Engineering Research Equipment: Real-Time Scheduling via Large-Scale Problem Reformation

所要求的设备是带有4个75 Mhz处理器的Silicon Graphics Power Challenge L Deskside服务器，将支持以下领域的研究：(1)通过大规模问题重构实现实时调度；(2)鲁棒制造系统仿真设计；(3)非并发排队网络的实时作业流控制；(4)安全节能的天然气网络。实时调度问题通常被表述为整数规划问题，通常不适合解决小问题。用新的变量集和约束重新表述这些问题显示了希望，导致大规模的问题需要解决。在目标寻求专家系统的基础上，利用一种新的、独特的中小型制造系统仿真环境，寻求一种满足多种性能标准的鲁棒系统设计，并且对不可控的环境或噪声参数不敏感。在非并发排队网络环境下，基于数学基础，在提高作业吞吐量的同时，将作业停留时间保持在可接受的范围内，需要适合实现实时控制的数值过程。离散优化和大规模整数规划的最新研究将用于评估网络安全和吞吐量之间的权衡，以及设计安全高效的天然气网络。购置的设备将在现有的以太网系统内联网，以便最大限度地利用四个重要的研究项目。研究问题的成功与否取决于高性能计算设备的可用性。

9622138 Boyd Acquisition of a The equipment requested is a Silicon Graphics Power Challenge L Deskside Server with four 75 Mhz processors will support research in the areas of: (1) real-time scheduling via large scale problem reformulation; (2) design of robust manufacturing systems using simulation; (3) real-time job flow control in non-concurrent queueing networks; and (4) safe and energy-efficient natural gas networks. Real-time scheduling problems are naturally formulated as integer programming problems and are commonly inadequate for solving even small problems. Reformulation of these problems with new sets of variables and constraints shows promise, resulting in large-scale problems to solve. A robust system design based on a goal seeking expert system and utilizing a new and unique simulation environment for manufacturing systems in small and medium size operations is sought that satisfies multiple performance criteria and is insensitive to uncontrollable environmental or noise parameters. Improved job throughput, while holding job sojourn times to within an acceptable range, based on mathematical foundations in the context of non-concurrent queueing networks require numerical procedures suitable for implementing real-time control. Recent research in discrete optimization and large-scale integer programming will be brought to bear on evaluating tradeoffs between network safety and throughput and in designing safe and efficient natural gas networks. The acquired equipment will be networked within an existing ethernet system to permit maximum utilization fro four important research projects. The success of the research problems is contigent upon the availability of high-performance computing equipment.