01 distributed system overview

Catalog

One , What is a distributed system

1, Concept

2, characteristic

3, Typical problems

Two ,CAP Theorem

1,C、A、P

2,CAP Theorem

3、 ... and ,BASE theory

1,BA、S、EC

Two ,BASE theory

---

The content of the article comes from ： Nanjing University / Star ring technology course , Big data theory and practice course Ⅰ

One , What is a distributed system

1, Concept

Hardware or software components （ service ） Distributed on different network computers , adopt The messaging Communicate and coordinate .

2, characteristic

Distribution

Equivalence

• equality ： There is no subject or subordinate
• Independent ： Own your own CPU And memory , Process data independently

concurrency

• external ： Hosting concurrent access from multiple clients
• Inside ： Homework （Job） Broken down into multiple tasks （Task）, Running concurrently on different nodes

Fault independence

• The failure of some nodes does not affect the normal use of the whole system

3, Typical problems

Abnormal communication

Network partition （ Split brain ）

• The system is split into two or more local clusters （ Partition ）
• Each partition operates independently , Simultaneous provision of services , Which leads to chaos

Node failure

• Down or dead

Tristate

• success 、 Failure and Overtime

Two ,CAP Theorem

1,C、A、P

C（Consistency, Uniformity ）

• meaning ： At the same time , The data is in Whether multiple copies of different nodes have identical values
• Strong consistency ： After data update , At the same time , Different read operations can get the latest value
• Weak consistency ： After data update , At the same time , Different read operations may not always get the latest value , There is no guarantee how long it will take to get the latest value

A（Availability, Usability ）

• meaning ： For each request , Whether the system can be in a limited （ Appoint ） To respond in a timely manner

P（Partition Tolerance, Partition tolerance ）

• meaning ： When a network partition occurs , The system can still provide external services to meet the consistency C And usability A Service for

2,CAP Theorem

describe ： Distributed systems in the same time segment , It's impossible to meet consistency at the same time C、 Usability A And partition fault tolerance P, Only two of them... Can be satisfied at most .

For distributed systems , Network partitioning cannot be avoided , Satisfy P The premise is the premise , So it is impossible to choose CA framework , Can only choose CP or AP framework .

for example ： When a network partition occurs , A node is writing

（1） If in order to guarantee C, Read and write operations on other nodes must be disabled , Then with A The conflict

（2） If in order to guarantee A, Other nodes can read and write normally , Then with C The conflict

choice CP or AP framework , The key is the business scenario . for example ： For banking businesses that must ensure strong consistency , Can only choose CP

3、 ... and ,BASE theory

1,BA、S、EC

BA（Basically Availability, Basic usability ）

• When the system fails , Ensure that the core functions and indicators are effective in advance , Partial loss of availability is allowed , Including the loss of response time 、 Loss of non core functions, etc

S（Soft State, Soft state ）

• Allow data to have intermediate states （ Not updated temporarily ）, And this status will not affect the overall availability
• Allow a certain delay in the synchronization of data copies on different nodes

EC（Eventually Consistency, Final consistency ）

• Copies of data distributed on different nodes , After a certain time of synchronization , To reach a consensus . for example ：Zookeeper、HDFS QJM Write more than half of the transaction strategy
• An upgraded version of weak consistency

Two ,BASE theory

describe ： The distributed system satisfies the partition fault tolerance P At the same time , Allow data soft state S The existence of , And implement basic Usability BA and Final Uniformity EC.

That is to satisfy P Under the premise of , Yes CAP Strong consistency in A And absolute availability C Make moderate compromises ：A-》BA、C-》EC

By tolerating temporary inconsistencies in some data （ Soft state ）, That is, sacrificing the strong consistency of data （ Ensure final consistency ）, To ensure that the core functions and indicators of the system are effective （ Basic available ）

BASE Theory is a summary of the distributed practice of large-scale Internet systems , It's right CAP The result of a trade-off between consistency and availability in .