CSR: Small: Collaborative Research: Enabling Cost-Effective Cloud HPC

该项目研究建立在公共云框架之上的新服务，以实现经济高效的高性能计算。PI将探索云计算的按需、弹性和可配置功能，以补充传统的超级计算机/集群平台。更具体地说，本研究旨在评估利用云实例的可配置性和分级定价模型构建动态云集群的效率。这一建议的科学价值在于为HPC创新地使用云计算的未开发功能，并战略地采用基于云的小型集群来为大型超级计算机开发/调优应用程序。通过这项研究，PI期望回答以下关键研究问题：(1)特定于工作负载的自动云集群配置；(2)具有成本意识和竞争意识的数据和任务协同调度；(3)自适应、集成的云实例配置和作业调度，以及用于降低云实例租赁成本的工作负载聚合。如果成功，这项研究将产生能够根据不同的HPC需求自适应地聚合、配置和重新配置云资源的工具，目的是为可扩展的并行应用提供低成本的研发环境。

The project examines novel services built on top of public cloudinfrastructure to enable cost-effective high-performance computing.The PIs will explore the on-demand, elastic, and configurable features ofcloud computing to complement the traditional supercomputer/clusterplatforms. More specifically, this research aims to assess the efficacyof building dynamic cloud-based clusters leveraging the configurabilityand tiered pricing model of cloud instances. The scientific value of thisproposal lies in the novel use of untapped features of cloud computingfor HPC and the strategic adoption of small, cloud-based clusters forthe purpose of developing/tuning applications for large supercomputers.Through this research, the PIs expect to answer key research questionsregarding: (1) automatic workload-specific cloud cluster configuration,(2) cost-aware and contention-aware data and task co-scheduling,and (3) adaptive, integrated cloud instance provisioning and jobscheduling, plus workload aggregation for cloud instance rental costreduction. If successful, this research will result in tools thatadaptively aggregate, configure, and re-configure cloud resources fordifferent HPC needs, with the purpose of offering low-cost R&Denvironments for scalable parallel applications.