Planning support for cooperating transactions in EPOS

Abstract

The paper observes that current transaction models in databases are too strict for cooperative and long transactions in software engineering. On the other hand, unrestricted concurrency and data sharing is not desirable either. The paper describes a way to reduce the number of conflicts that may arise when several users cooperate to solve a task using a common database. Manual interaction among the users are made easier by helping a project manager to better partition, schedule and connect the users' local activities as part of a common task. Given a proposed project division, we can assess the connections between subprojects through impact analysis. Based on this analysis, the project manager can choose to adjust the initial partitioning to reduce the dependencies between subprojects. The impact analysis can further help the project manager to schedule the identified subprojects. Lastly, such analysis can be used to suggest suitable cooperation protocols between subprojects, being scheduled in parallel. Such protocols will guide mutual propagation and sharing of temporary results, and associated negotiation. The paper also reports the preliminary implementation status of this work in the EPOS kernel software engineering environment.