Question

Which properties ensure that either all or none action will be carried on databasr?

Answer 1

In order to perform a Transaction in a database system and to make sure it works without any issues, there are few rules a Database Transaction should follow. These rules are the standards across all Relational Database systems (RDBMS) and are called ACID rules.
 
ACID stands for Atomicity, Consistency, Isolation and Durability. So let’s check what all these Rules states.

Check the details here for ACID properties of Transactions in an RDBMS.

–> A: Atomicity states that every Transaction should be atomic in nature. A Transaction in a Relational Database can contain either a single SQL statement or multiple SQL statements. Thus by Atomic Transaction it means “all or none”. Either all SQL statements/steps execute successfully in a transaction, or fail as a single unit and none of them should be treated as executed and the system should be returned to its original state.

For example: If account-A & account-B both having $2000 balance, you have to transfer $1000 from account-A to account-B, this will involves 2 steps. First withdrawal from account-A, and Second deposit in account-B. Thus, both the steps should be treated as single or atomic unit and at the end account-A should have $1000 & account-B should have $3000 balance. If in case after First step the system fails or any error occurs then first step should also be rolled-back and $1000 withdrawn from account-A should be re-deposited to it, maintaining $2000 back in both the accounts. Thus there should be no intermediate state where account-A has $1000 and account-B still has $2000 balance.

–> C: Consistency states that any Transaction happened in a database will take it from one consistent state to another consistent state. The data finally recorded in the database must be valid according to the defined Rules, Constraints, Cascades, Triggers, etc. If in case of any failure to these rules the changes made by any transaction should be rolled-back, this will put the system in earlier consistent state.

For example: If the money deposit process has any Trigger built on top of it. And at the time of money transfer any of the Trigger fails or any database node, the system should automatically Rollback the complete transaction and switch back the system to its previous consistent state before the transaction was started. Or if everything executes successfully then the system is committed to a new consistent state.

–> I: Isolation means Transactions performing same functions should run in Isolation and not in parallel to provide more concurrency to the data and avoiding dirty reads & writes. One need to use proper Transaction Isolation levels and locking in order to prevent this.

For example: If two people accessing a joint-account with $5000 balance from 2 terminals to withdraw money. Let’s say at same time John & Marry apply to withdraw $4000 from two different ATMs. If both the Transactions do not run in Isolation and run in parallel then both John & Marry will be able to withdraw $4000 each i.e. $8000 total from their account. To make sure this won’t happen Transactions should be not allowed to run in parallel, by setting Transaction Isolations and/or locking methods on the database objects.

–> D: Durability, a transaction should be durable by storing the data permanently and making it available in case of power failure, recovery from system failure, crash, any error, etc. All in all, the data should not get lost in any of the miss-happenings and one should be able to recover data from restore, logging and other methods.