Automating fine concurrency control in object-oriented databases

Four major problems that complicate read and write accesses to instances in object-oriented databases are discussed. The four problems are: difficulty in providing ad hoc commutativity relations, lacking overhead, lacking escalation, and pseudo-conflicts. It is shown that all of these problems can be solved by providing a simple form of commutativity. This kind of commutativity is syntactically extracted from the source codes of the methods at compile-time. Then, an efficient linear algorithm calculates the transitive access vectors. Finally, transitive access vectors are translated into classical access modes in order not to incur performance penalty at run-time. Related works on access vectors and field locking are reviewed.<
>