Concurrency control of performance tuning methodology

Locks and transactions

lock
Business
Indivisibility (Atomicity)
Uniformity (Consistency)
Isolation, (Isolation)
persistence (Durability)

Concurrent

The transaction isolation

Uncommitted read
Read committed
Repeatable
serialize
Read commit snapshot
snapshot

Lock upgrade granularity

That's ok
page
Expand
Partition
surface

Suggestions for using transactions

The shorter the transaction duration, the better
Avoid interacting with users during transactions
During data query , Try not to enable transactions
Make use of transaction isolation level and lock prompt

Deadlock

Cyclic deadlock
conversion deadlock
Distributed deadlocks

Cause of lock and related treatment

Time consuming queries or transactions
Incorrect transaction or transaction isolation level
The transaction is not handled correctly
Undetected distributed deadlock
Locking data granularity is too high or too low

The basic principle of

Prevent locking others

Transactions should not cross batches , And the shorter the better , Don't interact with users during transactions
Handle the timeout carefully and give up , Or execution error
Build the right index , Data tables have clustered indexes
Try not to start implicit transactions , Avoid opening transactions for a long time
Try to reduce the transaction isolation level
If allowed , You can try using lock hints

Prevent and handle deadlocks

Try to avoid or deal with the blockage as soon as possible
Access data in the same order
Let different connections use the same lock
Provide different data access paths

The solution after the birth and death lock

Set up Deadlock priority , Let unimportant affairs give up automatically
Welcome to join me for wechat exchange and discussion （ Please note csdn Add ）