I have a use case which includes a system that can trace data, to explain the case the system has to intercept each request and check if that request is already exist to do some map/reduce on the body, it suppose to be real time. I'm sorry that I can't elaborate more due to the sensitivity of the system I'm working on.
I'm focusing mainly on NoSql key/value store, so I'm thinking of Rocksdb / Riak. My question is does it fit for this use case or there is a better alternative?
First lets say that Riak and Rocksdb both are key/value storage but Riak is used as a distributed and highly available database but Rocksdb is an embedded database solution. Also neither of them is suitable for doing map reduce in real-time.
In the Hadoop ecosystem there are solutions which can be used for doing map reduce in real-time:
Riak with bitcask can be very fast in concepts of IO. If you want to speed up the process you can add some cache layer (redis for example), before you go with your key to riak.
But do you want to check either the value exists or not, and then, if not exists perform some map/reduce task by your site or it should be part of the key value store solution?
