CSR: Small: PROM in Clouds: Exploiting Scheduling for PeRformance OptiMization in Clouds

云计算正在成为一种有吸引力的计算模式，其中硬件和软件资源都可以从云提供商那里租用，以实现更好的规模经济，弹性，可用性和灵活性，以提供大量的服务。设计这样的云平台的一个主要挑战是提供高的和可预测的性能沿着服务质量（QoS）保证。性能的不可预测性归因于无数的因素，包括系统规模、工作负载动态性、虚拟化开销、资源共享和争用。本提案的目标是通过三个相互交织的任务来研究云计算中的性能增强和性能量化技术。首先，一个由高级作业调度器和精细作业调度器组成的综合调度框架，第二，开发云平台作业完成时间（JCT）的性能预测模型;第三，提供服务质量保证，一个控制理论模型将建议的调度和分析模型将被开发。通过在具有MapReduce和其他代表性云工作负载的真实的平台上进行测量，验证了本文提出的想法。如果成功地回答了这个问题，它将对支持云的应用程序域产生巨大的影响。这项工作的交叉性可以培育新的研究途径。除了建筑、分布式系统和性能建模等领域的本科生和研究生培训外，还将特别关注女性和少数族裔学生参与该项目。在这项研究中开发的建模工具和技术将公开提供。

Cloud computing is emerging as an attractive computing paradigm, whereboth hardware and software resources can be leased from a cloudprovider to achieve better economy of scale, elasticity, availability,and flexibility for a plethora of services. A major challenge indesigning such cloud platforms is in providing high and predictableperformance along with Quality of Service (QoS) guarantees. Theperformance unpredictability is attributed to a myriad of factorsincluding system scale, workload dynamism, virtualization overheads,resource sharing and contention.The objective of this proposal is to investigate performanceenhancement and performance quantification techniques in a cloudenvironment through three intertwined tasks.First, a comprehensive scheduling framework consisting of a high-leveljob scheduler and a fine-grained resource manager will be developedfor efficient management of cloud resources for performance enhancement.Second, performance prediction models will be developed for estimatingjob completion time (JCT) in a cloud platform.Third, for providing QoS guarantees, a control-theoretic modelincorporating the proposed scheduler and analytical models will bedeveloped. Measurements on real platforms with MapReduce and other representative cloud workloads will be used to validate theproposed ideas.This research addresses one of the main concerns of cloud computing -What are the performance implications if my application is moved to acloud? If successful in answering this question, it would have a tremendousimpact on the cloud-enabled application domains. The cross-cuttingnature of this work can foster new research avenues.In addition to undergraduate and graduate student training in areassuch as architecture, distributed systems, and performance modeling,special attention will be given to involving women and minoritystudents in this project. The modeling tools and techniques developedin this research will be made publicly available.