Communication Abstractions for Distributed Systems

This thesis presents techniques for reducing the complexity of communication in distributed systems. It presents a set of simple standards and tools that provide an efficient and consistent programming interface that can be used to implement a great variety of communication interactions both within and between various types of paradigms, abstractions, and entities. It also provides a framework based on these standards and tools, with which arbitrary communication systems can be constructed. Various communication paradigms and issues are examined in order to identify and categorize the fundamental aspects of communication. These fundamental aspects are then redefined and organized into a set of communication abstractions which is simple and general, rigorous and flexible, low-level and extensible. The generic communication model based on this set provides a framework for arbitrary communication systems. Further, the model utilizes the efficiencies of a single memory domain while providing a universal interface between various types of paradigms, abstractions, and entities across a wide spectrum of environments. The framework provided in the generic communication model is used to develop a specific model called the Buffer and Queue Model which provides a single efficient and consistent communications programming interface. The Buffer-Queue protocol extends this consistent programming interface transparently to a distributed environment. Finally, a few examples are presented using Buffers and Queues that illustrate how various complex communication facilities may easily be developed and how the use of a common fundamental base makes intermixing of locally defined user interfaces and conversion between higher-level protocols a relatively straightforward task.