Challenges in CISE: Metacomputing Environments for Optimization

9726385 Nocedal，豪尔赫西北大学 CISE中的挑战：优化的元计算环境 通过结合分布式硬件、网络硬件/软件 基础设施、优化工具和信息材料， 研究将产生元计算环境的优化功能强大 足以解决规模和复杂性前所未有的问题。 该项目集成了基础算法和面向应用的 研究优化与研究软件基础设施， 分布式系统管理。新的算法使用强大的，但 元计算机平台的异构、高延迟环境将 旨在解决目标应用领域的关键问题。的 优化工具将包括链接客户端工作站的接口 连接到元计算机服务器上环境将能够 与其他地方的数据收集和建模资源协同执行， 网络。 算法研究的重点领域包括组合优化， 整数规划、全局优化和非线性规划， 目标应用领域包括运输、电信、分子 构造和天气预报。软件基础设施研究将 重点关注资源分配、调度、持久性和安全性 最大限度地发挥环境的作用所需的机制， 优化工具及其用户的观点。该项目将使 使用Globus正在建造的GUSTO元计算试验平台 项目

9726385 Nocedal, Jorge Northwestern University Challenges in CISE: Metacomputing Environments for Optimization By combining distributed hardware, network hardware/software infrastructure, optimization tools, and informational material, this research will produce metacomputing environments for optimization powerful enough to tackle problems of unprecedented scale and complexity. The project integrates fundamental algorithmic and applications-oriented research in optimization with research in software infrastructure for distributed systems management. New algorithms that use the powerful but heterogeneous, high-latency environment of the metacomputer platform will be developed to solve key problems in targeted application areas. The optimization tools will include interfaces that link the client workstation to the metacomputer server in novel ways. The environment will be able to execute in tandem with data-gathering and modeling resources elsewhere in the network. Focus areas for the algorithmic research include combinatorial optimization, integer programming, global optimization, and nonlinear programming, with target application areas of transportation, telecommunications, molecular conformation, and weather forecasting. Software infrastructure research will focus on the resource allocation, scheduling, persistency, and security mechanisms needed to maximize the usefulness of the environment, from the viewpoint of the optimization tools and their users. The project will make use of the GUSTO metacomputing testbed being constructed by the Globus project.