multi-threaded concurrent import question

[dylan-tao]

Describe the feature:
Backend storage: Cassandra, multi-threaded concurrent import In addition to transactions and unique indexes to ensure data consistency, retry the import with transaction rollback, what other better way, the current performance is too low.

Describe a specific use case for the feature:
When importing data, enable multi-threaded concurrent writes, and the uniqueness of a node may have multiple inserts at the same time. The current solution is to start the transaction or commit or roll back as a whole. Execute the query first when inserting, and associate with the existing node if it exists, and insert if it does not exist. Since the node is diffuse derived, it is not possible to group and isolate the data through the associated node and then perform multi-threaded safe uniqueness import.