Understanding Distributed Transactions, Consistency Models, and Commit Protocols

Distributed transactions and distributed locks are challenging topics; this article starts from basic concepts, explaining transactions and the ACID properties (Atomicity, Consistency, Isolation, Durability) with banking examples.

It then discusses consistency models—strong, weak, and eventual consistency—and how sharding (vertical and horizontal) impacts transaction handling, introducing the problems that arise when data is split across multiple databases.

The article reviews the CAP theorem and BASE theory as theoretical foundations for distributed systems, highlighting the trade‑offs between consistency, availability, and partition tolerance.

Next, it details the Two‑Phase Commit (2PC) protocol, describing its voting and commit phases, the role of the coordinator and participants, and its drawbacks such as single‑point failure, synchronous blocking, and potential data inconsistency.

It proceeds to the Three‑Phase Commit (3PC) protocol, introducing a pre‑check phase and timeout strategies that reduce blocking time and improve reliability, while noting that efficiency concerns still limit its adoption.

Finally, the Try‑Confirm‑Cancel (TCC) pattern is presented as a service‑level alternative to 2PC, outlining its three steps (resource reservation, confirmation, and cancellation), workflow, and advantages for high‑performance systems.