Enhanced Commit Protocols for Low Latency Distributed Transactions

Abstract

Distributed transaction processing relies on commit protocols to ensure atomicity and consistency across multiple nodes. Conventional commit mechanisms such as the two phase commit protocol introduce significant latency due to their coordination intensive design, multiple communication rounds, and prolonged lock holding during commit execution. As distributed systems scale and transaction workloads increase, these factors collectively contribute to higher commit delays, reduced concurrency, and increased sensitivity to network variability. The blocking nature of coordinator driven commit execution further amplifies these challenges, particularly under moderate to high contention scenarios where participants must wait for global decisions before releasing resources. Existing commit protocols are designed with correctness as the primary objective, often at the cost of performance efficiency. These characteristics limit the ability of distributed transactional systems to meet low latency requirements, especially in environments where rapid transaction completion and high throughput are essential. This work focuses on addressing these limitations by examining protocol level enhancements aimed at reducing commit latency while preserving transactional correctness. By restructuring coordination flows and refining decision propagation mechanisms, the work seeks to establish a commit protocol design that better aligns with the performance demands of modern distributed transactional systems. The study emphasizes a systematic evaluation framework to analyze commit behavior across varying cluster sizes, enabling an understanding of how protocol design choices influence latency, coordination cost, and scalability. The objective is to demonstrate that commit latency can be effectively reduced through protocol optimization rather than application level adjustments, providing a foundation for designing efficient commit mechanisms suitable for large scale distributed transaction processing environments.