Recoverable Schedule (ACID)

acid

Definition

Recoverable Schedule (ACID)

A schedule is recoverable if for each pair of transactions $T_i,T_j$, where $T_j$ reads data items written by $T_i$, $T_i$ must commit before $T_j$.

Examples

Recoverable Schedule Example

Irrecoverable Schedule Example

Time	$T_1$	$T_2$	$A$
1	write($A$)		100 (uncomitted)
2		read($A$)	100
3		commit
4	abort

In this scenario:

1. $T_1$ writes a value to data item $A$. This value is not yet committed and resides in a memory buffer.
2. $T_2$ reads the uncommitted value of $A$ written by $T_1$. This is a dirty read.
3. $T_2$ then commits its changes. Since $T_2$ has committed, its results are now considered permanent.
4. Subsequently, $T_1$ fails for some reason and must be aborted. The system needs to roll back the changes made by $T_1$.

The problem here is that $T_2$, which has already committed, based its operations on a value from transaction $T_1$ that never successfully completed. The value of $A$ read by $T_2$ is now effectively a “ghost” value. Since $T_2$ has committed, we cannot simply roll it back. This leads to an inconsistent database state.