Blocking protocol for concurrency control

What is a blockade agreement

Using X Lock and S When locking data objects , Some rules need to be agreed , These rules are blocking agreements （Locking Protocol）.
When to apply for X Lock or S lock
Holding time
When to release
There are different rules for the way of blockade , There are different kinds of blockade agreements , They provide a certain guarantee for the correct scheduling of concurrent operations to varying degrees .

Three level blockade agreement

1. First level blockade protocol
2. Secondary blockade protocol
3. Three level blockade agreement

First level blockade protocol

First level blockade protocol

Business T Modifying data R It must be added before X lock , Not released until the end of the transaction .
Normal end （COMMIT）
Abnormal end （ROLLBACK）
The first level blocking protocol can prevent the loss of modification , And make sure that things T It's recoverable .
In the first level blockade agreement , If you just read the data and don't modify it , It doesn't need to be locked , So it's not guaranteed to be repeatable or unreadable “ dirty ” data .

Use blocking mechanism to solve the problem of lost modification

No changes lost
Business T1 reading A Make changes to A Add X lock
When T2 Ask again for A Add X Lock rejected
T2 Can only wait for T1 Release A Get the right after locking on A Of X lock
At this time T2 Read about A It's already T1 Updated value 15
T2 Click here to the new A Value is calculated , And the result value A=14 Write back to disk . Avoid losing T1 Update .

Secondary blockade protocol

Secondary blockade protocol

First level blocking protocol plus transactions T Reading data R It must be added before S lock , After reading, you can release S lock .
The secondary blocking protocol can prevent loss, modification and reading “ dirty ” data .
In the secondary blockade agreement , Because the data can be released after reading S lock , So it's not guaranteed to be repeatable .

Use the blocking mechanism to solve the read problem “ dirty ” Data problems

Do not read “ dirty ” data
Business T1 In the face of C Before making changes , First pair C Add X lock , Modify its value and write it back to disk
T2 Request in C add S lock , because T1 Already in C Added X lock ,T2 Can only wait for
T1 Be revoked for some reason ,C Restore to original value 100
T1 Release C Upper X After locking T2 get C Upper S lock , read C=100. Avoided T2 read “ dirty ” data

Three level blockade agreement

Three level blockade agreement

First level blocking protocol plus transactions T Reading data R It must be added before S lock , Not released until the end of the transaction .
The three-level blocking protocol can prevent loss and modification 、 Read dirty data and non repeatable data .

Use blocking mechanism to solve the problem of non repeatable reading

Repeatable
Business T1 reading A,B Before , First pair A,B Add S lock
Other affairs can only be right again A,B Add S lock , Instead of X lock , That is, other transactions can only read A,B, It can't be modified
When T2 To modify B And the application is right B Of X When the lock is rejected, you can only wait T1 Release B The lock on the
T1 Read again for checking A,B, Now read B Is still 100, The sum result is still 150, Can be read repeatedly
T1 Release at the end A,B Upper S lock .T2 To get the right B Of X lock

Summary of blockade agreement

The main differences between three-level agreements

What operations need to be blocked and when to release the lock （ That is, lock holding time ）

Different blocking protocols make transactions reach different consistency levels

The higher the level of blockade agreement , The more consistent

Welcome to join me for wechat exchange and discussion （ Please note csdn Add ）