No abstract

No abstract

Concurrency control for a real-time database must maintain both the traditional logical consistency constraints of data and transactions, and the additional temporal consistency constraints of data and transactions. Furthermore, the concurrency control should have the ability to express the trade-o that results from the inherent con ict between temporal and logical consistency constraints. The concurrency control should also be able to maintain and bound any imprecision that results from trading o logical consistency for temporal consistency.This dissertation presents a model and a concurrency control technique for real-time object-oriented databases. The model, called RTSORAC, is based on an object-oriented data model and it allows for the explicit expression of logical and temporal data consistency constraints. The concurrency control technique supports temporal and logical consistency, as well as bounded imprecision that results from their trade-o . It uses a semantic locking mechanism within each object and user-de ned compatibility o ver the methods of the object. The semantic-based compatibility function can specify when to sacri ce precise logical consistency to meet temporal consistency requirements. The concurrency control technique can also specify accumulation and bounding of any resulting logical imprecision.The dissertation presents a set of restrictions on the compatibility function and an object-oriented form of the epsilon-serializability correctness criterion OESR. It then presents a proof of global correctness that shows that the semantic locking technique, under the compatibility function restrictions, can guarantee OESR.To demonstrate how the semantic locking technique performs under varying conditions, the dissertation presents a set of tests that were conducted to compare the technique with other techniques. The techniques indicate that, in general, the semantic locking technique preserves transaction temporal consistency better than the other techniques tested, and it keeps data temporal inconsistency low.i ACKNOWLEDGEMENT I h a ve always thought of the work in this dissertation as a collaborative e ort between me and so many people who have helped me in many w ays, not just academically.First I must thank my advisor and friend Victor Fay-Wolfe, without whom I would not have known what a real-time database was. He has been an advisor in more than just my work and has provided me with all of the support a graduate student could ask for -and then some.I also thank Joan Peckham, who has been a role model for me in work and in life and in combining the two! My fellow graduate student Janet Prichard has been a supportive friend as well as a helpful colleague. I thank the entire RTSORAC research group for all of the valuable input I have gotten throughout the process of my research.I thank Manbir Sodhi and the rest of my committee for their support and help throughout the entire PhD process. Many thanks to John Kelvin Black, who provided me with a a solid testbed system upon which to des...

This paper describes the requirements for real-time

Abstract