Why does NewSQL eclipse the traditional relational database?

Traditional databases will still have a place. As for the advantages of NewsSQL, let's simply tell you:

First of all, there is a paper Pavlo-NewsSQL-Sigmodrec abroad about whether "middleware+relational database sub-database sub-table" is a distributed database of NewsSQL. According to the classification in this paper, Spanner, TiDB and OB are the first new architecture types. Sharding-Sphere, Mycat, DRDS and other middleware solutions are considered as the second (there is also a third cloud database in this paper, which will not be introduced in detail for the time being).

is the middleware (including SDK and Proxy)+traditional relational database (sub-database and sub-table) model a distributed architecture? I think so, because storage is indeed distributed and can scale out. But not a "pseudo" distributed database? From the perspective of the advanced structure, this also makes sense.

"Pseudo" is mainly reflected in the SQL parsing and execution plan generation of the middleware layer and the underlying DB, and the storage engine is based on B+Tree, which is actually redundant and inefficient in the distributed database architecture. In order to avoid causing a war of words between authenticity and distributed database, the NEWSSQL database in this paper refers to this new architecture NEWSSQL database.

what is the advantage of p>NewSQL database compared with middleware+sub-database and sub-table? Draw a simple architecture comparison diagram: the traditional database is designed for disk, and the storage management and concurrency control based on memory are not as efficient as NewSQL database; SQL parsing and execution plan optimization of middleware mode are repeated in middleware and database, which is inefficient. Compared with XA, the distributed transaction of NewSQL database is optimized with higher performance. The new architecture of NewSQL database storage design is multi-copy based on paxos (or Raft) protocol. Compared with the traditional database master-slave mode (there is also the problem of losing data after semi-synchronous to asynchronous), it has achieved real high availability and high reliability (RTO <: 3s，RPO=）； NewSQL database naturally supports data fragmentation, and the migration and expansion of data are automatic, which greatly reduces the work of DBA, and is transparent to applications, and there is no need to specify database and table keys in SQL.