Shulou(Shulou.com)05/31 Report--

Today, I will show you how to analyze CAP theory. The content of the article is good. Now I would like to share it with you. Friends who feel in need can understand it. I hope it will be helpful to you. Let's read it along with the editor's ideas.

CAP theory is an important theory in distributed system design. Although it provides a very useful basis for system design, it also brings a lot of misunderstandings. This paper will start with the background of the birth of CAP, then explain the theory, and finally analyze some new understandings of CAP in the current context to clarify some misunderstandings about CAP.

Background of the birth of CAP Theory

The theory of CAP arises from the debate of "data consistency VS availability". Brewer, the author of CAP, began to study cluster-based cross-regional systems (essentially early cloud computing) in the 1990s. For such systems, system availability is the primary goal, so they use caching or post-event updates to optimize system availability. Although these methods improve the availability of the system, they sacrifice the data consistency of the system.

Brewer put forward the theory of BASE (basic availability, soft state, ultimate consistency) in the 1990s, which was not accepted at that time. Because people still value the advantages of ACID, do not want to give up strong consistency. Therefore, Brewer put forward the CAP theory, the purpose is to broaden the design space of distributed systems, through the "three choices of two" formula, emancipate the mind, do not just focus on consistency.

Understanding the background of the birth of CAP, we can have a more in-depth understanding of CAP theory and its enlightenment. Although the view of "three choices and two" helps us to open up design ideas, it also brings a lot of misunderstandings. Next we will analyze them one by one. First, let's take a look at the explanation of CAP theory.

The Classical explanation of CAP Theory

CAP theorem is the most basic and critical theory in distributed system design. It points out that distributed data storage cannot meet the following three conditions at the same time.

Consistency: each read either gets the most recently written data or gets an error.

Availability (Availability): you get a (non-error) response per request, but there is no guarantee that the most recently written data will be returned.

Partition tolerance (Partition tolerance): although any number of messages are lost (or delayed) by the network between nodes, the system continues to run.

CAP's theorem states that in the case of network partitions, consistency and availability must be one of the two. When a network partition occurs (the network between different nodes fails or has a large delay), there is either loss of consistency (allowing data writing from different partitions) or loss of availability (service is stopped when the network partition is identified). When there is no network failure, that is, when the distributed system is running normally, consistency and availability can be satisfied at the same time. It should be noted here that the consistency in CAP's theorem is quite different from that in ACID database transactions. ACID C means that a transaction cannot break any database rules, such as the uniqueness of a key. By contrast, CAP's C only refers to consistency in the sense of a single copy, so it is only a strict subset of ACID consistency constraints.

CAP theory seems difficult to understand, but it can be deduced by grasping a core point without having to memorize it. When there is a network partition

If the system does not allow writes, it means that the availability of the system is reduced, but the data of different partitions can be consistent, that is, consistency is selected.

If the system allows writing, it means that the data between different partitions are inconsistent and the system availability is guaranteed, that is, selective availability.

New understanding of CAP

CAP is often misunderstood, largely because the scope of usability and consistency is often ambiguous when discussing CAP. If we do not first define the concepts of usability, consistency, and partition tolerance in specific scenarios, CAP will actually restrict the thinking of system design. First of all, since partitions rarely occur, there is no reason to sacrifice C or An if there are no partitions in the system. Secondly, the trade-off between C and A can occur repeatedly with very fine granularity in the same system, and each decision may be different because of the specific operation, or even because it involves specific data or users. Finally, all three properties can be measured to a certain extent, not black or white. Usability obviously varies continuously from 0% to 100%, consistency is divided into many levels, and even partitions can be subdivided into different meanings, for example, different parts of the system can have different perceptions of the existence of partitions.

What is zonal tolerance

In the real world, under normal circumstances, the communication between the nodes of the distributed system is reliable, and there will be no message loss or high delay, but the network is unreliable. There will always be occasional message loss or high message delay. At this time, nodes in different regions will be unable to communicate within a period of time, that is, partitioning occurs.

Partition tolerance means that a distributed system can continue to run and provide services when there is a network partition. Note that the ability to provide services here is different from the requirements of availability, which requires that any request can be responded to, which means that all nodes can provide services even if there is a network partition. The focus of partition tolerance is that after the emergence of network partitions, the system is still available (including partially available).

For example: a system that uses Paxos for data replication is a typical CP system. Even if there is a network partition, the primary partition can provide services, so it is partition tolerant. Another counterexample: systems that use 2PC for data replication do not have the feature of partition tolerance, and when network partitions occur, the whole system will block.

Scope of availability

Usability is intuitive: you get a (non-error) response for each request, but there is no guarantee that the most recently written data will be returned. In other words, for each node in a distributed system, it can respond to external requests, but does not require consistency.

The question that often puzzles us is what are the criteria for measuring system availability? In fact, the key point is the scope of usability, which is meaningless without the scope of usability in a specific scenario. When discussing usability, there should be specific scenarios to demarcate the boundary, but it is not rigorous to simply think that an algorithm meets the usability requirements, because there will be a lot of skills to make up for the correction in the engineering implementation.

For example: Google Docs is a typical AP system that can be used even when the network is down. The trick is that it goes into offline mode when it finds that the network is down, allowing users to continue editing and then merging the modified content after the network is restored. It can be found that for Google Docs, the user's browser is also a node of its system. When there is a network partition, it can still provide services to the user, but at the cost of giving up consistency, because the changes made by the user are known only locally. The server is not clear. So in this example, the scope of usability includes the user's browser, and the node of the distributed system that we generally understand must be the server's machine.

It is worth noting that in the real world, we generally do not pursue perfect availability, so the general saying is high availability, that is, to ensure the availability of as many node services as possible. This is one of the reasons why consistency algorithms like Paxos are becoming more and more popular.

Scope of consistency

When discussing consistency, we must make clear the scope of consistency, that is, the state is consistent within a certain boundary, and it is impossible to talk about consistency beyond the boundary. For example, when a network partition occurs in Paxos, complete consistency and availability can be guaranteed within a primary partition, while services outside the partition are not available. It is worth noting that when the system chooses consistency, that is, CP, when partitioning, it does not mean a complete loss of availability, depending on the implementation of the consistency algorithm. For example, the standard two-phase commit is completely unavailable when partitioning occurs, while Paxos ensures the consistency and availability of the primary partition.

After the above discussion, we can find that the scope requirements of usability are more stringent than the scope requirements of consistency. In CAP theory, usability requires the availability of the whole system, even if some nodes are unavailable, it is a violation of availability constraints. On the other hand, the requirement of consistency is not so high. When the network partition occurs, as long as the data consistency of the main partition is ensured, it is also considered that the system is consistent with the consistency constraint. Why would you say that? Because when there is a network partition, the client can get the latest value only by accessing the primary partition (accessing more than half of the nodes, if the values are all the same, it means that the accessed data is up-to-date). At this time, the system meets the requirements of consistency in CAP theory.

Manage Partition

Network partitioning is an inevitable occurrence in distributed systems. The classical CAP theory ignores network delay, but in the real world, network delay is closely related to partitioning. In other words, when the system is unable to reach an agreement in a limited time (high network delay), it means that partitioning has occurred. At this point, you need to choose between consistency and availability: choosing to continue to retry means choosing consistency and abandoning usability; abandoning data consistency to let the operation complete means that usability is selected. It is worth noting that abandoning data consistency when partitioning does not mean that it does not matter at all. general engineering implementations will use retries to achieve final consistency.

Through the above analysis, it can be found that balancing the impact of availability and consistency during partitioning is a key issue in distributed system design. Therefore, managing partitions not only requires proactive discovery of partitions, but also needs to prepare the recovery process for the impact that occurs during the partition. In other words, we can apply CAP theory from another perspective: how to choose between consistency and availability when the system enters partition mode.

There are three steps to managing partitions:

Partition start detected

Explicitly enter partition mode and restrict certain operations

Start the partition recovery process when communication is restored

When the system enters partition mode, there are two options:

Select consistency: for example, Paxos algorithm, only most of the primary partition can operate, other partitions are not available, when the network is restored, a small number of nodes synchronize data with most nodes.

Select availability: for example, Google Docs, enter offline mode when there is a partition, and other networks restore client and server data for merge recovery.

The above is the whole content of how to analyze CAP theory, more content related to how to analyze CAP theory can search the previous articles or browse the following articles to learn ha! I believe the editor will add more knowledge to you. I hope you can support it!

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.