Addressing techniques used in database object managers O 2 and ORION

Abstract

Addressing mechanisms used by the new generation of Data Base Management Systems (DBMS) differ significantly from traditional ones. Such changes are the direct result of new applications requirements such as office information systems (OIS) and computer aided design (CAD). In this context, object format requires different representations on disk and in main memory, and this is often valid for interobject references. It is evident that these mechanisms are closely linked to the mode of object-identity implementation, as well as clustering strategies. All of these functions are controlled by the object manager.This article describes these mechanisms through the implementation of two object managers for object-oriented DBMS: O 2 and ORION. We show how the performance of these systems depends on their memory management and addressing scheme. The two managers to be discussed merges techniques proposed by both data base field and object-oriented programming field. Their own mechanism differs, according to the way it handles distribution. ORION-1SX and O 2 have a Client/Server architecture, but each one uses a different approach for distributing of functionalities. ORION-1SX implements an object-server, whereas O 2 uses a page-server approach. An analysis of the two systems shows that they both use a two-level addressing mechanism. buffer management for objects in memory is diffent and more complex in ORION. On the other hand, the clustering strategies in O 2 have the advantage of being more dynamic and can be specified outside the schema.