Isn't that defeating the purpose of MondoDB? We're considering switching to it precisely because we're sick of two-tiered approach (MySQL+memcached) and MongoDB promises a built-in near-realtime caching.
Why would I want to lose the power of SQL if I'm not getting anything in return? I thought that "built-in memcached" was what I'd be getting. No?
Storage and caching are two different problems so it's reasonable to assume two separate solutions.
In this case, MongoDB is a leap ahead of MySQL; Redis is a leap ahead of Memcached. Both solve the core problems very well and include additional freebies (MongoDB's high-performance aggregate functions and Redis' atomic increments and set and list functions).
Why would I want to lose the power of SQL if I'm not getting anything in return? I thought that "built-in memcached" was what I'd be getting. No?