Enhancement of database access performance by improving data consistency in a non-relational database system (NoSQL)

TL;DR

This paper proposes a four-stage server-side model to improve data consistency and performance in MongoDB and other NoSQL databases, demonstrating significant throughput and latency improvements under update-heavy workloads.

Contribution

It introduces a novel four-stage server-side transaction management model tailored for MongoDB, enhancing data consistency and performance in NoSQL systems.

Findings

Improved throughput to 2864.726 ms from 2914.110 ms

Reduced read latency to 32806.275 ms from 26510.930 ms

Enhanced update latency to 51845.629 ms from 32457.662 ms

Abstract

This study aims to enhance data consistency in NoSQL databases, traditionally designed with BASE properties, as opposed to the strong consistency guaranteed by ACID principles in RDBMS. We introduce a comprehensive four-stage server-side model engineered explicitly for MongoDB. This model covers transaction management, bifurcation of read and write transactions, assessment of transaction readiness, and transaction execution via a specialized locking algorithm. Utilizing the Yahoo Cloud Services Benchmark (YCSB), particularly for update-heavy workloads (A, B, and F), our model exhibited significant improvements. Specifically, the average throughput, read, and update latencies improved to 2864.726 ms, 32806.275 ms, and 51845.629 ms, respectively, from the baseline metrics of 2914.110 ms, 26510.930 ms, and 32457.662 ms. These results demonstrate the efficacy of our proposed model in enhancing consistency not only in document-based NoSQL databases like MongoDB but also in other NoSQL database variants, including key-value, graph, and wide-column stores.