24-26| Transaction Management

& Concurrency Control

Database Systems
Transaction Management &
Concurrency Control

ACID, Commit, Rollback

 In this section, you will learn:
▪ What a database transaction is and what its
properties are
▪ How database transactions are managed
▪ What concurrency control is and what role
it plays in maintaining the databases
integrity

3
 Database Usage
▪ You are using multiple databases
Everyday knowingly or Unknowingly

Examples Uses
▪ Using ATM
▪ Using your Campus Learning
Management System
▪ Using Facebook
Database Usage
Database Usage
 What is a Transaction?

▪ A set of instructions used to perform a logical unit of

work.
▪ It generally represents a change in database.
▪ Collection of instructions

• Read from or write to a database, create a transaction.

• It can be a single SQL command or series of related
commands.
• In practical, transactions are formed by two or more database
requests.
• Database request is the equivalent of a single SQL
statement.
8
 What is a Transaction?
▪ Transaction changes the contents of the
database, must alter the database from one
consistent state to another.
▪ Consistent database state is one in which all
data integrity constraints are satisfied.
▪ All transactions are controlled and executed
by the DBMS to guarantee database
integrity.
9
 What is a Transaction?
▪ Logical unit of work
▪ Must be either entirely completed or
aborted
▪ No intermediate states are acceptable

10
 Example Transaction 1
▪ Register credit sale of 100 units of product
X to customer Y for $500
▪ Reduce product X’s quantity by 100
▪ Adding $500 to customer Y’s account
receivable
UPDATE PRODUCT
SET PROD_QOH = PROD_QOH - 100
WHERE PROD_CODE = ‘X’;
UPDATE ACCT_RECEIVABLE
SET ACCT_AMOUNT = ACCT_ AMOUNT + 500
11
WHERE ACCT_NUM = ‘Y’;
Transaction

12
Transaction States

13
 Properties of Transaction

▪ Atomicity
▪ Consistency
▪ Isolation
▪ Durability
15
 Transaction Properties

▪ A failed transaction cannot be resumed, it will be always restarted.

▪ Even If a transaction is 90% complete and it fails, it wont resume.
▪ ATM transactions are always restarted.

16
 Transaction Properties

▪ Before a transaction starts and after a transaction is completed, sum of money should be same.

A has 2000, B has 3000 : total is 5000

R(a) :2000
A=a-1000 (1000)
R(b) : 3000
B=b+1000 (4000)
W(b) : 4000
17
Total now: 4000+1000=5000
Consistency Example

18
Transaction Properties

19
20
Isolation Example

21
 Transaction Properties

Ensures that the result or effect of a committed

transaction persists in case of a system failure.
2
2
 What is a Schedule?
▪ Schedule: is a collection of transactions

▪ Serial Scheduling
▪ No other transaction can interfere unless the current transaction is executed.
▪ Maintains consistency
▪ Problem: waiting time, other transactions will have to wait
▪ Performance is degraded in terms of throughput: num of transactions
executed/time

▪ Parallel Scheduling
▪ Multiple transactions can be executed
▪ CPU is switched to enable concurrent transparency
▪ Read-write, write-read, write-write conflicts

23
 Concurrency Transparency

The design objective for a distributed database is to

ensure that even when multiple transactions are
running simultaneously, each transaction perceives
itself as the sole operation in the system. This means
that every transaction should operate as if it has
exclusive access to the database, regardless of other
concurrent activities. Consequently, the outcome of
executing multiple transactions concurrently should
mirror the results obtained if each transaction were
executed sequentially, one after the other.
24
 Failure Transparency

It refers to the extent to which errors and subsequent

recoveries of hosts and services within the system are
invisible to users and applications. For example, if a
server fails, but users are automatically redirected to
another server and never notice the failure, the
system is said to exhibit high failure transparency.

25
 Concurrency Control
▪ Coordinates simultaneous transaction execution in
multiprocessing database.

▪ Simultaneous execution of transactions over a shared

database can create several problems in data integrity and
consistency.

– Potential problems in multi-user environments

• Dirty Read Problem
• Unrepeatable Read Problem
• Phantom Read Problem
• Lost updates (write-write conflict) 26
 Dirty Read Problem
T1 T2 ▪ T2 read a value from T1 and commits it.
R(A) ▪ What if T1 fails later?
W(A) ▪ Incase of a failure in T1, transaction will
…. R(A) rollback.
…. Commit ▪ If T1 rollbacks, T2 has read a value that
…. does not exist.
failure ▪ This causes inconsistency in the database.
▪ Reading Uncommitted Data

27
 Unrepeatable Read Problem
T1 T2 ▪ When you read a variable multiple times and
R(A) each time you get a different value.
R=10
▪ T2 read A as 10, meanwhile T1 writes a
R(A)
R=10
different value on A (11).
W(A) ▪ If T2 reads A again, it gets to read a different
R=11 value (11), although it didn’t write on A.
…. R(A) ▪ T2 was thinking that it is performing in
R=11
isolation, however T1 changed A value.
….
▪ A read that cannot be repeated.
▪ Inconsistent retrievals
28
 Phantom Read Problem
T1 T2 ▪ T2 reads A for the first time.
R(A) ▪ T1 deletes A.
R(A) ▪ T2 reads A for the second time and gets an
del A error because A doesn’t exist.
…. R(A) ▪ When you aren’t able to read a variable for
the second time due to change in the
structure of the data items, that is the
phantom read problem.

29
 Lost Update
(write-write)Problem
T1 T2 ▪ T2 doesn’t read A, just writes on it. This is
R(A) referred to as blind write.
A=10 ▪ T1 reads A as 10.
W(A)
A=11 ▪ Writes 11 on it.
W(A) ▪ T2 performs a blind write and commits.
A=50
▪ T1 wrote a different value of A: 11, however
C
this value is being overwritten by T2, but T1
C
isn’t aware of it.
▪ 11 is being lost and updated to 50, but T1
doesn’t know this.
30
 The Scheduler
▪ How is the correct order? Who determines
the order?
▪ Fortunately, DBMS handles that tricky
assignment for us by using a built-in
scheduler
▪ Scheduler establishes order of concurrent
transaction execution

38
 The Scheduler
▪ Ensures execution of database operations to ensure
serializability
▪ Bases actions on concurrency control algorithms
▪ Locking
▪ Time stamping
▪ Ensures efficient use of computer’s CPU
▪ No scheduling, make transaction first comes first serve
(FCFS)
▪ Lost CPU cycles during processing of I/O
▪ If FCFS, one transaction will be performed at a time, while its I/O
operations CPU cycles will be wasted 39
 Read/Write Conflict Scenarios:
Conflicting Database Operations
Matrix

Table 9.9

40
 Concurrency Control
with Locking Methods
▪ Lock guarantees current transaction
exclusive use of data item (T2 does not
have to access to the data item that
currently being used by transaction t1)
▪ Acquires lock prior to access
▪ Lock is released (unlocked) when
transaction is completed. So other
transactions can lock the data
41
 Concurrency Control
with Locking Methods
▪ DBMS automatically initiates and enforces
locking procedures
▪ Managed by lock manager (built-in DBMS),
which is responsible for assigning and
policing the locks used by the transactions
▪ Lock granularity indicates level of lock use

42
 Lock Granularity
▪ Lock granularity indicates level of lock use
▪ Locking can take place at the following
levels:
▪ Database
▪ Table
▪ Page
▪ Row
▪ Even field (attribute)
43
Database-Level Locking Sequence
• A complete entity database is locked
• Preventing the use of any tables in the
database by transaction T2 while
transaction t1 is being executed.
• Good for batch processes
• Unsuitable for online multi-user DBMSs (it
waits for previous transaction to be
completed before execution of next
transaction)
44
Database-Level Locking Sequence

Figure 9.2
45
 Table-Level Locking Sequence
▪ A entire table is locked
▪ Preventing access to any row by
transaction T2 while transaction t1 is using
the table.
▪ Less restrictive than database locks
▪ Not suitable for multi-user DBMSs

46
Table-Level Lock Example

Figure 9.3
47
 Page-Level Locking Sequence
• Lock an entire diskpage, page
(equivalent of a disk lock), described
as a referenced section of a disk.
• A table can span several pages
• Each page contains several rows of
one or more tables
• Page-level locks are currently the most
frequently used multi-user DBMSs
locking method 48
 Page-Level Lock Example

Figure 9.4

If T2 wants to access the page locked by T1,

T2 must wait until T1 the page is unlocked by T1 49
 Row-Level Locking Sequence
▪ Much less restrictive than the locks
discussed above
▪ DBMS allows concurrent transactions to
access different rows of the same table,
even if the rows are locked on the same
page.
▪ Management requires high overhead cost

50
 Row-Level Lock Example

Figure 9.5

51
 Field (attribute)-Level Locking
Sequence
▪ Allows concurrent transactions to access
the same row, as long as they require the
use of different fields (attributes) within
the row.
▪ Most flexible multi-data access, it is rarely
done because it requires an extremely
high level of computer overhead.

52
 Lock Types: Binary Locks and
Shared/Exclusive Locks

▪ Binary locks
▪ Two states
▪ Locked (1)
▪ Unlocked (0)
▪ Transaction must unlock the object after its
termination
▪ Automatically managed or scheduled by DBMSs
▪ User does not need to concerned about locking
or unlocking data items
53
 Shared/Exclusive Locks
▪ Exclusive
▪ Used when potential for conflict exists.
▪ Issued when transaction wants to update
unlocked data.

54
 Shared/Exclusive Locks
▪ Shared
▪ Exists when concurrent transactions granted
READ access
▪ Produces no conflict for read-only transactions
▪ Issued when transaction wants to read and
Exclusive Lock not held on item

55
 Deadlocks
▪ Occurs when two transactions wait for
each other to unlock data

T1 ---- access data items X and Y

T2 ---- access data items Y and X

If T1 has not unlocked data items, T2 can not starts.

If T2 has not unlocked data items T1 can not continue.

56
 Deadlocks
▪ A deadlock occurs when two transactions have a
lock, each on a separate object, and they want
to acquire a lock on each other's object.
▪ A deadlock is a condition where two or more
transactions are waiting indefinitely for one
another to give up locks.
▪ Deadlock is said to be one of the most feared
complications in DBMS as no task ever gets
finished and is in waiting state forever.
57
In the student table, transaction T1 holds a lock on some Student rows and needs
to update some rows in the grade table. Simultaneously, transaction T2 holds locks
on some rows in the grade table and needs to update the rows in the Student table
held by Transaction T1.

Now Transaction T1 is waiting for T2 to release its lock and similarly, transaction T2
is waiting for T1 to release its lock. All activities come to a halt state and remain at a
standstill. It will remain in a standstill until the DBMS detects the deadlock and
aborts one of the transactions.
58
 Deadlock Prevention
▪ For a large database, the deadlock prevention
method is suitable.
▪ A deadlock can be prevented if the resources are
allocated in such a way that deadlock never
occurs.
▪ The DBMS analyzes the operations whether they
can create a deadlock situation or not, If they do,
that transaction is never allowed to be executed.

59
 Deadlocks
▪ In such a case, SQL detects the deadlock
automatically and solves the problem by
aborting one of the two transactions.

Prevention Methods
• Mutual Exclusion: that prevents simultaneous access to a shared
resource.
• Hold and Wait: infinite loop
• No preemption: Not temporarily interrupting an executing task
• Circular wait: Two or more processes wait for resources in a circular
order. To eliminate circular wait, we assign a priority to each
resource. 60
 Serializability
▪ Serializability is a concept in database systems
that ensures transactions are executed in a
manner that preserves consistency and
correctness of the database state, despite their
concurrent execution.
▪ It guarantees that the final outcome of executing
multiple transactions concurrently is equivalent to
the outcome that would be obtained if the
transactions were executed in some serial order,
one after the other.
61
 Serializability
T1 T2
R(a)
R(a) T1 T2
W(a)
W(a) When you get a loop, it
means it’s a parallel
schedule.

▪ Is it serializable?
▪ It means find a clone of this parallel schedule that is a serial scheduler.
▪ To convert it we can either make it T1 → T2 or T2→T1
▪ Two methods to check if its serializable.
▪ Conflict Equivalent serializable
62
▪ View serializable
 Conflict Serializability
▪ Conflict serializability is based on the notion of
conflicting operations between transactions.
▪ A schedule (or execution order) of transactions is
conflict serializable if it can be transformed into a
serial schedule by swapping non-conflicting
operations while preserving the order of conflicting
operations.
▪ Conflicting operations include read-write, write-read,
and write-write operations on the same data item by
different transactions.

63
 Conflict Equivalent
▪ Is S equivalent to S^
▪ Check for adjacent non-conflicting pairs

S S^
T1 T2 T1 T2
R(a) R(a)
W(a) W(a)
R(a) R(b)
W(a) R(a)
R(b) W(a)
64
 Conflict Equivalent
▪ Is S=S^
▪ Check for adjacent non-conflicting pairs

S S^
T1 T2 T1 T2
R(a) R(a)
W(a) W(a)
R(a) R(b)
W(a) R(a)
R(b) W(a) 65
 Conflict Equivalent
▪ Is S=S^
▪ Check for adjacent non-conflicting pairs

S S^
T1 T2 T1 T2
R(a) R(a)
W(a) W(a)
R(a) R(b)
R(b) R(a)
W(a) W(a) 66
 Conflict Equivalent
▪ Is S=S^
▪ Check for adjacent non-conflicting pairs

S S^
T1 T2 T1 T2
R(a) R(a)
W(a) W(a)
R(a) R(b)
R(b) R(a)
W(a) W(a) 67
 Conflict Equivalent
▪ Is S=S^
▪ Check for adjacent non-conflicting pairs

S S^
T1 T2 T1 T2
R(a) R(a)
W(a) W(a)
R(b) R(b)
R(a) R(a)
W(a) W(a) 68
 Conflict Equivalent
▪ If for S scheduler there is a conflict equivalent S^
then it means S is serializable
▪ S→S^→ serializable

S S^
T1 T2 T1 T2
R(a) R(a)
W(a) W(a)
R(b) R(b)
R(a) R(a)
W(a) W(a) 69
 Conflict Equivalent
T1 T2
▪ Another example R(a)
▪ Which are adjacent pairs here? W(a)
R(a)
W(a)
R(b)

70
 Conflict Equivalent
▪ Another example
▪ Which are adjacent pairs here?
▪ Are these non-conflicting?
▪ No, hence no change in positions

T1 T2
R(a)
W(a)
R(a)
W(a)
R(b) 71
Example 2: Conflict Equivalent
Serializable
Check conflict pairs in other transactions
and draw edge.
T1 T2 T3
Make Precedence Graph
R(x)
R(y)
Check for conflicting pairs.
R(x)
R(y)
Conflict of R(X) is W(X): so find W(X)
R(z)
in T2 and T3.
W(y)
W(Z) T1
T2
R(Z)
W(X)
W(Z) T3
10
Conflict Equivalent Serializable
Check conflict pairs in other
T1 T2 T3
transactions and draw edge.
R(x)
R(y)
R(x)
For W(Y): u’ll check for 2 conflicts
R(y)
R(Y), W(Y)
R(z)
W(y) T1 T2
W(Z)
R(Z)
W(X)
T3
W(Z) 73
Conflict Equivalent Serializable
Check conflict pairs in other
T1 T2 T3
transactions and draw edge.
R(x)
Does the graph have a loop/cycle?
R(y)
No, means its
R(x)
R(y)
R(z)
Conflict T1 T2
W(y)
Serializable
W(Z)
R(Z)
W(X)
T3
W(Z) 74
 Conflict Equivalent Serializable
If a schedule is conflict serializable, it means there will be an
equivalent schedule that is serialized.
So it is consistent.

Conflict → serializable → consistent

Serializable T1 T2

T3

75
 Conflict Equivalent Serializable
What combinations will be formed? What sequence of
transactions will be formed?
From the graph check which in-degree of the vertex is
zero.
T2 (no arrow towards T2) T1 T2

T3

76
 Conflict Equivalent Serializable
If a schedule is conflict serializable, it means there will be an
equivalent schedule that is serialized.
So it is consistent.
What combinations will be formed? Which sequence will be
formed?
From the graph check which T1
in-degree of the vertex is zero.

T2
T3 T3

77
 Conflict Equivalent Serializable
If a schedule is conflict serializable, it means there will be an
equivalent schedule that is serialized.
So it is consistent.
What combinations will be formed? Which sequence will be
formed?
From the graph check which T1
in-degree of the vertex is zero.

T2
T3
T1
78
▪ Que:Check serializability for this schedule.
T1 T2 T3
T1 T2
R(A)
W(A) Get a loop so wont
W(A) T3 draw further
W(A)
This schedule is not conflict
equivalent but is it serializable?
▪ We will use view serializable.

79
 View Serializability
▪ View serializability is based on the idea of the
database state seen by transactions.
▪ A schedule is view serializable if it produces
the same set of results (or views) as some
serial schedule, regardless of the specific
operations performed by transactions.
▪ View serializability is more general than
conflict serializability and considers the overall
effect of transactions on the database state
rather than focusing on specific conflicting
operations.
▪ Que:Check serializability for this schedule.
T1 T2 T3
T1 T2
R(A)
W(A) Get a loop so wont
W(A) T3 draw further
W(A)
This schedule is not conflict
equivalent but is it serializable?
▪ We will use view serializable.
Test for values of A in
T1 T2 T3
both tables.
R(A) A=100
W(A) A-40
W(A) A-40
W(A) A-20 81
▪ Check serializability for this schedule.
T1 T2 T3
T1 T2
R(A)
W(A) Get a loop so wont
W(A) T3 draw further
W(A)
This schedule is not conflict
equivalent but is it serializable?
▪ We will use view serializable.
Test for values of A in
T1 T2 T3
both tables.
R(A) A=100
The answers are same
W(A) A-40
hence both schedulers
W(A) A-40
are equivalent.
W(A) A-20 82
 View Serializable
▪ Rules
▪ For each data item in the transaction,
check these three rules:
1. Initial Read
2. Update
3. Final Update

83
 Example 1: View Serializable
▪ Check can this be solved through
T1 T2 T3 conflict equivalent method?
R(B)
W(A) ▪ Do you get a loop while checking
for conflicting pairs?
R(A)
R(A)
W(B)
W(B)
W(B)

84
 Example 1: View Serializable
T1 T2 T3
R(B)
A B
W(A)
R(A) Initial Read

R(A) Update

W(B) Final Update

W(B)
W(B)

85
 Example 1: View Serializable
T1 T2 T3
R(B)
A B
W(A)
R(A) Initial Read

R(A) Update

W(B) Final Update

W(B)
W(B)

▪ Initial read means if it is read

before updating that item. Since A
is written first hence this rule
won’t be applied for A. 22
 Example 1: View Serializable
T1 T2 T3
R(B)
A B
W(A)
R(A) Initial Read T2

R(A) Update T2 T1, T2, T3

W(B) Final Update T2 T3

W(B)
W(B)

▪ To find the sequence of transactions, we see that our initial read for
B is T2 and final update is T3, so sequence becomes
▪ T2 → T1→ T3
23
▪ You can check by assigning values to A and B.
 Example 2: View Serializable
T1 T2 T3
R(Z)
R(Y)
W(Y) X Y Z
R(Y) Initial T1 T2 T2
R(Z) Read
R(X) Update T1, T2 T2,T3,T1 T2

W(X) Final T2 T1 T2
Update
W(Y)
W(Z)
R(Z)
R(Y)
W(Y)
W(X) 24
 Example 2: View Serializable
T1 T2 T3
X Y Z
R(Z)
Initial T1 T2 T2
R(Y)
Read
W(Y)
Update T1, T2 T2,T3,T1 T2
R(Y)
Final T2 T1 T2
R(Z) Update
R(X)
W(X) Now, we must find a sequence of transactions.
W(Y) Since Z has T2 only in all columns, this can be
ignored.
W(Z)
See for X, Y
R(Z)
X: T1→T2, Y: T2→ T3→T1
R(Y)
X and Y, both contradict each other.
W(Y)
Hence No order consists for this schedule.
W(X)
So, its not view serializable. 25
 Practice Ques:
T1 T2 T3 Find if this schedule is
R(A) serializable.
W(A)
R(A) Make precedence graph.
W(A) If you find loop, then check
W(A) with view serializable method.

90
 Practice Ques:
Make precedence graph, it
T1 T2 T3 has loops
R(A)
W(A) Make the table for view
R(A) serializable and find the
W(A) sequence.
W(A)
A
Initial Read T1
T1 T2 Update T2, T1, T3
Final Update T3

T3 Answer: T1 → T2 → T3 91