问题
We have different products that rely on relational databases for various reasons, basically the transactional nature of the operations (atomicity, consistency, etc.). This is not going to change any time soon.
Given this scenario, are there any possible justifications to export the data to a NoSQL solution? Maybe Datawarehousing, Analytics, etc.
Any comments are welcome.
回答1:
"Data" is just a vague generality without a data structure. "Relational" means the data structure is relations/tables with generic queries. (Not automatically interleaved execution of semantically atomic concurrent transactions, which just happens to become feasible given relational set-at-a-time operators.) Need for specialized data structures or queries suggests augmenting a relational system interface or implementation. "NoSQL" is a catch-all for all sorts of such systems with special-case abstractions and/or implementations. But they don't need to be non-relational in interface. They just need to be not what current "relational" DBMSs provide. Given that they exist, one can reasonably export to (or import from) them when needed.
Unfortunately the relational model is not generally understood (including by "relational" DBMSs), so instead these systems reject what they could simply extend.
来源：https://stackoverflow.com/questions/44313877/reasonable-export-of-relational-to-non-relational-data