SGER: Dynamic Scheduling and Resource Allocation in Hetergeneous Computer Environments

本研究解决了包含各种架构和计算能力的网络化多用户、多计算机系统的操作问题。特别关注的问题是将软件分配给计算机作业以避免不平衡负载，以及将作业分配给一组相同处理器的动态调度。这些问题的解决对于从网络异构计算环境向“元计算机”发展至关重要。即，在各种可用的计算机之间自动分配工作的高效、方便的计算环境。政府现正申请一笔启动资助，以促进为期一年的初步研究，以研究开发这种元计算机的可行性。伊利诺斯大学的国家超级计算应用中心（NCSA）对上述问题非常感兴趣。NCSA将提供分析和建立异构系统模型所需的数据，并将作为本研究中发展的思想的试验台。通过解决软件分配和分配给一组处理器的作业的动态调度问题，共同研究。他们将分析NCSA关于系统运行的丰富数据，将确定性和随机运筹学研究结合起来开发解决方案，并使用仿真建模来测试他们的解决方案。

This research addresses the operation of networked multi-user, multi-computer systems encompassing a variety of architectures and computing capabilities. The problems of particular concern are assignment of software to computers jobs to avoid unbalanced loads, and dynamic scheduling of jobs assigned to a set of identical processors. Resolution of these problems is vital to progress from networked heterogeneous computing environments to a "meta-computer," i.e., an efficient, convenient computing environment which automatically allocates job among a variety of available computers. A startup grant is being requested to facilitate initial research over one year to study the feasibility of the development of such a meta- computer. The issues described above are of great interest to the National Center for Supercomputing Applications (NCSA) at the University of Illinois. NCSA will provide the data required to analyze and build models of heterogeneous systems and will serve as a test bed for the ideas developed in this research. The co-investigators by addressing the problems of software assignment and dynamic scheduling of jobs assigned to a set of processors. They will analyze NCSA's wealth of data on system operation, bring together determinitic and stochastic operations research to develop solutions, use simulation modeling to test their solutions.