Distributed cap theory

CAP The theory is distributed system 、 In particular, the most discussed theory in the field of distributed storage . among C For consistency (Consistency),A For usability (Availability),P Represents partition fault tolerance (Partition tolerance).CAP Theory tells us C、A、P The three cannot be satisfied at the same time , Only two of them can be satisfied at most .

CAP A brief introduction to the theory

CAP The theory is distributed system 、 In particular, the most discussed theory in the field of distributed storage . among C For consistency (Consistency),A For usability (Availability),P Represents partition fault tolerance (Partition tolerance).CAP Theory tells us C、A、P The three cannot be satisfied at the same time , Only two of them can be satisfied at most .

CAP Two out of three

• Uniformity (Consistency): A write operation returns success , Then all subsequent read requests must read this new data ; If the return fails , Then all read operations cannot read this data . All nodes access the same latest data .
• Usability (Availability): High availability for data updates , Requests can be processed in a timely manner , Not always waiting for , Even if a node fails .
• Partition tolerance (Partition tolerance): Can tolerate network partition , When the network is disconnected , The separated nodes can still provide services to the outside world .

Yes CAP Theoretical understanding

understand CAP The simplest theory is to imagine two replicas on either side of the partition , That is, the network between the two replicas is disconnected , Can't communicate .

• If one copy is allowed to update , It will lead to inconsistent data , That is, to lose C nature .
• If in order to ensure consistency , Set the replica on one side of the partition to unavailable , So lost again A nature .
• Unless two copies can communicate with each other , To guarantee C And promise A, This in turn leads to the loss of P nature .

Generally speaking, distributed systems using network communication , Can't give up P nature , Then we can only make a difficult choice in terms of consistency and availability .

CAP The expression of theory serves its purpose well , It broadens the thinking of distributed system designers , Design a variety of systems under a variety of alternatives . There have been countless new systems in the past decade , As a result, there have been quite a number of debates on the relative relationship between consistency and usability .

CAP In depth theoretical understanding

stay CAP Twelve years after the theory was put forward , The author came out to refute the rumor .“ Two out of three ” There has always been a misleading formula , It oversimplifies the relationship between properties :

• First , Because zoning rarely happens , So there's no reason to sacrifice when there's no partition in the system C or A.
• secondly ,C And A The trade-offs can occur repeatedly in the same system with very small granularity , And every decision may be made because of specific operation , Even because it involves specific data or users .
• Last , All three properties can be measured in degree , It's not black or white with or without . Usability is obviously in 0% To 100% Between successive changes , There are many levels of consistency , Even zoning can be subdivided into different meanings , For example, different parts of the system can have different cognition about whether there is a partition .

So consistency and usability are not incompatible , Either this or that .Paxos、Raft The equally distributed consistency algorithm is the witness of a good balance between consistency and availability .

If you reprint , Please indicate the source ： Open source byte   https://sourcebyte.cn/article/176.html