Two Phase Locking Protocol

What is two phase locking?

Two-phase locking is a concurrency control method in a database and transaction processing system. It ensures that transactions are executed in a serializable manner by requiring that a transaction must obtain a lock on all of its data before it can modify them.

A transaction is called that it is following the Two-Phase Locking protocol if Locking and Unlocking can be done in two phases. These two phases are the growing phase, where the transaction acquires locks, and the shrinking phase, where it releases locks. This mechanism ensures that transactions are executed atomically and ensures data consistency.

Now let’s see both phases in detail:

Growing phase

In this phase, transactions acquire locks on the data they need to access. Locks can either be shared (read) locks or exclusive (write) locks. The growing phase ensures that transactions are executed in a serializable manner and ensures data consistency.

Shrinking phase

In this phase, transactions release all the locks they have acquired. The shrinking phase allows other transactions to access the data that was previously locked and ensures that resources are freed up for other transactions.

Types of two phase locking

There are three types of two phase locking:

Strict two-phase locking

This type of two-phase locking requires transactions to follow a strict locking protocol where locks must be acquired in a specific order and released in the reverse order.

The locks are held until the end of the transaction, and no transactions can modify the data until the lock is released.

Strict two-phase locking ensures the highest level of data consistency, but it can result in deadlocks if transactions request conflicting locks.

Conservative two-phase locking

This type of two-phase locking is less restrictive than strict two-phase locking and allows transactions to release locks before the end of the transaction.

The locks can be released if the transaction determines that it no longer needs to access the data.

This type of two-phase locking is less prone to deadlocks, but it can result in lower levels of data consistency.

Rigorous Two-Phase Locking

Rigorous Two-Phase Locking (RTPL) is a type of two-phase locking protocol that is used in database management systems and transaction processing systems.

RTPL is a strict implementation of two-phase locking that requires transactions to obtain all the locks they need before they can start modifying the data.

Once the locks are acquired, they are held until the end of the transaction, and no other transaction can modify the same data

Cascading Roll back in two phase locking

Cascading rollback is a phenomenon that can occur in Two-Phase Locking (2PL) systems, where a transaction's failure to commit causes the failure of other transactions that depend on it.

If a transaction in a 2PL system encounters an error or deadlock, it may need to be rolled back, which means that all the changes made by the transaction are undone. If other transactions have already acquired locks on the data that was modified by the transaction, they will also need to be rolled back to maintain data consistency.

Cascading rollback can result in a chain reaction, where the failure of one transaction causes the failure of multiple transactions, causing multiple rollbacks.

Example:

Here is an example of cascading rollback in a Two-Phase Locking (2PL) system:

Suppose there are three transactions T1, T2, and T3, and they are all accessing the same data.

1. T1 starts and acquires an exclusive lock on data A.
2. T2 starts and tries to acquire a lock on data A, but it is blocked by T1's lock.
3. T3 starts and acquires a lock on data B.
4. T2 acquires a lock on data B.
5. T1 encounters an error and needs to be rolled back. As a result, T1 releases its lock on data A and all the changes made by T1 are undone.
6. T2, which was dependent on T1's changes to data A, must also be rolled back because it now has a conflicting lock on data B.
7. T3, which had acquired a lock on data B, must also be rolled back because its changes were based on the state of data B before T2 acquired a lock on it.

In this example, the failure of T1 caused a chain reaction, resulting in the cascading rollback of T2 and T3. This can lead to a significant overhead in terms of time and resources and can also cause performance degradation in the system.

To avoid cascading rollback, 2PL systems typically implement various concurrency control and deadlock detection algorithms to prevent deadlocks from occurring.

Deadlock in Two-Phase Locking

Deadlock is a situation that can occur in Two-Phase Locking (2PL) systems, where two or more transactions are blocked and waiting for each other to release locks, resulting in a standstill.

In a 2PL system, transactions must obtain locks on the data they need to access before they can start modifying it. If two transactions request conflicting locks, one of them may need to wait for the other to release its lock before it can proceed.

Deadlocks can occur in 2PL systems when two or more transactions hold locks on data that the other transactions need to access.

For example, suppose two transactions T1 and T2 are executing concurrently and they both need to access data A and B. If T1 acquires a lock on data A and then tries to acquire a lock on data B, and T2 acquires a lock on data B and then tries to acquire a lock on data A, a deadlock will occur because T1 and T2 are waiting for each other to release their locks.