I am actually can not understand the real advatage or killer feature of MongoDB storage.
As I understand currently MongoDB implements exclusive per-database write lock that is even worse from the concurrency point of view than per-table write locks of MySQL's MyISAM storage engine.
Also MongoDB uses memory mapped files which are mapped to virtual memory (not physical RAM) and flush data to disk every second (without journaling). With write ahead log (journal) the data is also not durable because group commits are performed in certain time intervals. So we receive not durable and not in-memory storage at the same time. (There was a time when Redis also used swapping but they wisely removed this feature since 2.5 version)
So considering the above, what is the point of using MongoDB in production (especially highload) applications?