Newsql Early understanding

The term newsql is presented in a research paper by 451 Group's analyst Matthew Aslett. It refers to a new type of database system that challenges old database vendors. Newsql is the abbreviation for a variety of new scalable/high-performance databases that not only have NoSQL storage management capabilities for massive amounts of data, but also maintain traditional database support for acid and SQL features.

Michael Stonebraker, chief technology officer at VOLTDB, a distributed database company, says NoSQL databases offer good scalability and flexibility, but they also have their own shortcomings. Because the Sql,nosql database system is not used, it does not have the characteristics of highly structured queries. NoSQL Other issues include actions that do not provide acid (atomicity, consistency, isolation, and durability). In addition, different NoSQL databases have their own query language, which makes it difficult to standardize application interfaces. Stonebraker indicates that the lag of a database system can often be attributed to a number of factors. such as the buffer pool maintained by the database system for the purpose of recovery logs, and the data fields that manage locks and locks. These behaviors were found to consume 96% of the resources of the system in the VOLTDB test.

While the internal structure of the Newsql system varies greatly, there are two notable common features: (1) They all support relational data models, and (2) they all use SQL as their primary interface. The first known newsql system, called H-store, is a distributed parallel memory database system. At present, the Newsql system is broadly divided into three categories:

New architecture

The first type of newsql system is a completely new database platform, and they all adopt different design methods. They are roughly divided into two categories:

(1) This type of database works on a distributed cluster node, where each node has a subset of data. The SQL query is executed on the node where the query fragment is sent to its own data. These databases can be scaled linearly by adding additional nodes. Existing databases of this type are: Google spanner, Voltdb, Clustrix, NuoDB.

(2) These database systems usually have a single primary node data source. They have a set of nodes for transaction processing, which, after receiving a specific SQL query, takes all the data it needs from the primary node and executes the SQL query and returns the results.

SQL engine

The second category is the highly optimized SQL storage engine. These systems provide the same programming interface as MySQL, but the scalability is better than the built-in engine InnoDB. This kind of database system has: Tokudb, Memsql.

Transparent fragmentation

This kind of system provides the slicing middleware layer, the database automatic segmentation is running at multiple nodes. This kind of database package expands: Scalebase,dbshards, Scalearc.

So what are the existing newsql system vendors:

We will include (sequentially random) Clustrix, Geniedb, Scalarc, schooner, Voltdb, Rethinkdb, Scaledb, Akiban, CodeFutures, Scalebase, Translattice and Nimbusdb, as well as drizzle, a MySQL cluster with NDB and MySQL with Handlersocket. The latter includes Tokutek and Justone DB. The related "Newsql as a service" category includes Amazon relational database services, Microsoft Sqlazure,xeround and Fathomdb.