Analysis of new mechanism--nosql mode in devouring large data storage domain

Over the past few years, a burgeoning large data-storage mechanism is gobbling up large data-storage markets. This storage solution differs significantly from the traditional RDBMS, which is called NoSQL.

The following key members are in the NoSQL world, including

Google BigTable, HBase, hypertable

Amazon Dynamo, Voldemort, Cassendra, Riak

Redis

CouchDB, MongoDB

And these solutions have some common features

Based on key-value storage

The system runs on a mass of ordinary machines.

Data is distributed in clusters after partitioning and replication

Loosen the requirement for data consistency (because of the cap theorem).

The important criteria for choosing NoSQL are the CAP (consistency, availability, and partition tolerance), which is what we call consistency, availability, and partitioning tolerance. But the CAP principle requires that only two of the consistency, availability, and partition tolerance can be selected in a distributed system.

This article is designed to extract the common technology behind these solutions in order to understand more deeply the meaning of application design. This article does not compare these solutions and does not recommend the use of a particular product.

The underlying data model can be viewed as a large hashtable (key/value store)

API Access Basic form:

Get (Key): extracts the corresponding value of the given key

Put (Key,value): Creates or updates a corresponding value for a given key

Delete (key): Delete keys and their associated values

Perform user-defined functions in a server environment with a more advanced API

Execute (Key, twist, parameters): Invokes a value that corresponds to the given key of the operation, with special data (such as list, Set, map, etc.)

MapReduce (Keylist, Mapfunc, Reducefunc): Key calls to Range MapReduce

The underlying infrastructure consists of a large number of (hundreds) of inexpensive, common, unreliable machines that are made up of networks. Each machine is a separate physical node (PN). Each PN software configuration is the same, but the CPU, memory, hard drive, and so on will be different. Each PN runs a different number of virtual nodes based on different hardware configurations.

(Responsible editor: The good of the Legacy)