No description available.

No description available.

No description available.

No description available.

No description available.

This dissertation addresses the problem of writing portable programs for parallel computers, including shared memory, distributed, and non-uniform memory access architectures. The basis of our approach is to separate the expression of the algorithm from the machine-dependent details that are necessary to achieve good performance. The method begins with a statement of the algorithm in a classic, explicitly parallel, manner. This basic program is then annotated to specify architecture-dependent details such as scheduling and mapping. These ideas have been cast in terms of a programming language, Par, which provides flexible facilities for a range of programming styles, from shared memory to message passing. Par is used to specify both the algorithm and the implementation of the annotations.

This dissertation addresses three fundamental problems with the network software as it is currently designed and implemented: the poor performance of high modular (or layered) protocols, network software's inability to keep up with the rapid changes in the networking technology and applications demands, and the inordinate amount of time it takes to produce new network protocols. These problems are solved through the use of a new platform for the implementation and execution of network protocols, a new methodology for the design of network protocols and a new network architecture a new network architecture. Avoca is the result of taking a coordinated approach to network software design, implementation, and standardization and consists of three parts: the Avoca platform, the Avoca methodology and the Avoca Network Architecture. The Avoca platform is a small operating systems kernel designed from scratch to implement network protocols efficiently. The Avoca methodology (or Meta-Protocol) is set of rules governing the design and implementation of network protocols implemented on the Avoca platform. The Avoca Network Architecture is a novel architecture explicitly designed to support the rapidly changing networking environment. Using Avoca highly layered network software can be implemented efficiently. Avoca proves that modularity is not inherently slow. Avoca supports the encapsulation, underspecification, composition and reuse of protocols will demonstrated. Avoca shows that network software is amenable to the use of software engineering techniques to improve the protocol implementation process. Finally, Avoca demonstrates that a network architecture flexible enough to support a rapidly changing networking environment is possible.

Recent studies point to major problems in today's software systems. Problems in cost, reliability, maintainability, and poor responsiveness to user requirements have their origin in the early phases of the system development effort. Although increasing awareness of poor design practice has stimulated several research efforts toward making the "detailed design" process more "systematic," there is a significant need for a computer-aided methodology to help designers cope with the complex design process. A framework is established for organizing activities in support of one important aspect of "detailed design," the organization of processes into appropriate process groups and program modules. A computer-aided methodology is presented for analysis of variety of inter-process relationships in the determination of effective modularizations. The proposed methodology extends current software engineering practice through partial automation of an important software engineering problem, the effective structuring of processes according to multiple design criteria. Multiple design criteria are used to determine inter-process relationships. The system accomodates a number of design criteria including volume of data transport, distribution of data references, information and control distribution. The methodology begins with the assignment of a graph structure to subsystem components and their interdependencies. The resulting graph is partitioned in determination of subgraphs (modules) with strong intra-dependencies and weak inter-dependencies. The set of subgraphs define modules which satisfy principles of high module strength and low module coupling. The decomposition method used also produces a hierarchical structure of modules with little resource sharing. The resulting design limits "reference distribution" and "information distribution" between modules, which results in reduction of complexity of the total structure. Analytical tools in support of these activities are presented to illustrate support of the methodology by a pilot study.

No description available.

No description available.