How can databases be structured to efficiently serve the diverse needs of transactions? This paper investigates vertical partitioning, a technique to divide a database relation into fragments aligned with transaction requirements. By grouping attributes frequently accessed together, vertical partitioning aims to optimize database performance in various contexts, enhancing both speed and efficiency. The paper explores its application in single-type devices, multi-level memory systems, and distributed databases, emphasizing local transaction processing. A two-phase approach is presented: initially driven by empirical functions, followed by cost optimization within the application environment. Implemented algorithms and examples highlight the practical benefits, showcasing its potential to improve data accessibility and reduce processing overhead in database systems.
Published in ACM Transactions on Database Systems, this paper contributes to the journal's core focus on database design and optimization. By addressing the problem of vertical partitioning, the research offers insights into improving database performance and resource utilization, topics of significant interest to the journal's readership.