CSR-AES: Adaptive Optimization for Dynamically Discovered Hardware and Software Resources

CSR-AES：动态发现的硬件和软件资源的自适应优化

• 批准号: 0720471
• 负责人: Rudolf Eigenmann
• 金额: $ 30万
• 依托单位: Purdue University
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2007
• 资助国家: 美国
• 起止时间: 2007-08-01 至 2012-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=0720471&HistoricalAwards=false
• 关键词: CSR; AES; Adaptive; Optimization; Dynamically

This project pursues the vision of a community that shares resources in an openmanner. Similar to today's web publication and discovery of documents, computerapplications, machine cycles, and databases can be announced openly and used byothers worldwide.A big issue in realizing this vision is to create facilities that help managethe diversity of available resources and adapt them to the users' needs. Thus,published software modules may be adapted to compose new application programs;discovered programs may be tuned to the user's platform; distributedapplications may run at high performance, as they adapt to changing bandwidthand latency characteristics of the target environment. This project developssuch facilities. Building on results of prior NSF support, an adaptiveengine will dynamically improve the performance of distributed computerapplications; it will adaptively compose software modules from components; andit will support computer applications that roam across discovered machineplatforms. These components are being integrated in the iShare Internet sharingsystem, which provides the environment for resource publication, discovery, andremote use.Components of this system are being created in related NSF projects, which synergistically complement the present project.This work addresses a fundamental problem in computer system and softwaredesign: optimal design decisions are often impossible because insufficientinformation is available at design-time. The adaptive engine moves designdecisions to the time application software executes, when this information becomesavailable.The project is expected to yield significant performance improvements of theinvolved compilers, composers, and application software, substantially facilitating the way the involved resources can be shared by the global community.

该项目追求以开放的方式共享资源的社区愿景。与今天的网络发布和发现文档类似，计算机应用程序、机器周期和数据库可以公开发布并供全世界的其他人使用。实现这一愿景的一个大问题是创建有助于管理可用资源多样性并使其适应用户需求的设施。因此，已发布的软件模块可以改编成新的应用程序；发现的程序可以调整到用户的平台；分布式应用程序可以以高性能运行，因为它们可以适应目标环境不断变化的带宽和延迟特征。本项目开发这些设施。基于先前NSF支持的结果，自适应引擎将动态地提高分布式计算机应用的性能；将组件自适应地组合成软件模块；它还将支持在已发现的机器平台上漫游的计算机应用程序。这些组件被集成到iShare互联网共享系统中，该系统为资源发布、发现和远程使用提供了环境。该系统的组成部分正在相关的国家科学基金会项目中创建，这些项目将协同补充本项目。这项工作解决了计算机系统和软件设计中的一个基本问题：由于在设计时可用的信息不足，通常不可能做出最佳设计决策。自适应引擎将设计决策移动到应用程序软件执行的时间，当这些信息变得可用时。该项目有望显著提高所涉及的编译器、作曲家和应用软件的性能，极大地促进了所涉及的资源可以被全球社区共享的方式。