Static allocation in Distributed Objectbase Systems: A graphical approach

Abstract

In recent years, object oriented (OO) models have emerged as preferred data models for a wide range of application domains. This is primarily due to the capability of the OO-models to manage and represent information of any arbitrary complexity. Distributed Objectbase Systems (DOBS) combines the benefit of distributed processing and the power of manipulation and abstraction of complex information to provide a powerful environment for distributed computing that is not available in conventional distributed database systems. Application performance and processing cost in a DOBS are greatly influenced by communication overhead involved in accessing nonlocal data and making remote method invocations. Efficient distribution design is mandatory to ensure optimal performance of the distributed system at minimum cost. Distribution design in DOBS involves fragmentation of the objectbase and allocation of the resulting class fragments between the nodes of the network.We adopt a top-down approach for the distribution design for a DOBS and assume that the global conceptual schema is partitioned into a set of class fragments. We focus our attention to optimal allocation of the fragments between the sites of the network subject to a set of constraints. Our entity of allocation is a class fragment. The allocation scheme defines the local conceptual schema at every site.The problem of allocation has been addressed for distributed file systems (DFS) and distributed (relational) database systems (DDBS). The additional complexity introduced by the object model has only recently been investigated with no result appearing that address the specific problem of allocation. This paper addresses the problem of object allocation in DOBS. A general allocation taxonomy is defined based on data model, degree of redundancy and design objective, and different allocation models are classified based on this taxonomy. An allocation model for DOBS is formulated and an algorithm is presented for static nonredundant allocation in DOBS.KeywordsAllocation SchemeAllocation ProblemCommunication CostAllocation ModelInitial AllocationThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.