基于全数据的云存储系统实时性能建模理论及方法研究

项目名称： 基于全数据的云存储系统实时性能建模理论及方法研究

项目编号： No.61472323

项目类型： 面上项目

立项/批准年度： 2015

项目学科： 自动化技术、计算机技术

项目作者： 张晓

作者单位： 西北工业大学

项目金额： 80万元

中文摘要： 云存储是IaaS的重要组成部分，是大量云计算应用的基础。这些应用需要低成本、高性能且稳定的数据存储服务。由于云存储系统是一个多用户、分布式的大规模动态系统，现有的评测及建模方法不适用于云存储系统的性能分析。传统建模方式仅能反映系统的统计规律，无法获得真实系统的特性。随着大数据相关研究的发展，云存储系统中数据的采集、存储和处理都有了突破。对复杂系统进行全面数据采集并建模已成为一种新的建模途径。本项目拟提出一种基于全数据的性能预测模型，该模型对云存储系统进行持续数据采集，结合云存储服务性能指标，分析并预测云存储系统级的性能和瓶颈。研究内容包括:1.利用集成的全数据建立自适应的反馈性能分析及预测模型。2.数据缺失和采集延迟情况下，劣质数据的检测与修复方法，以及延时数据的利用方法。

中文关键词： 系统建模；性能预测；性能优化

英文摘要： Cloud storage systems provide sharable public data storage service to public. This leads new application model and it becomes a new trend in internet technology development. Because there are so many of internet applications and private data stored in the public data storage platform, the performance of data access almost decides whether related services work well. Performance measurement and analysis is the basis of system maintenance and performance optimization. The existing methods take several characteristics into model based on some hypothesis. This modeling approach requires an understanding of the mechanism of the storage system. This model can only get statistical results and it cannot reflect the true characteristic of a real system. With the development of big data, people can collect, storage and process big data. It's a new approach to modeling a complex system with massive data collected. The goal of our project is to make full use of all the data collected in cloud storage systems .This project aims to build a dynamic self-adaptive performance model to estimate runtime performance data using machine learning. The research including: 1. self-adaptive performance modeling using all data gather from trace/log and sensors. 2.how to use uncertain data in massive data and how to handle data transmission delay.

英文关键词： System modelling;Performance Predict;Performance optimization