Mutexes and spinlocks

• The mutex After the lock failed , Threads will Release CPU , To other threads ;

• spinlocks After the lock failed , Threads will Busy waiting , Until it gets the lock ;

The mutex , It's a kind of 「 An exclusive lock 」, When locking fails , From user mode to kernel state , Let the kernel switch threads for us ;

What is the context of thread switching ？

When two threads belong to the same process , Because virtual memory is shared , So when switching , Virtual memory, these resources remain stationary , Only the private data of the thread needs to be switched 、 Data that is not shared, such as registers

spinlocks , It's through CPU Provided CAS function （Compare And Swap）, stay 「 User mode 」 Complete locking and unlocking operations , No active thread context switch is generated , So compared to mutexes , It's going to be faster , It's also less expensive .

Mutexes and spinlocks are the most basic locks , Read write lock can choose one of these two locks according to the scene .

If you can be sure that the locked code has a short execution time , You shouldn't use mutexes , Instead, spin locks should be used , Otherwise use mutex .

Optimistic lock and pessimistic lock

Concept

Optimism lock ： Optimistic lock is very optimistic when operating data , Think other people won't change the data at the same time . So optimistic locks don't lock , It's just to judge whether someone else has modified the data during the update ： If someone else modifies the data, they give up the operation , Otherwise, perform the operation .

Pessimistic locking ： Pessimistic lock is more pessimistic when operating data , Think that others will modify the data at the same time . Therefore, when operating data, lock the data directly , The lock will not be released until the operation is completed ; No one else can modify the data during locking .

Realization

Optimism lock ： There are two main ways to achieve this ：CAS Mechanism and Version number mechanism

Pessimistic locking ： The implementation method is locking , Lock can be a code block lock （ Such as Java Of synchronized keyword ）, It can also be data locking （ Such as MySQL The platoon lock in ）

Applicable scenario

Functional limitation

Compared with pessimistic lock , There are more restrictions on the application of optimistic lock , Whether it's CAS Or version number mechanism .

for example ,CAS Only atomicity of single variable operation can be guaranteed , When multiple variables are involved ,CAS There's nothing we can do , and synchronized Then we can lock the whole code block to deal with . Another example is the version number mechanism , If query When it comes to watches 1, and update When it comes to watches 2, It's also hard to implement optimistic locks with simple version numbers .

The intensity of the competition

• When the competition is not fierce ( The probability of concurrency conflict is small ) when , Optimistic lock has more advantages , Because pessimistic locks lock code blocks or data , Other threads cannot access at the same time , Affect concurrency , In addition, adding and releasing locks need to consume extra resources

• When the competition is fierce ( There is a high probability of concurrent conflicts ) when , Pessimistic lock has more advantages , Because optimistic locks often fail when performing updates , Need to keep retrying , waste CPU resources , Exit mechanism can be introduced , If the number of retries exceeds a certain threshold, it fails to exit . Of course , More importantly, avoid using optimistic locks in a highly competitive environment .

Questioning

• Optimistic lock plus lock ？  Optimistic lock itself is not locked , But sometimes optimistic locking may cooperate with locking operation （MYSQL It's locked ）

• CAS What are the disadvantages ？  ABA problem , That is, a thread changes the data in memory , however CAS The compared value is returned to the original value by the thread ,CAS Operation succeeded , The more effective solution is to introduce version number , Every time the value in memory changes , Version No +1, It's going on CAS In operation , Not only compare values in memory , Also compare version numbers , Only when neither of them has changed ,CAS To execute successfully