CSR: Medium: Collaborative Research: Architecting Performance Sensitive Applications for the Cloud

云计算为IT组织提供了创建高度可扩展的分布式应用程序的能力，同时提供了极具吸引力的成本节约优势。然而，考虑到云平台的丰富配置选项集、共享的多租户性质以及诸如计划维护等活动所产生的动态特性，架构、配置和调整云应用(延迟敏感型网络应用和海量数据处理应用)以满足其严格的性能要求是一项挑战。该项目正在开发新的方法、算法和系统，使应用架构师能够(1)明智地跨多个云数据中心构建应用，同时考虑应用性能要求、成本节约目标以及由云组件的性能和成本模型指导的云定价方案；(2)通过统计机器学习技术自动学习有效的应用配置和配置-性能预测模型；以及(3)通过在较短时间尺度上的事务重新分配和在较长时间尺度上的应用程序迁移来创建能够适应云环境中的持续动态的应用程序。该研究的影响是多重的：(1)使IT组织能够通过将其操作最佳地转移到云上来显著降低成本；(2)基于操作上部署的应用程序创建基准并收集将提供给研究界的工作负荷跟踪；(3)使开发的算法和系统作为开源软件广泛可用；(4)为泰国全国范围的医疗保健云的设计提供信息；(4)在本科生和研究生课程中引入与声音计算相关的主题；以及(6)培养多名博士、硕士和本科生，明确努力从代表不足的少数群体中招收和培训学生。

Cloud computing offers IT organizations the ability to creategeo-distributed, and highly scalable applications while providingattractive cost-saving advantages. Yet, architecting, configuring, andadapting cloud applications (latency-sensitive web applications andbulk data processing applications) to meet their stringent performancerequirements is a challenge given the rich set of configurationoptions, shared multi-tenant nature of cloud platforms, and dynamicsresulting from activities such as planned maintenance.This project is developing novel methodologies, algorithms, andsystems that can enable application architects to (1) judiciouslyarchitect applications across multiple cloud data-centers whileconsidering application performance requirements, cost savingobjectives, and cloud pricing schemes guided by performance and costmodels of cloud components; (2) automatically learn effectiveapplication configurations and configuration-to-performance predictionmodels through statistical machine learning techniques; and (3) createapplications that can adapt to ongoing dynamics in cloud environmentsthrough transaction reassignment over shorter time-scales, andapplication migration over longer time-scales.The impact of this research is multi-fold: (1) Enable IT organizationsto significantly reduce costs by optimally moving their operations tothe cloud; (2) create benchmarks based on operationally deployedapplications and collecting workload traces which will be madeavailable to the research community; (3) make developed algorithms andsystems widely available as open source software; (4) inform thedesign of a nation-wide health-care cloud in Thailand; (4) introducecloud computing related topics in the undergraduate and graduatecurriculum; and (6) train multiple Ph.D., M.S., and undergraduatestudents, with explicit effort to recruit and train students fromunder-represented minority groups.