[database] must know and be able at the end of the term ----- Chapter 11 concurrency control

Chapter 11 concurrency control

1、 The concept of transaction and ACID nature

Concept ： Is a user-defined sequence of database operations , All or nothing of these operations , It is an indivisible work unit

ACID The nature of ：

A(Atomicity)： Atomicity , Either do it all or don't do it at all

C（Consistency）： Uniformity , Only the results of successful transaction submission , That is, only when it is successfully completed can it be submitted , You won't submit half done

I（Isolation）： Isolation, , One transaction cannot be disturbed by other transactions

D（Durability）： Continuity , Once the transaction is committed , Its change is permanent

2、 Data inconsistency caused by concurrent operations

1） Lost data

2） read “ dirty ” data

3） It can't be read repeatedly

3、 Common blocking techniques for concurrency control

1） Exclusive lock ==>X lock ==> Write lock ： Readable and modifiable

2） Shared lock ==>S lock ==> Read the lock ： Readable but not modifiable

4、 The meaning of blockade 、 Basic blocking types （ Exclusive lock 、 Shared lock ） And its compatibility

meaning ： Business T Before operating on an object , Send a request to the system first , Lock it

type ： Exclusive lock 、 Shared lock

Compatibility ：

1） added X After the lock , Transactions cannot lock them

2） added S After the lock , Transactions can only add S lock

5、 The granularity of the blockade

meaning ： The size of the blocked object is called “ Blocking granularity ”

characteristic ：

1） The larger the particle size , That is, the larger a single object , Then the smaller the data unit that can be blocked , The number of concurrency decreases , Less overhead

2） The smaller the particle size , That is, the smaller a single object , Then the more data units that can be blocked , The number of concurrent increases , Spending more

6、 Know the cause and prevention of deadlock

reason ： Two or more transactions send requests to block mutually blocked resources at the same time , Form a dead wait

Ways to prevent deadlocks ：

1） One time blockade ： It is to lock all the data to be used at one time , Otherwise, do not execute

2） Sequential blockade ： Specify a blocking order for data objects , Everything is blocked in this order