Introduction to Transaction Processing

Concepts and Theory

 Introduction to Transaction Processing

 Single-User System: At most one user at a time can

use the system.
 Multiuser System: Many users can access the system
concurrently.
 Concurrency
– Interleaved processing: concurrent execution of
processes is interleaved in a single CPU
– Parallel processing: processes are concurrently
executed in multiple CPUs.

2
 Introduction to Transaction Processing
 A Transaction: logical unit of database processing that
includes one or more access operations (read -retrieval,
write - insert or update, delete).
 A transaction (set of operations) may be stand-
alone specified in a high level language like SQL
submitted interactively, or may be embedded within a
program.
 Transaction boundaries: Begin and End transaction.
 An application program may contain several
transactions separated by the Begin and End
transaction boundaries.

17-3
 Introduction to Transaction Processing
SIMPLE MODEL OF A DATABASE (for
purposes of discussing transactions):
 A database - collection of named data items
 Granularity of data – size of data item like a field, a
record , or a whole disk block
 Basic operations are read and write
– read_item(X): Reads a database item named X into a
program variable. To simplify our notation, we assume
that the program variable is also named X.
– write_item(X): Writes the value of program variable X
into the database item named X.

4
 Introduction to Transaction Processing

READ AN D WRITE OPERATION S:

 Basic u nit of d ata transfer from the d isk to the
com pu ter m ain m em ory is one block. In general, a
d ata item (w hat is read or w ritten) w ill be the field
of som e record in the d atabase, althou gh it m ay be a
larger u nit su ch as a record or even a w hole block.
 read_item(X) command includes the follow ing
steps:
1. Find the ad d ress of the d isk block that contains item X.
2. Copy that d isk block into a bu ffer in m ain m em ory (if that
d isk block is not alread y in som e m ain m em ory bu ffer).
3. Copy item X from the bu ffer to the p rogram variable nam ed X.

5
 Introduction to Transaction Processing
READ AN D WRITE OPERATION S (cont.):
 w rite_item(X) command includes the follow ing
steps:
1. Find the ad d ress of the d isk block that contains
item X.
2. Copy that d isk block into a bu ffer in m ain m em ory
(if that d isk block is not alread y in som e m ain
m em ory bu ffer).
3. Copy item X from the p rogram variable nam ed X
into its correct location in the bu ffer.
4. Store the u pd ated block from the bu ffer back to
d isk (either im m ed iately or at som e later point in
tim e).
6
 FIGURE
Two sample transactions. (a) Transaction T1.
(b) Transaction T2.

7
 Introduction to Transaction Processing
Why Concurrency Control is needed:
 The Lost Update Problem.
This occu rs w hen tw o transactions that access the
sam e d atabase item s have their operations
interleaved in a w ay that m akes the valu e of som e
d atabase item incorrect.
 The Temporary Update (or D irty Read) Problem.
This occu rs w hen one transaction u pd ates a d atabase
item and then the transaction fails for som e reason.
The u pd ated item is accessed by another transaction
before it is changed back to its original valu e.

8
 Introduction to Transaction Processing

Why Concurrency Control is needed (cont.):

 The Incorrect Summary Problem .
If one transaction is calcu lating an aggregate
su m m ary fu nction on a nu m ber of record s w hile
other transactions are u p d ating som e of these
record s, the aggregate fu nction m ay calcu late som e
valu es before they are u pd ated and others after they
are u pd ated .

9
Some problems that occur when concurrent
execution is uncontrolled. (a) The lost update
problem.

10
 Some problems that occur when concurrent
execution is uncontrolled. (b) The temporary update
problem.

11
Some problems that occur when concurrent execution is
uncontrolled. (c) The incorrect summary problem.

12
 Introduction to Transaction Processing
Why recovery is needed:
(What cau ses a Transaction to fail)
1. A computer failure (system crash): A hard w are or
softw are error occu rs in the com pu ter system d u ring
transaction execu tion. If the hard w are crashes, the
contents of the com pu ter’s internal m em ory m ay be
lost.
2. A transaction or system error : Som e operation in the
transaction m ay cau se it to fail, su ch as integer overflow
or d ivision by zero. Transaction failu re m ay also occu r
becau se of erroneou s param eter valu es or becau se of a
logical program m ing error. In ad d ition, the u ser m ay
interru p t the transaction d u ring its execu tion.

13
 Introduction to Transaction Processing
Why recovery is needed (cont.):
3. Local errors or exception conditions d etected by the
transaction:
- certain cond itions necessitate cancellation of the
transaction. For exam ple, d ata for the transaction m ay
not be fou nd . A cond ition, su ch as insu fficient accou nt
balance in a banking d atabase, m ay cau se a transaction,
su ch as a fu nd w ithd raw al from that accou nt, to be
canceled .
- shou ld be p rogram m ed in the transaction itself.
4. Concurrency control enforcement: The concu rrency
control m ethod m ay d ecid e to abort the transaction, to
be restarted later, becau se it violates serializability or
becau se several transactions are in a state of d ead lock .

14
 Introduction to Transaction Processing

Why recovery is needed (cont.):

5. D isk failure: Som e d isk blocks m ay lose their d ata
becau se of a read or w rite m alfu nction or becau se of a
d isk read / w rite head crash. This m ay happ en d u ring
a read or a w rite operation of the transaction.
6. Physical problems and catastrophes: This refers to an
end less list of problem s that inclu d es p ow er or air-
cond itioning failu re, fire, theft, sabotage, overw riting
d isks or tapes by m istake, and m ou nting of a w rong
tap e by the operator.

15
 Transaction and System Concepts

A transaction is an atom ic u nit of w ork that is either

com pleted in its entirety or not d one at all. For
recovery p u rposes, the system need s to keep track of
w hen the transaction starts, term inates, and com m its
or aborts.
Transaction states:
 Active state
 Partially com m itted state
 Com m itted state
 Failed state
 Term inated State

16
 Transaction and System Concepts
Recovery m anager keeps track of the follow ing
op erations:
 begin_transaction: This m arks the beginning of
transaction execu tion.
 read or w rite: These specify read or w rite operations
on the d atabase item s that are execu ted as p art of a
transaction.
 end_transaction: This specifies that read and w rite
transaction op erations have end ed and m arks the end
p oint of transaction execu tion. At this point it m ay be
necessary to check w hether the changes introd u ced
by the transaction can be perm anently applied to the
d atabase or w hether the transaction has to be aborted
becau se it violates concu rrency control or for som e
other reason.
17
 Transaction and System Concepts

Recovery m anager keeps track of the follow ing operations

(cont):
 commit_transaction: This signals a successful end of the
transaction so that any changes (u pd ates) execu ted by
the transaction can be safely committed to the d atabase
and w ill not be u nd one.
 rollback (or abort): This signals that the transaction has
ended unsuccessfully, so that any changes or effects that
the transaction m ay have app lied to the d atabase m u st
be undone.

18
 Transaction and System Concepts

Recovery techniques use the following operators:

 undo: Sim ilar to rollback excep t that it ap p lies
to a single op eration rather than to a w hole
transaction.
 redo: This sp ecifies that certain transaction
operations m u st be redone to ensu re that all the
op erations of a com m itted transaction have been
ap p lied su ccessfu lly to the d atabase.

19
State transition diagram illustrating the states for
transaction execution.

20
 Transaction and System Concepts

The System Log

 Log or Journal : The log keeps track of all transaction
op erations that affect the valu es of d atabase item s. This
inform ation m ay be need ed to perm it recovery from
transaction failu res. The log is kept on d isk, so it is not
affected by any type of failu re except for d isk or
catastrophic failu re. In ad d ition, the log is p eriod ically
backed u p to archival storage (tape) to gu ard against
su ch catastrophic failu res.
 T in the follow ing d iscu ssion refers to a u niqu e
transaction-id that is generated au tom atically by the
system and is u sed to id entify each transaction:
21
 Transaction and System Concepts
The System Log (cont):
Types of log record:
1. [start_transaction,T]: Record s that transaction T has
started execu tion.
2. [w rite_item ,T,X,old _valu e,new _valu e]: Record s that
transaction T has changed the valu e of d atabase item X
from old _valu e to new _valu e.
3. [read _item ,T,X]: Record s that transaction T has read
the valu e of d atabase item X.
4. [com m it,T]: Record s that transaction T has com p leted
su ccessfu lly, and affirm s that its effect can be
com m itted (record ed p erm anently) to the d atabase.
5. [abort,T]: Record s that transaction T has been aborted .
22
 Transaction and System Concepts
The System Log (cont):
 p rotocols for recovery that avoid cascad ing
rollbacks d o not requ ire that read op erations
be w ritten to the system log, w hereas other
p rotocols requ ire these entries for recovery.
 strict p rotocols requ ire sim p ler w rite entries
that d o not inclu d e new _valu e.

23
 Transaction and System Concepts
Recovery using log records:
If the system crashes, w e can recover to a consistent
d atabase state by exam ining the log and u sing one of
the techniqu es d escribed in later sections.
1. Becau se the log contains a record of every w rite
op eration that changes the valu e of som e d atabase
item , it is p ossible to undo the effect of these w rite
operations of a transaction T by tracing backw ard
throu gh the log and resetting all item s changed by a
w rite op eration of T to their old _valu es.
2. We can also redo the effect of the w rite op erations of a
transaction T by tracing forw ard throu gh the log and
setting all item s changed by a w rite operation of T
(that d id not get d one perm anently) to their
new _valu es. 24
 Transaction and System Concepts
Commit Point of a Transaction:
 D efinition: A transaction T reaches its commit point
w hen all its operations that access the d atabase have
been execu ted su ccessfu lly and the effect of all the
transaction operations on the d atabase has been
record ed in the log. Beyond the com m it p oint, the
transaction is said to be committed, and its effect is
assu m ed to be permanently recorded in the d atabase.
The transaction then w rites an entry [com m it,T] into
the log.
 Roll Back of transactions: N eed ed for transactions
that have a [start_transaction,T] entry into the log bu t
no com m it entry [com m it,T] into the log.

25
 Transaction and System Concepts
Commit Point of a Transaction (cont):
 Redoing transactions: Transactions that have w ritten
their com m it entry in the log m u st also have record ed
all their w rite op erations in the log; otherw ise they
w ou ld not be com m itted , so their effect on the
d atabase can be redone from the log entries. (N otice
that the log file m u st be kept on d isk. At the tim e of a
system crash, only the log entries that have been
written back to disk are consid ered in the recovery
p rocess becau se the contents of m ain m em ory m ay be
lost.)
 Force w riting a log: before a transaction reaches its
com m it p oint, any portion of the log that has not been
w ritten to the d isk yet m u st now be w ritten to the d isk.
This p rocess is called force-w riting the log file before
com m itting a transaction.
26
 Desirable Properties of Transactions

ACID properties:
 Atomicity: A transaction is an atom ic u nit of
p rocessing; it is either p erform ed in its entirety
or not p erform ed at all.

 Consistency preservation: A correct execu tion

of the transaction m u st take the d atabase from
one consistent state to another.

27
 Desirable Properties of Transactions

ACID properties (cont.):

 Isolation: A transaction shou ld not m ake its u p d ates
visible to other transactions u ntil it is com m itted ; this
property, w hen enforced strictly, solves the tem porary
u p d ate p roblem and m akes cascad ing rollbacks of
transactions u nnecessary.
 D urability or permanency: Once a transaction changes
the d atabase and the changes are com m itted , these
changes m u st never be lost becau se of su bsequ ent
failu re.

28
 Characterizing Schedules based on
Recoverability
 Transaction schedule or history: When transactions are
executing concurrently in an interleaved fashion, the order of
execution of operations from the various transactions forms
what is known as a transaction schedule (or history).

 A schedule (or history) S of n transactions T1, T2, ..., Tn:

It is an ordering of the operations of the transactions subject to
the constraint that, for each transaction Ti that participates in
S, the operations of Ti in S must appear in the same order in
which they occur in Ti. Note, however, that operations from
other transactions Tj can be interleaved with the operations of
Ti in S.
29
 Characterizing Schedules based on
Recoverability
Schedules classified on recoverability:
 Recoverable schedule: One where no committed
transaction needs to be rolled back.
A schedule S is recoverable if no transaction T in S commits
until all transactions T’ that have written an item that T reads
have committed.
 Cascadeless schedule: One where every transaction reads
only the items that are written by committed transactions.
Schedules requiring cascaded rollback: A schedule in
which uncommitted transactions that read an item from a
failed transaction must be rolled back.

30
 Characterizing Schedules based on
Recoverability
Schedules classified on recoverability (cont.):
 Strict Schedules: A schedule in which a transaction
can neither read or write an item X until the last
transaction that wrote X has committed.

31
 Characterizing Schedules based on
Serializability
 Serial schedule: A schedule S is serial if, for every
transaction T participating in the schedule, all the
operations of T are executed consecutively in the
schedule. Otherwise, the schedule is called nonserial
schedule. Hence, in a serial schedule, only one
transaction at a time is active-the commit (or abort)
of the active transaction initiates execution of the
next transaction.
 Serializable schedule: A schedule S (possibly
concurrent) is serializable if it is equivalent to some
serial schedule of the same n transactions.
32
 Characterizing Schedules based on
Serializability
 Result equivalent: Two schedules are called result
equivalent if they produce the same final state of the
database.
 Conflict equivalent: Two schedules are said to be
conflict equivalent if the order of any two conflicting
operations (read and write, write and read, and
write and write on the same data item) is the same in
both schedules.
 Conflict serializable: A schedule S is said to be
conflict serializable if it is conflict equivalent to
some serial schedule S’.
33
 Characterizing Schedules based on
Serializability
 Being serializable is not the same as being serial

 Being serializable im p lies that the sched u le is a

correct sched u le.
– It w ill leave the d atabase in a consistent state.
– The interleaving is appropriate and w ill resu lt in a
state as if the transactions w ere serially execu ted , yet
w ill achieve efficiency d u e to concu rrent execu tion.

34
 Characterizing Schedules based on
Serializability

Serializability is hard to check.

– Interleaving of operations occurs in an operating
system through some scheduler
– Difficult to determine beforehand how the
operations in a schedule will be interleaved.

35
 Characterizing Schedules based on
Serializability
Practical approach:
 Come up with methods (protocols) to ensure
serializability.
 It’s not possible to determine when a schedule begins
and when it ends. Hence, we reduce the problem of
checking the whole schedule to checking only a
committed project of the schedule (i.e. operations
from only the committed transactions.)
 Current approach used in most DBMSs:
– Use of locks with two phase locking

36
 Characterizing Schedules based on
Serializability
 View equivalence: A less restrictive definition of
equivalence of schedules

 View serializability: definition of serializability

based on view equivalence. A schedule is view
serializable if it is view equivalent to a serial
schedule.

37
 Characterizing Schedules based on
Serializability
Two schedules are said to be view equivalent if the following
three conditions hold:
1. The same set of transactions participates in S and S’, and S
and S’ include the same operations of those transactions.
2. For any operation Ri(X) of Ti in S, if the value of X read by
the operation has been written by an operation Wj(X) of Tj
(or if it is the original value of X before the schedule started),
the same condition must hold for the value of X read by
operation Ri(X) of Ti in S’.
3. If the operation Wk(Y) of Tk is the last operation to write
item Y in S, then Wk(Y) of Tk must also be the last operation
to write item Y in S’.

38
 Characterizing Schedules based on
Serializability
The premise behind view equivalence:
 As long as each read operation of a transaction reads
the result of the same write operation in both
schedules, the write operations of each transaction
must produce the same results.
 “The view”: the read operations are said to see the
the same view in both schedules.

39
 Characterizing Schedules based on
Serializability
Relationship between view and conflict equivalence:
 The two are same under constrained write
assumption which assumes that if T writes X, it is
constrained by the value of X it read; i.e., new X =
f(old X)
 Conflict serializability is stricter than view
serializability. With unconstrained write (or blind
write), a schedule that is view serializable is not
necessarily conflict serialiable.
 Any conflict serializable schedule is also view
serializable, but not vice versa.
40
 Characterizing Schedules based on
Serializability
Relationship between view and conflict equivalence
(cont):
Consider the following schedule of three transactions
T1: r1(X), w1(X); T2: w2(X); and T3: w3(X):
Schedule Sa: r1(X); w2(X); w1(X); w3(X); c1; c2; c3;

In Sa, the operations w2(X) and w3(X) are blind writes, since T1
and T3 do not read the value of X.

Sa is view serializable, since it is view equivalent to the serial

schedule T1, T2, T3. However, Sa is not conflict serializable,
since it is not conflict equivalent to any serial schedule.
41
 Characterizing Schedules based on
Serializability
Testing for conflict serializability
Algorithm:
1. Looks at only read_Item (X) and write_Item (X) operations
2. Constructs a precedence graph (serialization graph) - a graph
with directed edges
3. An edge is created from Ti to Tj if one of the operations in Ti
appears before a conflicting operation in Tj
4. The schedule is serializable if and only if the precedence graph
has no cycles.

42
 FIGURE
Example of serializability testing. (a) The READ
and WRITE operations of three transactions T1, T2,
and T3.

43
 FIGURE (continued)
Example of serializability testing. (b) Schedule E.

44
 FIGURE (continued)
Another example of serializability testing. Precedence graph
for Schedule E.

45
 FIGURE (continued)
Example of serializability testing. (c) Schedule F.

46
 FIGURE (continued)
Another example of serializability testing. Precedence graph
for Schedule F.

47
 Characterizing Schedules based on
Serializability
Other Types of Equivalence of Schedules
 Under special semantic constraints, schedules that
are otherwise not conflict serializable may work
correctly. Using commutative operations of addition
and subtraction (which can be done in any order)
certain non-serializable transactions may work
correctly

48
 Characterizing Schedules based on
Serializability
Other Types of Equivalence of Schedules(cont.)
Example: bank cred it / d ebit transactions on a given item are
separable and commutative.
Consid er the follow ing sched u le S for the tw o transactions:
Sh : r1(X); w 1(X); r2(Y); w 2(Y); r1(Y); w 1(Y); r2(X); w 2(X);
Using conflict serializability, it is not serializable.
H ow ever, if it cam e from a (read ,u pd ate, w rite) sequ ence as
follow s:
r1(X); X := X – 10; w 1(X); r2(Y); Y := Y – 20;r1(Y);
Y := Y + 10; w 1(Y); r2(X); X := X + 20; (X);
Sequ ence explanation: d ebit, d ebit, cred it, cred it.
It is a correct schedule for the given semantics
49
 Transaction Su pport in SQL2

 A single SQL statement is always considered to

be atomic. Either the statement completes
execution without error or it fails and leaves the
database unchanged.
 With SQL, there is no explicit Begin Transaction
statement. Transaction initiation is done
implicitly when particular SQL statements are
encountered.
 Every transaction must have an explicit end
statement, which is either a COMMIT or
ROLLBACK.
50
 Transaction Su pport in SQL2

Characteristics specified by a SET

TRANSACTION statement in SQL2:
 Access mode: READ ONLY or READ WRITE. The
default is READ WRITE unless the isolation level of
READ UNCOMITTED is specified, in which case
READ ONLY is assumed.
 Diagnostic size n, specifies an integer value n,
indicating the number of conditions that can be held
simultaneously in the diagnostic area. (Supply user
feedback information)

51
 Transaction Su pport in SQL2

Characteristics specified by a SET

TRANSACTION statement in SQL2 (cont.):
 Isolation level <isolation>, where <isolation> can be
READ UNCOMMITTED, READ COMMITTED,
REPEATABLE READ or SERIALIZABLE. The
default is SERIALIZABLE.
With SERIALIZABLE: the interleaved execu tion of
transactions w ill ad here to ou r notion of
serializability. H ow ever, if any transaction execu tes
at a low er level, then serializability m ay be violated .

52
 Transaction Su pport in SQL2
Potential problem with lower isolation levels:
 Dirty Read: Reading a value that was written by a
transaction which failed.
 Nonrepeatable Read: Allowing another transaction to
write a new value between multiple reads of one
transaction.
A transaction T1 may read a given value from a table.
If another transaction T2 later updates that value and
T1 reads that value again, T1 will see a different value.
Consider that T1 reads the employee salary for Smith.
Next, T2 updates the salary for Smith. If T1 reads
Smith's salary again, then it will see a different value for
Smith's salary.
53
 Transaction Su pport in SQL2

Potential problem with lower isolation levels

(cont.):
 Phantoms: New rows being read using the same read
with a condition.
A transaction T1 m ay read a set of row s from a
table, p erhap s based on som e cond ition sp ecified
in the SQL WH ERE clau se. N ow su ppose that a
transaction T2 inserts a new row that also satisfies
the WH ERE clau se cond ition of T1, into the table
u sed by T1. If T1 is rep eated , then T1 w ill see a row
that previou sly d id not exist, called a phantom.

54
 Transaction Su pport in SQL2
Sample SQL transaction:
EXEC SQL whenever sqlerror go to UNDO;
EXEC SQL SET TRANSACTION
READ WRITE
DIAGNOSTICS SIZE 5
ISOLATION LEVEL SERIALIZABLE;
EXEC SQL INSERT
INTO EMPLOYEE (FNAME, LNAME, SSN, DNO, SALARY)
VALUES ('Robert','Smith','991004321',2,35000);
EXEC SQL UPDATE EMPLOYEE
SET SALARY = SALARY * 1.1
WHERE DNO = 2;
EXEC SQL COMMIT;
GOTO THE_END;
UNDO: EXEC SQL ROLLBACK;
THE_END: ...

55