The Database and Caching

With relational databases, object caching is usually a separate facility (such as memcached), which makes sense as even a RAM page cache hit is a fairly expensive operation with a relational database (joins may be required, and the data must be transformed into an object representation). Further, memcached type solutions are more scaleable than a relational database.

Mongo eliminates the need (in some cases) for a separate object caching layer. Queries that result in file system RAM cache hits are very fast as the object's representation in the database is very close to its representation in application memory. Also, the MongoDB can scale to (almost) any level and provides an object cache and database integrated together, which is very helpful as there is no risk of retrieving stale data from the cache. In addition, the complex queries a full DBMS provides are also possible.