With the emergence of NoSQL databases, many large applications have migrated from relational databases (RDB) due to their superior flexibility and performance. Database migration from RDB to NoSQL databases involves schema transformation and data migration, which is not straightforward. The challenge lies in that RDB stores data in normalized form, whereas NoSQL supports denormalization. To address the challenge of schema transformation, this paper proposes a model called query-based denormalization using hypergraph (QBDNH) from RDB to the NoSQL database. The model takes the inputs from existing relational tables and queries and transforms them into the denormalized NoSQL model using hypergraphs. The approach overcomes limitations like complex relationship representation and data access pattern coverage of existing graph-based denormalization techniques. The proposed model reduces the overall time, cost, and effort needed to transform the schema manually. To validate the effectiveness of QBDNH, the experiments are conducted on the TPC-H dataset, and the performance of QBDNH is compared to existing graph-based denormalization models such as TLD, CLDA, and Kuszera. The evaluation is carried out in two parts: the first part analyzed the query speedup factor, while the second part measured efficiency improvement based on query pipeline execution. The results revealed that QBDNH achieved a notable query performance improvement with speedup factors of 1.29, 1.35, and 1.40 compared to existing TLD, CLDA, and Kuszera models. Furthermore, QBDNH significantly enhanced pipeline utilization compared to TLD and Kuszera.

随着NoSQL数据库的出现，许多大型应用程序由于其卓越的灵活性和性能而从关系数据库（RDB）迁移。数据库从RDB迁移到NoSQL数据库涉及到模式转换和数据迁移，这并不简单。挑战在于 RDB 以规范化形式存储数据，而 NoSQL 支持非规范化。为了解决模式转换的挑战，本文提出了一种从 RDB 到 NoSQL 数据库的基于查询的反规范化使用超图（QBDNH）的模型。该模型从现有关系表和查询中获取输入，并使用超图将它们转换为非规范化的 NoSQL 模型。该方法克服了现有基于图的非规范化技术的复杂关系表示和数据访问模式覆盖等限制。所提出的模型减少了手动转换模式所需的总体时间、成本和精力。为了验证 QBDNH 的有效性，在 TPC-H 数据集上进行了实验，并将 QBDNH 的性能与现有的基于图的反规范化模型（例如 TLD、CLDA 和 Kuszera）进行了比较。评估分两部分进行：第一部分分析查询加速因素，第二部分衡量基于查询管道执行的效率提升。结果显示，与现有 TLD、CLDA 和 Kuszera 模型相比，QBDNH 实现了显着的查询性能提升，加速因子分别为 1.29、1.35 和 1.40。此外，与 TLD 和 Kuszera 相比，QBDNH 显着提高了管道利用率。