Unit 6 Databases, Serialisability, Transactions and 2 Phase Locking

Question 1

What is a database?

Answer 1

A database is an organised collection of data, and is said to be a form of persistent storage.

Question 2

What are the (ACID) desirable properties of a transaction?

Answer 2

Atomicity - either all steps occur, or none at all
Consistency - The system remains in a consistent state
Isolation - the transaction cannot reveal any part of its state til the transaction is committed.
Durability - once committed the result will persist.

Question 3

What are the methods of UserTransaction?

Answer 3

begin(), commit(), rollback()

Question 4

What are the states of a Transaction?

Answer 4

ACTIVE - initial state and while it is executing
PARTIALLY COMMITTED - the state after the final operation has been executed. It remains here while the system writes information to disk.
COMMITTED - the state after successful completion.
FAILED - when normal exection cannot proceed
ABORTED - following rollback() ie all effects are undone and the database is restored to its original state.

Question 5

Compare the concept of a transaction with that of a coarse grained atomic action. What are their differences and similarities?

Answer 5

Both are as series of steps that need to be implemented ‘as one’. They give the appearance of indivisibility.
The difference is that a transaction has FAILURE ATOMICITY (if it fails it is as though it never happened in the first place, if it succeeds it is made persistent).

Question 6

What are the possible next states following PARTIALLY COMMITTED for a transaction?

Answer 6

It can progress to COMMITTED once the write completes and the data is made persistent.
Or if there is a problem it will enter FAILED and then ABORTED.

Question 7

What does the TP system do?

Answer 7

The TP (Transaction Processing) system manages the correct and efficient execution of transactions.

Question 8

Serialisability - What is a schedule?

Answer 8

A schedule is a preserved, chronological order of operations.

Question 9

Serialisibility - What is a serial schedule?

Answer 9

A serial schedule is one where the transactions are executed in strict serial order.

Question 10

Give the definition of serialisibility.

Answer 10

Serialisability is a schedule for a group of transactions that behaves as though it were executed in serial order. ie the results are the same

Question 11

Is it a problem if serial schedules do not give the same final results for the data object involved?

Answer 11

No it is not. The main concern is whether the data objects are in a consistent state.

Question 12

What are conflicting operations?

Answer 12

Operations conflict if the order in which they are executed affects the result (e.g read/write)

Question 13

How is the notion of conflicting operations used to determine whether a schedule is serialisable?

Answer 13

We need to determine what the conflicting operations are. Then we need to determine the order they are executed by transactions. If the oPrrder is the same then we know the schedule is equivalent to a serial schedule..

Question 14

Precedence graphs can be used to show conflicting operations of transactions. What would the presence of a cycle in a precedence graph mean?

Answer 14

It would mean the schedule it represents is not serialisable.

Question 15

Why is it important to make schedules of multiple concurrent transactions serialisable?

Answer 15

To keep the transaction ISOLATION property intact (ACID).

Question 16

What is the aim of two-phase locking?

Answer 16

To ensure the effects of the transactions on shared data is serially equivalant.

Question 17

How does two phase locking work?

Answer 17

A transaction is either in acquire phase or release phase. Once the transaction starts to release it cannot acquire any further locks.

Question 18

What is the difference between strict 2PL and 2PL?

Answer 18

In the strict form all locks are released on commit.

In the non-strict form the locks are gradually released

Question 19

What are the two classes of concurrency control mechanism?

Answer 19

Pessimistic algorithms assume there are likely to be conflicts and take measures to avoid them.
Optimistic algorithms assume conflicts will be rare, and can dealt with after the event.

Question 20

Name two types of pessimistic algorithm

Answer 20

Two-phase locking and Time Stamp Ordering.

Question 21

How does Time Stamp ordering work as an algorithm of concurrency control?

Answer 21

Each object has a time stamp associated with the transaction that accessed it most recently. If another transaction requests an operation that conflicts with one that has been carried out previously on this object the time stamp is compared. If the transaction’s time stamp is earlier than the one recorded by the object the transaction must abort and restart with a later timestamp.
Basically the time stamp cannot go backwards.

Question 22

Name some advantages of Time Stamp Ordering.

Answer 22

it is efficient and has fewer deadlock problems than locking.
It is simple - few overheads.

Question 23

Name some disadvantages of time-stamp ordering.

Answer 23

It can suffer from high levels of aborts.

May be more restrictive than 2PL.

Question 24

What is the main goal of optimistic concurrency control?

Answer 24

To avoid the overheads of locking managment, deadlock prevention or detection.

Question 25

When would you use an optimistic concurrency control algorithm?

Answer 25

When the likelihood of clients accessing the same data is low, and a fast response is important.

Question 26

In 2PL what is a shared lock and why might we want to use one?

Answer 26

A shared lock is useful for non-conflicting operations, such as two read operations. If a shared lock can be used it increases concurrency.

Question 27

What is a simple alternative to deadlock detection?

Answer 27

Timeouts.

Question 28

What are cascading aborts and what causes them?

Answer 28

Cascading aborts occur when one transaction aborts and this leads to other transactions having to abort. It means work has to be repeated.

Question 29

Distinguish between volatile memory, persistent storage and stable storage.

Answer 29

Volatile memory would be lost if the device is switched off. Persistent storage remains after the device is switched off. Stable storage is stored at least twice for added security.

Question 30

Which of the ACID properties of transactions are addressed through crash resilience?

Answer 30

Atomicity - transactions are all or nothing | Durability - crash resilience ensures that if the transaction has committed its results are stored in persistent memory.

Question 31

What is meant by rolling back.

Answer 31

In the even that a transaction cannot happen then the system is restored to the state is was in prior to the transaction starting.

Question 32

In Logging, what might happen if log information is written to the persistent store after the data is updated by the atomic operation is written there?

Answer 32

There would be no record of the data in the event a roll back was required. Logging should be 'write-ahead'.

Question 33

What is shadowing? What essential feature is required to ensure that shadowing is implemented successfully?

Answer 33

Shadowing is where a copy of the original state of the data is kept and another copy is updated. Once the updated copy is committed, then the original copy can be deleted. The essential feature is that the replacement of the relevant part of the persistent store is done atomically.

Question 34

What is the purpose of JDBC?

Answer 34

Provides a standard interface for communication with a large number of databases.

Question 35

What are the steps involved in setting up a connection between a Java application and a database using JDBC?

Answer 35

1. The application causes the database driver to be loaded. 2. The application requests a connection for the particular database and the driver manager requests all registered drivers to respond with a connection. 3. The driver manager returns a connection to the application. 4. communication through the JDBC connection to the database.