An adaptive consistency and parallelism model in distributed databases

Abstract

This paper proposes and investigates a novel model for adaptive management of consistency and parallel processing in distributed databases for high-load information systems. The goal of the model is to overcome the fundamental trade-off between ensuring a high level of data consistency (achieved through increased quorum parameters and synchronization among replicas) and the requirements for low latency and high system throughput under intensive parallel workloads. The proposed model formalizes a dynamic mechanism for adapting key parameters of a distributed system, in particular consistency levels and the degree of parallelism, based on continuous monitoring and analysis of the system state. The model implements a closed-loop adaptive control cycle and is designed using a microservices architecture with containerization, which ensures scalability and flexibility of configuration. Experimental results demonstrate that the model reduces the average request processing latency by 20–40% and increases throughput by 15–30% through adaptive parallelism control, while maintaining a stable level of data consistency, reflected in a reduced conflict rate under dynamically changing workloads. The proposed approach has practical significance for optimizing the performance and reliability of modern information systems, particularly in domains such as financial technologies, e-commerce, the Internet of Things, and cloud services.

Problems in programming 2026; 2: 28-36