云数据中心基于应用共存特性的混合调度研究

项目名称： 云数据中心基于应用共存特性的混合调度研究

项目编号： No.61202060

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 计算机科学学科

项目作者： 宋莹

作者单位： 中国科学院计算技术研究所

项目金额： 25万元

中文摘要： 随着云数据中心规模的不断增加和云应用的层出不穷，为解决资源利用最大化与应用共存冲突导致服务质量下降之间的矛盾，当前云数据中心普遍采用基于虚拟化技术的共享服务器和非虚拟化的独占服务器并存的使用模式。在该大规模混合场景下，如何将大量云应用自动、高效地映射到异构服务器成为当前云数据中心面临的一大挑战。本课题拟研究以下三个问题：1）针对如何预测具有不同资源需求特性的应用在不同负载下资源需求问题，考虑构建云应用及应用组合对异构服务器规模需求的预测模型；2）针对如何定量刻画云应用在共享资源时相互影响程度的问题，考虑进行细粒度应用资源占用模式及共存冲突分析；3）针对如何将大量云应用高效映射到大规模异构服务器的问题，考虑设计基于应用共存非线性叠加的应用自动部署方法。通过该项目的研究，将在ACM或IEEE期刊或会议上发表学术论文6篇以上，同时将为云数据中心提供高效的应用分析及部署方法，促进云数据中心的发展。

中文关键词： 调度；云数据中心；应用共存；；

英文摘要： With the rapid increasing of the scale of cloud data centers as well as the amount of cloud applications, the coexistence of the dedicated physical servers and the virtualization-based shared servers is the typical usage mode in current cloud data centers to resolve the contradiction between the maximum resource utilizaiton and the QoS (quality of service) decrease resulted from the resource sharing by applications. In such a large-scale hybrid scenario, how to map the large amount of cloud applications to the heterogeneous servers automatically and efficiently is the burning challenge faced by today's cloud data centers. In this project, we will research on the following three problems. Firstly, as to the problem of how to determine the server requirements by applications with diverse resource requirement characteristic under different workload, we will model the prediction of the server requirements by a cloud application as well as by the combination of cloud applications. Secondly, as to the problem of how to quantitatively determine the effect degree among cloud applications when they share the same physical server, we will analyze the fine-grained resource usage and coexistence characteristic of cloud applications. Lastly, as to the problem of how to deploy a large amount of applications with diverse resou

英文关键词： Scheduling；Cloud data center；application coexistence；；