/ˌaɪ.səˈleɪ.ʃən/
noun — "operations shielded from external interference."
Isolation is a property of transactions in computing and database systems that ensures concurrent transactions execute independently without undesired interaction. Each transaction appears to operate in isolation from others, preventing phenomena such as dirty reads, non-repeatable reads, and phantom reads. This property preserves data consistency and integrity in multi-user or multi-process environments.
Technically, isolation is one of the four ACID properties (Atomicity, Consistency, Isolation, Durability) that define reliable transactions. Database management systems implement isolation through locking mechanisms, multi-version concurrency control (MVCC), or serialization strategies. Different isolation levels—such as Read Uncommitted, Read Committed, Repeatable Read, and Serializable—offer trade-offs between consistency guarantees and performance.
In workflow terms, consider two concurrent bank transactions: one transferring funds from Account A to B, and another calculating interest on Account B. Isolation ensures that each transaction sees a consistent view of Account B. The interest calculation cannot observe partial updates from the transfer, preventing incorrect balances.
At a lower level, isolation also applies to threads or processes manipulating shared memory. Atomic operations, mutexes, and semaphores enforce temporary isolation, preventing race conditions and maintaining predictable behavior.
Conceptually, isolation acts like a private workspace: every transaction or operation executes in its own bubble, invisible to others until it completes, ensuring integrity and consistency across the system.
See ACID, Atomicity, Concurrency.