IMPLEMENTATION TECHNIQUES FOR MAIN MEMORY DATABASE SYSTEMS

IMPLEMENTATION TECHNIQUES FOR MAIN MEMORY DATABASE SYSTEMS

© 1984 ACM 0-89791-128-8/84/0006/0001 $00.75 | David J. DeWitt1, Randy H. Katz2, Frank Olken3, Leonard D. Shapiro4, Michael R. Stonebraker2, David Wood2
This paper explores the changes necessary to enable relational database systems to utilize large amounts of main memory. It evaluates various access methods, including AVL and B+ trees, and hash-based query processing strategies. The study finds that B+ trees are the preferred storage mechanism unless more than 80-90% of the database fits in main memory. Hash-based query processing is advantageous for large memory situations. The paper also discusses algorithms for relational database operations, such as join, projection, and aggregate functions, and evaluates their performance under different conditions. Additionally, it examines recovery mechanisms for memory-resident databases, proposing techniques like pre-committed transactions and checkpointing to improve transaction throughput and reduce recovery times. The authors conclude that while B+ trees remain the preferred access method, hash-based algorithms are optimal for large memory environments. Future research directions include buffer management, virtual memory effects, and concurrency control.This paper explores the changes necessary to enable relational database systems to utilize large amounts of main memory. It evaluates various access methods, including AVL and B+ trees, and hash-based query processing strategies. The study finds that B+ trees are the preferred storage mechanism unless more than 80-90% of the database fits in main memory. Hash-based query processing is advantageous for large memory situations. The paper also discusses algorithms for relational database operations, such as join, projection, and aggregate functions, and evaluates their performance under different conditions. Additionally, it examines recovery mechanisms for memory-resident databases, proposing techniques like pre-committed transactions and checkpointing to improve transaction throughput and reduce recovery times. The authors conclude that while B+ trees remain the preferred access method, hash-based algorithms are optimal for large memory environments. Future research directions include buffer management, virtual memory effects, and concurrency control.
Reach us at info@study.space