Draw a precedence graph for the following schedule that will represent any conflicting pairs of operations
computer science
Description
Question 1:
i.
Draw a
precedence graph for the following schedule that will represent any conflicting
pairs of operations where for R1(X), R is a read operation, 1
is the transaction T1, and X is the data item X (so in
this case the operation is
"Transaction T1 Reads X"): R1(Y)W3(X)R2(Z)R1(Z)W4(Y)W4(X)W3(Z)
ii.
Is this
schedule conflict serializable? If not, what are the possible serial orderings
that it is equivalent to?
iii.
The loosest
form of the four standard transaction isolation levels is READ UNCOMMITTED,
while the strictest is SERIALIZABLE.
Compare and contrast these levels with reference to problems that you
may face with either.
Question 2: A
locking approach such as shared/exclusive locking, on its own does not
guarantee conflict serializability.
i.
Describe an
example of a transaction schedule where READ(X), WRITE(X) as well as
READ_LOCK(X), WRITE_LOCK(X) and UNLOCK(X) operations are indicated for each
transaction involved, showing and explaining how things can go wrong without
the 2PL growing-shrinking ordering requirement.
ii.
Define the
terms referred to by the ACID acronym and for each property provide a
high-level example of what can go wrong without it (e.g. an example of
what might happen in a database system if transactions are NOT run as
atomic units (Atomicity), etc.
Question
3: Consider the following sequence
of 10 interleaved operations by two transactions T1 and T2 (assume
no other transactions are running). Data item X begins with the value 40
1.
T1 START TRANSACTION
2.
T2 START TRANSACTION
3.
T1 X = READ(X)
4.
T2 X = READ(X)
5.
T1 X = X + 30
6.
T2 X = X - 5
7.
T1 WRITE(X)
8.
T2 WRITE(X)
9.
T1 COMMIT
10.
T2 COMMIT
Outline what you expect will happen if
these attempts to execute in the following environments:
·
2PL where every
READ gets a shared lock.
·
2PL but with
non-locking reads (e.g. with MySQL MVCC-based default settings)
