Object group interactions in distributed systems

Abstract

A distributed system is an information system where discrete components of the overall processing activities may be located in more than one system, at more than one geographical location, or where there is any reason which necessitates explicit communications among the components. The purpose of this thesis is to establish the foundation for object group interactions in distributed systems. Object groups provide a basis for replication and parallelism which increase the opportunities to improve performance, availability or fault tolerance. However, multiple members in the object group and coordination between them add a major degree of complexity; a large number of objects are involved and the membership is changed continuously. Comprehensive support for the object groups is a way of selectively hiding this complexity. Multiple objects are configured into an object group and behave towards other objects as if the multiple objects were collectively one object. An object can also join or leave an existing group at any time. Infrastructure and primitives to support dynamic management of object groups are introduced. Intra-group interactions is distinguished from inter-group interactions. The inter-group interactions hides the existence of a group and is shown as an interaction between singleton objects as a whole, while intra-group interactions deal with the coordination between group members. The group management and organization of inter-group interactions via intra-group interactions are achieved via a group agent acting as a proxy on behalf of all members. Group Execution Protocol enables a client group to invoke an operation in a server group with hiding the complexity of the group interactions. Group Execution Protocol provides a remote procedure call service for groupto-group interactions for an object-oriented distributed processing systems.