Object persistence in heterogeneous database environments

Abstract

1 .O Introduction Corporate IT trends indicate increased focus on shared enterprise level data access. Such access is viewed as crucial weaponry of the IT arsenal. Successful implementation of such global data access is impeded by the numerous persistence mechanisms used within each corporation. Due to investments in traditional software solutions, there is a mounting need to integrate legacy systems with " faster-to-market " object based technologies. This workshop concentrated on the significant issues encountered when performing integration of run-time objects, persistent objects and data residing in relational and other persistence mechanisms. Fourteen corporations and one university were represented in this workshop. The workshop proved to be a real life testament for groups experiencing the practical challenges associated with object and data integration. The workshop was broken into three intervals. This first interval offered an opportunity for the authors of accepted papers to briefly discuss their experiences and solutions. Additionally, authors identified their 6 hottest topics. The second interval consisted of work group discussions centered around four topic groupings. The third interval consisted of presentations by representatives of each work group. Motivations, challenges and insights to the topics in each work group were reviewed in these presentations. Several of the same topics were discussed in more than one work group. Hence, the work group findings are categorized into: The motivation for this work group was to avoid supporting a " least common denominator " approach for dissimilar transaction models. Work group participants pointed out that one approach to solving the dissimilar transaction model problem was to support a generalized transaction model. This would circumvent the benefits of tailoring specific transaction models toward the requirements of any particular application. Difficulties lie in the synchronization of databases supporting long transactions with databases supporting short transactions. Long transactions span the life of clients connected to the databases. Additionally, objects checked back in to the database become versioned. Information about the long transactions would have to be embedded within the application data for data residing on other persistence mechanisms. The work group briefly discussed the issue of nested transactions. Due to the complex interrelated nature of objects, there is a requirement to nest transactions relative to object boundaries. Nested transactions are not typically supported within current relational and other non object based persistence mechanisms. Other architectures supported more sophisticated levels of access control. For example, request management was supported for data access and service …