What is a distributed ACID-compliant database?

A distributed ACID-compliant database is a system that ensures transactional reliability (ACID) across multiple servers or nodes spread over a network. ACID stands for Atomicity, Consistency, Isolation, and Durability. Atomicity guarantees that a transaction either fully completes or fails without partial changes. Consistency ensures data remains valid according to defined rules before and after transactions. Isolation prevents concurrent transactions from interfering with each other, and Durability ensures committed data survives failures. In a distributed setup, these properties must hold even when data is partitioned or replicated across locations, which introduces challenges like network delays and node failures.

To achieve ACID compliance in a distributed environment, databases use protocols like two-phase commit (2PC) or consensus algorithms such as Raft or Paxos. For example, Google Spanner employs a combination of synchronized clocks (TrueTime) and distributed locking to maintain consistency globally. CockroachDB uses Raft consensus to replicate data across nodes, ensuring atomicity and durability even if some nodes fail. These systems coordinate nodes to agree on transaction outcomes before committing, ensuring atomicity. Consistency is maintained through strict validation of data rules across replicas, while isolation often relies on techniques like timestamp ordering or multi-version concurrency control (MVCC) to manage concurrent access without locks.

Such databases are critical for applications requiring high reliability and global scalability, like financial systems or inventory management. For instance, a banking app using a distributed ACID database can process transactions across regions while ensuring balances remain accurate and overdrafts are prevented. However, enforcing ACID in distributed systems can trade off latency, as coordinating nodes introduces delays. Developers must weigh these trade-offs: systems like Amazon Aurora optimize for specific scenarios (e.g., read-heavy workloads) by relaxing some isolation levels. Choosing a distributed ACID database depends on whether strict data correctness justifies the complexity and potential performance costs.