What is database concurrent access

Uncontrolled concurrent access to the same data by multiple users may result in incorrect data being accessed and stored
DBMS must provide concurrency control mechanism
Transaction is the basic unit of concurrency control and recovery

affairs

concept
- From the user's point of view, some operations on the database should be a whole , that is, an independent work unit, inseparable
- Example: Electronic funds transfer (transfer a sum of money from account A to account B) may be considered by the customer as an independent operation, while DBS is composed of several operations
definition
- A transaction is a sequence of database operations defined by the user . These operations are either done or not done. It is an indivisible unit of work.
- Transaction and program are two concepts
  - In a relational database, a transaction can be a single SQL statement, a set of SQL statements, or an entire program
  - An application usually consists of multiple transactions
- Transactions are the fundamental unit of recovery and concurrency control
Two ways of defining transactions
- explicit way
  - The start of a transaction is explicitly controlled by the user or automatically implied by the DBMS
  - The end of the transaction is explicitly controlled by the user
- implicit way
  - When the user does not explicitly define a transaction, the DBMS automatically divides the transaction by default

Explicitly define transactions

BEGIN TRANSACTION ←—— 事务开始
    SQL语句1
    SQL语句2
    ......
    COMMIT ←—— 事务结束



BEGIN TRANSACTION ←—— 事务开始
    SQL语句1
    SQL语句2
    ......
    ROLLBACK ←—— 事务结束

transaction end statement
- COMMIT
  - Transaction ended normally
  - Commit all operations of the transaction (read+update)
  - All updates to the database in the transaction take effect permanently
- ROLLBACK
  - transaction terminated abnormally
  - A failure occurred during the operation of the transaction, and it cannot continue to execute, and all update operations of the transaction are rolled back
  - The transaction is rolled back to the state it was in when it started
ACID properties
- Atomicity _ _
  - A transaction is a logical unit of work for a database
    - All operations included in the transaction are either done or not done
- Consistency _ _
  - The result of transaction execution must be to change the database from one consistent state to another consistent state
    - Consistent state: the database contains only the results of successful transaction commits
    - Inconsistent state: the database contains the results of failed transactions
  - Consistency and atomicity are closely related.
    Example: bank transfer: withdraw 10,000 yuan from account A and deposit it into account B.
    Define a transaction that includes two operations. The
    first operation is to subtract 10,000 from account A The
    second operation is to add 10,000 yuan to account B.
    These two operations can be done entirely or not at all, and
    the database is in a consistent state.
    If only one operation is performed, the user will logically An error occurred, 10,000 yuan was missing, and the database was in an inconsistent state
- Isolation _ _
  - The execution of a transaction cannot be interfered with by other transactions
    - The operations and data used within a transaction are isolated from other concurrent transactions
    - Transactions executed concurrently cannot interfere with each other
- Durability _ _
  - permanence also known as permanence
    - Once a transaction is committed, its changes to the data in the database should be permanent
    - Subsequent other operations or failures should not have any effect on its execution results
- Ensuring transaction ACID properties is an important task in transaction processing
- Factors that destroy the ACID properties of transactions
  - When multiple transactions are running in parallel , the operations of different transactions are executed cross-executed
    - The DBMS must ensure that the cross-running of multiple transactions does not affect the ACID properties of these transactions, especially atomicity and isolation
  - The transaction was forcibly stopped during operation
    - The DBMS must ensure that a transaction that is forcibly terminated has no impact on the database or other transactions
Multi-transaction execution mode
- Transaction serial execution
  - Only one transaction runs at a time, and other transactions must wait until the end of this transaction before they can run
  - Cannot make full use of system resources and give full play to the characteristics of database shared resources
- cross concurrency
  - The parallel execution of transactions is that the parallel operations of these parallel transactions run alternately
  - It is a concurrency method in a single processor system, which can reduce the idle time of the processor and improve the efficiency of the system
- Simultaneous mode
  - In a multi-processor system, each processor can run one transaction, and multiple processors can run multiple transactions at the same time, realizing the real parallel operation of multiple transactions
  - The most ideal way of concurrency, but subject to the hardware environment
Problems caused by concurrent execution of transactions
- Uncontrolled operation of multiple users concurrently accessing the same data may access and store incorrect data, destroying transaction isolation and database consistency
- DBMS must provide concurrency control mechanism
- The concurrency control mechanism is one of the important indicators to measure the performance of a DBMS

Inconsistency between concurrent operations and data

Causes:
① Caused by concurrent operations of two transactions
② In the case of concurrent operations, $T_{1}$ the $T_{2}$ scheduling of the operation sequence of the two transactions is random
③ If the above scheduling sequence is executed, $T_{1}$ the transaction modification will be discarded
④ Because of the fourth step After $T_{2}$ the transaction modifies A and writes it back, it overwrites $T_{1}$ the modification of the transaction

Data inconsistency caused by concurrent operations
- lost modification
- non-repeatable read
  - read-update
    - Transaction 1 reads a certain data
    - Transaction 2 modified it
    - When transaction 1 reads the data again, it gets a different value from the previous one
  - read-delete
    - Transaction 1 reads certain data records from the database according to certain conditions
    - Transaction 2 deleted some of the records
    - When transaction 1 reads data again according to the same conditions, it is found that some records disappear mysteriously
  - read-insert
    - Transaction 1 reads certain data records from the database according to certain conditions
    - Transaction 2 inserted some records
    - When transaction 1 reads data again according to the same conditions, it finds some more records
- read "dirty" data

Scheduling of Concurrent Operations

The computer system's scheduling of parallel operations in parallel transactions is random, and different scheduling may produce different results
The scheduling strategy that serializes all transactions must be the correct scheduling strategy
- If there are no other transactions running at the same time during the running of a transaction, that is to say, it is not disturbed by other transactions, then it can be considered that the running result of the transaction is normal or expected

serializable scheduling

Serial execution of transactions in a different order may also produce different results but since it does not put the database in an inconsistent state, all can be considered correct
Parallel execution of several transactions is correct
- iff the result is the same as executing them serially in some order. This parallel scheduling strategy is called serializable scheduling

Serializability is the only criterion for the correctness of
parallel transactions Example: Now there are two transactions, including the following operations:
transaction 1: read B; A=B+1; write back to A;
transaction 2: read A; B=A +1; Write back to B;
Assume that the initial value of A is 2, and the initial value of B is 2.

A Serial Scheduling Policy	B Serial Scheduling Policy
C Since its execution result is different from (a) and (b), it is a wrong scheduling. non-serializable schedule	D Since its execution result is the same as that of the serial schedule (a), it is a correct schedule. serializable scheduling

Serializable Scheduling Policy

In order to ensure the correctness of parallel operations, the parallel control mechanism of DBMS must provide certain means to ensure that the scheduling is serializable
Theoretically speaking, the scheduling strategy that prohibits the execution of other transactions when a certain transaction is executed must be serializable scheduling, which is also the simplest scheduling strategy
But this approach is practically infeasible because it prevents users from fully sharing database resources

practice questions

A method to ensure the correctness of concurrent operation scheduling

block method
- Two-stage lock (2PL) protocol
time stamping method
optimistic approach

Concurrency Control - Overview and Concurrency Control Scheduling