The authors expand on their previous work on concurrency control for OODBMSs and cite all the significant related work. The idea is to propose, evaluate, and model adaptive concurrency control approaches. The authors use and study adaptive fine-grained locking options for client/server OODBMSs. They study the performance of diverse locking granularities at the page and object levels; they also deal with consistency (using a callback scheme).

The paper states that, for reasons of simplicity and communications efficiency, a number of OODBMSs use page server architectures (where the data page is the minimum unit of transfer and client caching). Despite their efficiency, page servers are often criticized as being too restrictive when it comes to concurrency, because they use pages as the minimum locking unit as well. The paper shows how to support object-level locking in a page server context. Several approaches are described, including an adaptive granularity approach that uses page-level locking for most pages but switches to object-level locking when finer-grained sharing is demanded. Each of these approaches is based on extending the idea of callback locking.

OODBMSs phrase requests in terms of objects. They store objects in units of pages and manage them in terms of data pages. The data shipping approach can manage object-to-page mapping at different granularity levels. The two client/server approaches analyzed in the paper are page server and object server. The final context points are client cache consistency and cache consistency maintenance protocols to ensure a consistent view of the database.

This paper extends some of the authors’ previous work, where they claim to have demonstrated that page servers can support fine-grained data sharing. An adaptive locking algorithm and a lock de-escalation technique that works in the context of hierarchy lock granularities are presented. The latter uses coarse locks that de-escalate into fine-grained locks (for example, pages or records).

The granularity issues are client/server data transfer, concurrency control, and replica management. Client-server data transfer deals with client requests to servers. Concurrency control, or locking, involves multilevel (multiuser) requests. Coarse granular locks have less overhead, but can lead to false sharing and thus to conflicts. Replica management (callbacks) involves intertransaction issues, including caching copies of the data. It requires both replica management and locking. The paper proposes that dynamic adaptive approaches can result in performance gains.

The key contribution of this paper is that it defines five approaches and analyzes their performance based on the three granularity criteria. The authors provide a comprehensive description of each and thoroughly describe the modeling and results. The approaches analyzed and modeled in the paper are object server;page server; page server with object locking with object callbacks; page server with object locking with adaptive callbacks; and page server with adaptive locking with adaptive callbacks.

This excellent reference paper is required reading for those who work with, research, and develop OODBMSs. It provides a formal framework by which to understand the intuitive results. The modeling results also provide new insights into locking strategies for this genre of OODBMS applications.