I was reading about noSQL and relational DBMSs. After going through many posts, I am not able to find out exactly how noSQL systems perform horizontal scaling and how it is more efficient than RDBMS scaling.
After all, RDBMSs also do replication and sharding.
How this is fast in case of NoSql?
Wiki relational DBMSs. They offer generic querying and integrity with a certain computational speed/storage cost via modeling in terms of relations--tables of rows that make some true statement about the situation--generally plus concurrent and distributed atomic transactions. Read Wikipedia re NoSQL. This is a catch-all term to refer to systems that "scale" in various ways by the foregoing of and/or specializing in assorted functionality offered by current RDBMS interfaces and implementations, depending on the particular NoSQL system (or, given some shared feature, class of system). Since "relational" is about characterizing and querying state as tables and not about particular performance or other properties, "NoSQL" is actually a misconceived misnomer.
First of all, there is no technology or product called "NoSQL." NoSQL is a broad category describing dozens of different technologies, which each work differently. Please don't use the term "NoSQL" as if it's a single implementation.
Most NoSQL products achieve some scalability improvements by sacrificing the ACID features that are provided by default by RDBMS.
