I. Introduction
-
CAP理论
: A distributed system can not meet the consistency (Consistence), availability (Availability), partition fault tolerance (Partition tolerance), can only meet two (CA / CP / AP). -
Consistency (Consistence): After a write operation is complete, any read can obtain the latest value of the write operation is written;
-
Availability (Availability): non-faulty nodes within a reasonable period of time to make the response is reasonable;
-
Partition fault tolerance (Partition tolerance): a node or network partitions in a distributed system when a fault occurs, the entire system can still provide to meet the consistency and availability of services.
-
BASE理论
: System requirements for relaxation strong consistency of data at a time, allowing the system to achieve eventual consistency. -
Basic available (Basically Available): a system failure occurs, but still can be used;
-
Soft state (Soft state): multiple copies of data of different nodes of the system allows the presence of delay in the data;
-
The final consistency (Eventually consistent): data finally be able to achieve a consistent state.
二. Students, Cf, of
- CA: single-point cluster, meet consistency, system availability, scalability poor. Such as the traditional Oracle database;
- CP: meet consistency, fault tolerance partition systems, performance is not particularly high. As the Redis, Mongodb ;
- AP: meet availability, fault-tolerant system partition, a low consistency requirements. As most of the site architecture.