On the performance issues of object-based buffering

Due to the nature of supported applications, object-oriented DBMSs need efficient mechanisms for the retrieval of complex objects and the navigation along semantic links among objects. Object clustering is effective when the primary access pattern follows the primary relationship used for clustering, but not so effective when multiple relationships are involved. Object buffering provides a possible solution to compensate the limitation of clustering schemes. The paper investigates the advantages of using an additional layer of object-based buffer pools on top of a traditional page-based buffer pool, and proposes a profile-based buffering scheme to optimize the efficiency of object buffer pools. The performance evaluation based on the semantic implementation of profiles was conducted.<<ETX>>