Summary
NoSql databases have fast write throughput and fast lookup on primary key. And they also have good scalability and reliability.
However, in order to search non-primary attributes, secondary index should be maintained and this could hurts write performance.
This paper compares two types of secondary index:
Embedded Indexes: lightweight filters embedded inside the primary table
Stand-Alone Indexes: separate data structure
by implementing 2 Embedded Indexes and 3 Stand-Alone Indexes on top of LevelDB.
They conclude that : the embedded indexes offer superior write throughput and are more space efficient, whereas the stand-alone secondary indexes achieve faster query response times. So the optimal choice depends on the application workload.

