Transactions & Concurrency Control | Database Management System (DBMS) - Computer Science Engineering (CSE) PDF Download

ACID Properties

A transaction is a single logical unit of work which accesses and possibly modifies the contents of a database. Transactions access data using read and write operations.
In order to maintain consistency in a database, before and after transaction, certain properties are followed. These are called ACID properties.

Atomicity

By this, we mean that either the entire transaction takes place at once or doesn’t happen at all. There is no midway i.e. transactions do not occur partially. Each transaction is considered as one unit and either runs to completion or is not executed at all. It involves following two operations.
Abort: If a transaction aborts, changes made to database are not visible.
Commit: If a transaction commits, changes made are visible.
Atomicity is also known as the ‘All or nothing rule’.
Consider the following transaction T consisting of T1 and T2: Transfer of 100 from account X to account Y.

Transactions & Concurrency Control | Database Management System (DBMS) - Computer Science Engineering (CSE)

If the transaction fails after completion of T1 but before completion of T2.( say, after write(X) but before write(Y)), then amount has been deducted from X but not added to Y. This results in an inconsistent database state. Therefore, the transaction must be executed in entirety in order to ensure correctness of database state.

Consistency

This means that integrity constraints must be maintained so that the database is consistent before and after the transaction. It refers to correctness of a database. Referring to the example above,
The total amount before and after the transaction must be maintained.
Total before T occurs = 500 + 200 = 700.
Total after T occurs400 + 300 = 700.
Therefore, database is consistent. Inconsistency occurs in case T1 completes but T2 fails. As a result T is incomplete.

Isolation

This property ensures that multiple transactions can occur concurrently without leading to inconsistency of database state. Transactions occur independently without interference. Changes occurring in a particular transaction will not be visible to any other transaction until that particular change in that transaction is written to memory or has been committed. This property ensures that the execution of transactions concurrently will result in a state that is equivalent to a state achieved these were executed serially in some order.
Let X= 500, Y = 500.
Consider two transactions T and T”.

Transactions & Concurrency Control | Database Management System (DBMS) - Computer Science Engineering (CSE)

Suppose T has been executed till Read (Y) and then T’’ starts. As a result , interleaving of operations takes place due to which T’’ reads correct value of X but incorrect value of Y and sum computed by
T’’: (X+Y = 50, 000+500=50, 500)
is thus not consistent with the sum at end of transaction: 
T: (X+Y = 50, 000 + 450 = 50, 450).
This results in database inconsistency, due to a loss of 50 units. Hence, transactions must take place in isolation and changes should be visible only after a they have been made to the main memory.

Durability:

This property ensures that once the transaction has completed execution, the updates and modifications to the database are stored in and written to disk and they persist even is system failure occurs. These updates now become permanent and are stored in a non-volatile memory. The effects of the transaction, thus, are never lost.

The ACID properties, in totality, provide a mechanism to ensure correctness and consistency of a database in a way such that each transaction is a group of operations that acts a single unit, produces consistent results, acts in isolation from other operations and updates that it makes are durably stored.

Concurrency Control -Introduction

Concurrency Control deals with interleaved execution of more than one transaction. In the next article, we will see what is serializability and how to find whether a schedule is serializable or not.

What is Transaction?

A set of logically related operations is known as transaction. The main operations of a transaction are:

Read(A): Read operations Read(A) or R(A) reads the value of A from the database and stores it in a buffer in main memory.

Write (A): Write operation Write(A) or W(A) writes the value back to the database from buffer.

Let us take a debit transaction from an account which consists of following operations:

  1. R(A);
  2. A=A-1000;
  3. W(A);

Assume A’s value before starting of transaction is 5000.

  • The first operation reads the value of A from database and stores it in a buffer.
  • Second operation will decrease its value by 1000. So buffer will contain 4000.
  • Third operation will write the value from buffer to database. So A’s final value will be 4000.

But it may also be possible that transaction may fail after executing some of its operations. The failure can be because of hardware, software or power etc. For example, if debit transaction discussed above fails after executing operation 2, the value of A will remain 5000 in the database which is not acceptable by the bank. To avoid this, Database has two important operations:

Commit: After all instructions of a transaction are successfully executed, the changes made by transaction are made permanent in the database.

Rollback: If a transaction is not able to execute all operations successfully, all the changes made by transaction are undone.

Properties of a transaction

Atomicity: As a transaction is set of logically related operations, either all of them should be executed or none. A debit transaction discussed above should either execute all three operations or none.If debit transaction fails after executing operation 1 and 2 then its new value 4000 will not be updated in the database which leads to inconsistency.

Consistency: If operations of debit and credit transactions on same account are executed concurrently, it may leave database in an inconsistent state.

  • For Example, T1 (debit of Rs. 1000 from A) and T2 (credit of 500 to A) executing concurrently, the database reaches inconsistent state.
  • Let us assume Account balance of A is Rs. 5000. T1 reads A(5000) and stores the value in its local buffer space. Then T2 reads A(5000) and also stores the value in its local buffer space.
  • T1 performs A=A-1000 (5000-1000=4000) and 4000 is stored in T1 buffer space. Then T2 performs A=A+500 (5000+500=5500) and 5500 is stored in T2 buffer space. T1 writes the value from its buffer back to database.
  • A’s value is updated to 4000 in database and then T2 writes the value from its buffer back to database. A’s value is updated to 5500 which shows that the effect of debit transaction is lost and database has become inconsistent.
  • To maintain consistency of database, we need concurrency control protocols which will be discussed in next article.  The operations of T1 and T2 with their buffers and database have been shown in Table 1.
     

Transactions & Concurrency Control | Database Management System (DBMS) - Computer Science Engineering (CSE)

Table 1

Isolation: Result of a transaction should not be visible to others before transaction is committed. For example, Let us assume that A’s balance is Rs. 5000 and T1 debits Rs. 1000 from A. A’s new balance will be 4000. If T2 credits Rs. 500 to A’s new balance, A will become 4500 and after this T1 fails. Then we have to rollback T2 as well because it is using value produced by T1. So a transaction results are not made visible to other transactions before it commits.

Durable: Once database has committed a transaction, the changes made by the transaction should be permanent. e.g.; If a person has credited $500000 to his account, bank can’t say that the update has been lost. To avoid this problem, multiple copies of database are stored at different locations.

What is a Schedule?

A schedule is series of operations from one or more transactions. A schedule can be of two types:

  • Serial Schedule: When one transaction completely executes before starting another transaction, the schedule is called serial schedule. A serial schedule is always consistent. e.g.; If a schedule S has debit transaction T1 and credit transaction T2, possible serial schedules are T1 followed by T2 (T1->T2) or T2 followed by T1 ((T1->T2). A serial schedule has low throughput and less resource utilization.
  • Concurrent Schedule: When operations of a transaction are interleaved with operations of other transactions of a schedule, the schedule is called Concurrent schedule. e.g.; Schedule of debit and credit transaction shown in Table 1 is concurrent in nature. But concurrency can lead to inconsistency in database.  The above example of concurrent schedule is also inconsistent.

Question: Consider the following transaction involving two bank accounts x and y:

  1. read(x);
  2. x := x – 50;
  3. write(x);
  4. read(y);
  5. y := y + 50;
  6. write(y);

The constraint that the sum of the accounts x and y should remain constant is that of?

  1. Atomicity
  2. Consistency
  3. Isolation
  4. Durability

Solution: As discussed in properties of transactions, consistency properties says that sum of accounts x and y should remain constant before starting and after completion of transaction. So, the correct answer is B.

Next article- Serializability of Schedules

How to test if two schedules are View Equal or not ?

Two schedules S1 and S2 are said to be view equal iff following below conditions are satisfied :

1) Initial Read

 If a transaction T1 reading data item A from initial database in S1 then in S2 also T1 should read A from initial database.

Transactions & Concurrency Control | Database Management System (DBMS) - Computer Science Engineering (CSE)

Transaction T2 is reading A form initial database.

2) Updated Read

If Ti is reading A which is updated by Tj in S1 then in S2 also Ti should read A which is updated by Tj.

Transactions & Concurrency Control | Database Management System (DBMS) - Computer Science Engineering (CSE)

Above two schedule are not view equal as in S1 :T3 is reading A updated by T2, in S2 T3 is reading A updated by T1.

3) Final Write operation 

If a transaction T1 updated A at last in S1, then in S2 also T1 should perform final write operations.

Transactions & Concurrency Control | Database Management System (DBMS) - Computer Science Engineering (CSE)

Above two schedule are not view as Final write operation in S1 is done by T1 while in S2 done by T2.

View Serializability: A Schedule is called view serializable if it is view equal to a serial schedule (no overlapping transactions).

Conflict Serializability

As discussed in Concurrency control, serial schedules have less resource utilization and low throughput. To improve it, two are more transactions are run concurrently. But concurrency of transactions may lead to inconsistency in database. To avoid this, we need to check whether these concurrent schedules are serializable or not.

Conflict Serializable: A schedule is called conflict serializable if it can be transformed into a serial schedule by swapping non-conflicting operations.

Conflicting operations: Two operations are said to be conflicting if all conditions satisfy:

  • They belong to different transaction
  • They operation on same data item
  • At Least one of them is a write operation

Example: –

  • Conflicting operations pair (R1(A), W2(A)) because they belong to two different transactions on same data item A and one of them is write operation.
  • Similarly, (W1(A), W2(A)) and (W1(A), R2(A)) pairs are also conflicting.
  • On the other hand, (R1(A), W2(B)) pair is non-conflicting because they operate on different data item.
  • Similarly, ((W1(A), W2(B)) pair is non-conflicting.

Consider the following schedule:

S1: R1(A), W1(A), R2(A), W2(A), R1(B), W1(B), R2(B), W2(B)

If Oi and Oj are two operations in a transaction and Oi< O(Ois executed before Oj), same order will follow in schedule as well. Using this property, we can get two transactions of schedule S1 as:

T1: R1(A), W1(A), R1(B), W1(B)
T2: R2(A), W2(A), R2(B), W2(B)

Possible Serial Schedules are: T1->T2 or T2->T1

-> Swapping non-conflicting operationR2(A) and R1(B) in S1, the schedule becomes,

S11: R1(A), W1(A), R1(B), W2(A), R2(A), W1(B), R2(B), W2(B)

-> Similarly, swapping non-conflicting operations W2(A) and W1(B) in S11, the schedule becomes,

S12: R1(A), W1(A), R1(B), W1(B), R2(A), W2(A), R2(B), W2(B)

S12 is a serial schedule in which all operations of T1 are performed before starting any operation of T2. Since S has been transformed into a serial schedule S12 by swapping non-conflicting operations of S1, S1 is conflict serializable.

Let us take another Schedule:

S2: R2(A), W2(A), R1(A), W1(A), R1(B), W1(B), R2(B), W2(B)

Two transactions will be:

T1: R1(A), W1(A), R1(B), W1(B)
T2: R2(A), W2(A), R2(B), W2(B)

Possible Serial Schedules are: T1->T2 or T2->T1

Original Schedule is:

S2: R2(A), W2(A), R1(A), W1(A), R1(B), W1(B), R2(B), W2(B)

Swapping non-conflicting operations R1(A) and R2(B) in S2, the schedule becomes,

S21: R2(A), W2(A), R2(B), W1(A), R1(B), W1(B), R1(A), W2(B)

Similarly, swapping non-conflicting operations W1(A) and W2(B) in S21, the schedule becomes,

S22: R2(A), W2(A), R2(B), W2(B), R1(B), W1(B), R1(A), W1(A)

In schedule S22, all operations of T2 are performed first, but operations of T1 are not in order (order should be R1(A), W1(A), R1(B), W1(B)). So S2 is not conflict serializable.

Conflict Equivalent: Two schedules are said to be conflict equivalent when one can be transformed to another by swapping non-conflicting operations. In the example discussed above, S11 is conflict equivalent to S1 (S1 can be converted to S11 by swapping non-conflicting operations). Similarly, S11 is conflict equivalent to S12 and so on.

Note 1: Although S2 is not conflict serializable, but still it is conflict equivalent to S21 and S21 because S2 can be converted to S21 and S22 by swapping non-conflicting operations.

Note 2: The schedule which is conflict serializable is always conflict equivalent to one of the serial schedule. S1 schedule discussed above (which is conflict serializable) is equivalent to serial schedule (T1->T2).

Question: Consider the following schedules involving two transactions. Which one of the following statement is true?

S1: R1(X) R1(Y) R2(X) R2(Y) W2(Y) W1(X)

S2: R1(X) R2(X) R2(Y) W2(Y) R1(Y) W1(X)

  • Both S1 and S2 are conflict serializable
  • Only S1 is conflict serializable
  • Only S2 is conflict serializable
  • None

Solution: Two transactions of given schedules are:

 T1: R1(X) R1(Y) W1(X)
 T2: R2(X) R2(Y) W2(Y)

Let us first check serializability of S1:

S1: R1(X) R1(Y) R2(X) R2(Y) W2(Y) W1(X)

To convert it to a serial schedule, we have to swap non-conflicting operations so that S1 becomes equivalent to serial schedule T1->T2 or T2->T1. In this case, to convert it to a serial schedule, we must have to swap R2(X) and W1(X) but they are conflicting. So S1 can’t be converted to a serial schedule.

Now, let us check serializability of S2:

S2: R1(X) R2(X) R2(Y) W2(Y) R1(Y) W1(X)

Swapping non conflicting operations R1(X) and R2(X) of S2, we get

S2’: R2(X) R1(X) R2(Y) W2(Y) R1(Y) W1(X)

Again, swapping non conflicting operations R1(X) and R2(Y) of S2’, we get

S2’’: R2(X) R2(Y) R1(X) W2(Y) R1(Y) W1(X)

Again, swapping non conflicting operations R1(X) and W2(Y) of S2’’, we get

S2’’’: R2(X) R2(Y) W2(Y) R1(X) R1(Y) W1(X)

which is equivalent to a serial schedule T2->T1.

So, correct option is C. Only S2 is conflict serializable.

Recoverability of Schedules

We have discussed the basics of Transactions and Schedules in Concurrency Control (Introduction) article. As discussed, a transaction may not execute completely due to hardware failure, system crash or software issues. In that case, we have to rollback the failed transaction. But some other transaction may also have used values produced by failed transaction. So we have to rollback those transactions as well.

Transactions & Concurrency Control | Database Management System (DBMS) - Computer Science Engineering (CSE)

Above table shows a schedule with two transactions, T1 reads and writes A and that value is read and written by T2. T2 commits. But later on, T1 fails. So we have to rollback T1. Since T2 has read the value written by T1, it should also be rollbacked. But we have already committed that. So this schedule is irrecoverable schedule.

Irrecoverable Schedule: When Tj is reading the value updated by Ti and Tj is committed before commit of Ti, the schedule will be irrecoverable.

Transactions & Concurrency Control | Database Management System (DBMS) - Computer Science Engineering (CSE)

Table 2 shows a schedule with two transactions, T1 reads and writes A and that value is read and written by T2. But later on, T1 fails. So we have to rollback T1. Since T2 has read the value written by T1, it should also be rollbacked. As it has not committed, we can rollback T2 as well. So it is recoverable with cascading rollback.
Recoverable with cascading rollback: If Tj is reading value updated by Ti and commit of Tj is delayed till commit of Ti , the schedule is called recoverable with cascading rollback.

Transactions & Concurrency Control | Database Management System (DBMS) - Computer Science Engineering (CSE)

Table 3 shows a schedule with two transactions, T1 reads and writes A and commits and that value is read by T2. But if T1 fails before commit, no other transaction has read its value, so there is no need to rollback other transaction. So this is a cascadeless recoverable schedule.

Cascadeless Recoverable: If Tj reads value updated by Ti only after Ti is commited, the schedule will be cascadeless recoverable.

Question: Which of the following scenarios may lead to an irrecoverable error in a database system?
(A) A transaction writes a data item after it is read by an uncommitted transaction.
(B) A transaction reads a data item after it is read by an uncommitted transaction.
(C) A transaction reads a data item after it is written by a committed transaction.
(D) A transaction reads a data item after it is written by an uncommitted transaction.

Answer: See the example discussed in Table 1, a transaction is reading a data item after it is written by an uncommitted transaction, the schedule will be irrecoverable.

The document Transactions & Concurrency Control | Database Management System (DBMS) - Computer Science Engineering (CSE) is a part of the Computer Science Engineering (CSE) Course Database Management System (DBMS).
All you need of Computer Science Engineering (CSE) at this link: Computer Science Engineering (CSE)
62 videos|66 docs|35 tests

Top Courses for Computer Science Engineering (CSE)

FAQs on Transactions & Concurrency Control - Database Management System (DBMS) - Computer Science Engineering (CSE)

1. What is a transaction in computer science engineering?
Ans. In computer science engineering, a transaction refers to a logical unit of work performed on a database. It consists of a sequence of operations, such as reading or writing data, that must be executed as a single, indivisible unit. Transactions ensure data integrity and consistency within a database.
2. What is concurrency control in computer science engineering?
Ans. Concurrency control in computer science engineering refers to the management of simultaneous execution of multiple transactions in a database system. It aims to ensure that transactions do not interfere with each other and maintain data consistency. Techniques like locking, timestamp ordering, and optimistic concurrency control are used to achieve concurrency control.
3. How does locking help in concurrency control?
Ans. Locking is a technique used in concurrency control to prevent conflicts between concurrent transactions accessing the same data. When a transaction needs to read or modify data, it acquires a lock on the data item. This lock restricts other transactions from accessing the data until the lock is released, ensuring that only one transaction can modify the data at a time and maintaining data consistency.
4. What is the role of timestamps in concurrency control?
Ans. Timestamps play a crucial role in concurrency control. Each transaction is assigned a unique timestamp, which represents its order of execution. Timestamp ordering technique uses these timestamps to schedule transactions in a way that maintains data consistency. Transactions with earlier timestamps are given priority, and conflicts between transactions are resolved based on their timestamps.
5. What is optimistic concurrency control?
Ans. Optimistic concurrency control is a technique used in database systems to manage concurrent transactions. Instead of locking data items, optimistic concurrency control allows multiple transactions to proceed concurrently without acquiring locks. Before committing, each transaction checks if any other transaction has modified the data it accessed. If conflicts are detected, the transaction is rolled back and re-executed. This technique reduces the need for locking and improves concurrency, but it requires additional overhead for conflict detection and resolution.
62 videos|66 docs|35 tests
Download as PDF
Explore Courses for Computer Science Engineering (CSE) exam

Top Courses for Computer Science Engineering (CSE)

Signup for Free!
Signup to see your scores go up within 7 days! Learn & Practice with 1000+ FREE Notes, Videos & Tests.
10M+ students study on EduRev
Related Searches

past year papers

,

Exam

,

shortcuts and tricks

,

Viva Questions

,

Transactions & Concurrency Control | Database Management System (DBMS) - Computer Science Engineering (CSE)

,

Free

,

Objective type Questions

,

Important questions

,

Extra Questions

,

practice quizzes

,

study material

,

ppt

,

Semester Notes

,

pdf

,

video lectures

,

Transactions & Concurrency Control | Database Management System (DBMS) - Computer Science Engineering (CSE)

,

Transactions & Concurrency Control | Database Management System (DBMS) - Computer Science Engineering (CSE)

,

Previous Year Questions with Solutions

,

MCQs

,

Summary

,

Sample Paper

,

mock tests for examination

;