A descriptive model for active message inter-process communication.

Electronic data processing > Distributed processing.

Telecommunication > Message processing.

The concept of active message inter-process communication (AMIPC) has been introduced, described and studied. An active message contains or consists of an executable procedure or script. It supports communication between a sender and a recipient by executing in the vicinity of the recipient, and representing the sender to the recipient. Like a conventional computer program, the active message can, in theory, incorporate unlimited amounts of logic and/or data. For example it may engage in a dialogue with the designated recipient, create and send other messages, route itself around a network, etc. The research was motivated by the recognition of limitations in conventional message inter-process communication (MIPC), which can lead to delay induced inconsistent behavior in distributed software systems. The potential for active messages to dynamically make resource optimizing decisions make them an attractive alternative to conventional messages > as the fundamental construct in MIPC. The primary objective of the research was to formulate a model for describing the primitive objects and operations in AMIPC, and to demonstrate its utility by using it to design some AMIPC facilities. A secondary objective was to obtain some preliminary evidence of AMIPC relative performance compared to conventional MIPC. A descriptive model for AMIPC was successfully constructed. Experimental AMIPC facilities, based on this model and motivated by a military C₂ case study, were then designed. Also designed was an experimental message specification language (MSL). Next, the experimental AMIPC facilities design was implemented at a prototype level. Finally, the prototype implementation was operated, in both AMIPC and MIPC communication modes, to collect some preliminary performance data for both modes. An analysis of the preliminary results is provided, along with suggestions on how they should be used. The significance of this work is that it introduces the concept of active messages to distributed software systems, in general, and inter-process communication in specific. It also demonstrates that systems which support AMIPC, though not yet well understood, could be designed and implemented. The research does not show, however, that AMIPC should be immediately considered as a serious alternative to conventional MIPC.

Academic theses