Fundamentals of Distributed Systems: CAP Theory, ACID, BASE, Idempotency, and Distributed Transaction Protocols

CAP Theory : Consistency means all replicas hold the same state at any moment; Availability ensures the system responds correctly even when some nodes fail; Partition tolerance guarantees service continuity despite network partitions.

Local Transaction ACID Properties : Atomicity – all operations succeed or none; Consistency – data satisfies integrity constraints before and after; Isolation – concurrent transactions do not interfere; Durability – once committed, data persists.

BASE Theory : Basically Available – core services remain functional during failures; Soft State – data may be in intermediate states without affecting overall availability; Eventually Consistent – data will become consistent after a bounded time. BASE sacrifices some strong consistency to gain higher availability compared to ACID.

Idempotency Design : An operation is idempotent if repeated executions with the same parameters produce the same result, e.g., calling orderProcess() multiple times deducts inventory only once and returns true.

Distributed Transaction Classification – Two‑Phase Commit (2PC) and Three‑Phase Commit (3PC) : 3PC adds a timeout mechanism for both coordinator and participants and inserts a preparation phase between the first two phases, reducing resource locking and blocking time. However, it incurs performance overhead and can cause long lock durations.

Compensation Transactions – TCC and Saga : Both use a compensation (undo) operation for each action, requiring idempotent design. They simplify error handling but increase code size and reduce maintainability.

Message Transactions : Consistency is achieved by wrapping local operations and message sending in a single transaction, ensuring both succeed or fail together. Downstream services consume the message to perform corresponding actions, achieving eventual consistency. Example product: RocketMQ.

Distributed Transaction Product Frameworks

JDTS : Provides high‑availability services where any node can handle requests; supports horizontal and vertical scaling.

Seata : Open‑source solution offering AT, TCC, SAGA, and XA modes for a one‑stop distributed transaction platform.

Global Transaction Service (GTS) : Enables high‑performance transaction consistency across micro‑services, integrating with databases (RDS, MySQL, PostgreSQL), RPC frameworks (Spring Cloud, Dubbo, HSF), and message queues to support distributed, multi‑database, message, and service‑chain transactions.