Distributed environment Features :
Distributive
Concurrency - the program runs concurrent operation is very common, such as uniform resource simultaneously access the same multiple nodes in a distributed system such as databases, distributed storage -
Disorder - message communication between processes in order of appearance will be inconsistencies -
Distributed under existing environmental problems:
Network communication problem - Network itself is not reliable
Problems network partition (split brain) - due to abnormalities occurring when the network between the network nodes in the distributed system portion of the delay is increasing, resulting in all the nodes of the distributed architecture only some of the nodes can communicate. eg for example, now has a controlling five redis from, if there are three anomalies which can not emerge from the communications network, in which it is possible to select one of three new master. So the question comes, there will be two masters. Such data will duplicate the problem (the two clusters).
Tri-state: the state will continue to exist in addition to time-out success or failure in a distributed architecture.
Distributed transaction (atomic consistency isolation, durability)
Centralization and decentralization:
Centralization: the equivalent of only a leader, but the leader if sick how to do, there is generally used standby program includes cold standby and hot standby
Cold standby: eg only one leader, the leader is lost, people will be inactive activation
Hot Standby: eg a two leaders ready to work a
Decentralization: micro bus service removed
Distributed architecture is used in many architectural thinking: When a cluster fails in the cluster will automatically heat to elect a new leadership, the most typical: zookeeper / etcd
CAP / BASE theory