Relational Database V. S. Non-relational database, relational database v. s

Relational Database V. S. Non-relational database, relational database v. s
The most important feature of a relational database is transaction consistency: The read and write operations of a traditional relational database are transactions and ACID-based, this feature allows relational databases to be used in almost all systems with consistency requirements, such as typical banking systems.

However, in Web applications, especially SNS, consistency is not so important, user A's content is tolerable when user B sees updates of the same user C content, or when two people see updates of the same friend's data, the time difference is several seconds, therefore, the biggest feature of relational databases is useless, at least not that important.

On the contrary, the huge cost paid by relational databases to maintain consistency is poor read/write performance. SNS such as Weibo and facebook have extremely high requirements on concurrent read/write capabilities, relational databases are no longer able to cope with this situation. (In reading, traditionally, to overcome the defects of relational databases and improve performance, a level-1 memcache is added to the static Web page. In SNS, the changes are too fast, memchache is powerless.) Therefore, a new data structure storage must be used to replace relational databases.

Another feature of relational databases is that they have a fixed table structure. Therefore, their scalability is very poor. In SNS, system upgrades and feature increases often mean huge changes in data structures, this is also difficult for relational databases and requires new structured data storage.

As a result, non-relational databases emerged. Since it is impossible to use a structured data storage to meet all new requirements, non-relational databases are not strictly a type of database, it should be a collection of structured data storage methods.

It must be emphasized that the persistent storage of data, especially the persistent storage of massive data, still requires a relational database veteran.