Berkeley DB in multi-threaded applications

Question

What is the best way to use berkeley DB descriptors (bdb) in a multi-threaded application?

• Is it better for each thread to open its own descriptor; or,

• is it better to open one descriptor and make each thread txn_begin {} txn-> commit ()?

Which one is better? I am using the Transactional Data Store with the DB_THREAD flag.

thanks

Answer 1

Have you read this guide for bdb in a multi-thread environment?