Matakuliah
Tahun
Versi
: M0184 / Pengolahan Data Distribusi
: 2005
:
Session – 12
CONCURRENCY CONTROL
USER ACCESS CONTROL
OBJECTIVES
• The problem around user access control
(concurrency transaction)
• User access control – schedules and
serialization
Problem in Concurrent
transaction
• Lost Update problem
• Violation of integrity constraint
• Inconsistent retrieval problem
Lost Update problem
Begin transaction T1
read balance(x)
balance(x) = balance(x)-100
Time
if balance(x) < 0 then
print “No Fund!”
abort T1
end if
write balance(x)
write balance(y)
balance(y) = balance(y)+100
write balance(y)
Commit T1
Begin transaction T2
read balance(x)
balance(x) = balance(x) + 100
write balance(x)
Commit T2
Note :
X = 100
The increase of X to 200 by T2
will be overwritten by the
decrement to ) by T1, thereby
losing 100
Violation by Integrity
Constraint
• When two transaction
are allowed to
execute concurrently
without being
synchronized.
SURGEON
SName
Mary
Tom
Operation
Tonsillectomy
Tonsillectomy
Mary
Appendectomy
SCHEDULE
SName Operation
Mary
……..
Date
Tonsillectomy 04.04.200
5
……………
……….
Violation by Integrity Constraint
Begin transaction T3
read SCHEDULE where date= 04.04.2005
read SURGEON where
SURGEON.SName =
SCHEDULE.SName and
SURGEON.Operation = “Appendectomy”
if not found then abort T3
SCHEDULE.operation = “Appendectomy”
Commit T3
Begin transaction T4
read SCHEDULE where date= 04.04.2005
read SURGEON where
SURGEON.SName = “Tom” And
SURGEON.Operation = SCHEDULE.Operat
if not found then abort T4
SCHEDULE.SName = “Tom”
Commit T4
Note :T3 changes the operation schedule on 04.04.2005 from a
tonsillectomy to an appendectomy. T4 changes the surgeon
assigned to 04.04.2005 to Tom.
The effect of these two transaction is to produce database state
which is inconsistent. Tom now is qualified to operate on
04.04.2005 and perform appendectomy, which is not qualified to do
so, as shown in tables before.
Inconsistent Retrieval Problem
Begin transaction T1
read balance(x)
balance(x) = balance(x)-100
if balance(x) < 0 then
print “No Fund!”
abort T1
end if
write balance(x)
write balance(y)
balance(y) = balance(y)+100
write balance(y)
Commit T1
Begin Transaction T5
sum = 0
do while not end of relation
read balance(a)
sum = sum + balance(a)
::::
::::
read balance(x)
sum = sum + balance(x)
::::
::::
read balance(y)
sum = sum + balance(y)
::
Commit T5
Schedule and Serialization
• A transaction consists pf a sequence of Reads
and Writes to the database. The entire sequence
of reads and writes by all concurrent
transactions in database taken together is
known as Schedule. A schedule S is generally
written :
S = [O1, O2, O3,O4,……,On]
O : indicate Read or Write operation
Schedule and Serialization
Cont’d
• A Serial Schedule is one in which all read
and write of each transaction are grouped
together so that the transaction are run
sequentially one after the other.
EXAMPLE
T6
T7
Begin Transaction
T7
Read X
Time
Begin Transaction
T6
read Y
Y=Y+1
write Y
Commit T6
Read Y
Y=Y+1
Write Y
Commit T7
T8
Begin transaction T8
read X
X=X+1
write X
Commit T8
EXAMPLE Cont’d
• The Schedule S for that example :
S=
[R7(X),R8(X),W8(X),R6(Y),W6(Y),R7(Y),W7(Y)]
• Serial Schedule SR :
SR =
[R6(Y),W6(Y),R7(X),R7(Y),W7(Y),R8(X),W8(X)]