Perspectives in Computing: Locking Performance in Centralized Databases reports some results from a project whose aim was the development of a performance model for concurrency control algorithms. This book proposes a model that is powerful enough to help users understand, compare, and control the performance of the algorithms.
Organized into four chapters, this book begins with an overview of the model in its simplest form. This text then explains locking, which uses a combination of blocking and restarts to achieve correctness. Other chapters test the model and the method of analysis by comparing the model's predictions to simulation results wherein the prediction and simulations indicate that restarts can cause thrashing. This book discusses as well the formula for throughput and for the restart rate. The final chapter deals with the adaptation of the model to timestamping and distributed algorithms.
This book is a valuable resource for mathematicians and research workers.
Inhalt
Introduction
Chapter One: The Model
Chapter Two: The No-Waiting Case
2.1 The Equations
2.2 Validation
2.3 Multiple Transaction Classes
2.4 Queries and Updates
2.5 Variable Length Transactions
2.6 Summary
Chapter Three: The Waiting Case
3.1 Blocking
3.2 Three Approximations
3.3 Waiting Time
3.4 Restart Rate
3.5 Validation
3.6 Comparison to the No-Waiting Case
3.7 Comparison to Static Locking
3.7.1 Incremental Static Locking
3.7.2 Atomic Static Locking
3.8 Nonuniform Access
3.9 Shared Locks
3.10 Effect of Parametric Changes
3.10.1 Resource Contention
3.10.2 Closed-Form Expressions
3.10.3 Changes in N
3.10.4 Changes in k
3.10.5 Changes in D
3.11 Summary
Chapter Four: Conclusion
4.1 Past Work
4.1.1 Experimental Studies
4.1.2 Analytic Models: Dynamic Locking
4.1.3 Analytic Models: Static Locking
4.2 Conclusions about the Model
4.3 Future Work
Appendix: Simulation
Glossary
References
Index