1. Complex query services for JSON nested documents and arrays
2. The amount of data on the proposed business layer also do a horizontal split, relying entirely on MongoDB own cluster mechanism may have performance issues (Baidu op Summary)
3. Visual write query performance is inferior to MySQL (even if writing w=0 is not as good as MySQL)
4. Memory consumption is serious, SSD effect is not obvious
5. Querying indexes is important (this is true for all db)
6. Easy to add fields and indexes online (mysql5.6 already supports online DDL or some tools)
7. No transaction, no Line lock, no join
Project background
Commodity NoSQL platform, data volume 20 million (a machine to do, because the resource application does not come down so did not do the cluster, and do the time is not the core business), visual complex query around 400ms can respond
MongoDB Usage Summary