CAREER: Network Coding Theory for Distributed Storage

职业：分布式存储的网络编码理论

• 批准号: 1055099
• 负责人: Georgios-Alex Dimakis
• 金额: $ 45万
• 依托单位: University of Southern California
• 依托单位国家: 美国
• 项目类别: Continuing Grant
• 财政年份: 2011
• 资助国家: 美国
• 起止时间: 2011-02-01 至 2013-07-31
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1055099&HistoricalAwards=false
• 关键词: CAREER; Network; Coding; Theory; Distributed

Modern data centers are becoming the next computing platform for the Internet. These massive farms of computer servers are an indispensable component of the information age, enabling applications like web search, social networks, cloud computing, as well as distributed file and video sharing. The problem of storing and processing massive amounts of information over networks is significant and challenging.Still, most existing distributed storage systems represent information by suboptimal replication schemes combined with ad-hoc networking protocols. This research is focused on understanding, on a fundamental level, how to represent, store, and process information in distributed storage systems.Specifically, this research involves the design of novel distributed storage codes that use network coding theory to address modern storage challenges. As is well known, general multi-source information theory problems can be notoriously intractable. This research focuses on the framework of storage networks and builds on recent results of the PI to investigate: 1) Dynamics of coded storage systems. How to maintain coded information representations over networks. Prior work has demonstrated the key role of network coding for such problems but the fundamental bounds remain unknown for several cases of interest. 2) Interference Alignment for Network Coding. Interference alignment techniques were introduced to mitigate wireless interference at the physical layer but also have surprising applicability in distributed storage and other network coding problems. This research is developing a novel mathematical framework that interprets interference alignment techniques through optimization theory. This project is strongly coupled with educational developments and outreach through an online wiki bibliography and online tutorial presentations.

现代数据中心正在成为互联网的下一个计算平台。这些庞大的计算机服务器群是信息时代不可或缺的组成部分，支持Web搜索、社交网络、云计算以及分布式文件和视频共享等应用。在网络上存储和处理大量信息是一个重要而具有挑战性的问题，然而，大多数现有的分布式存储系统都是通过结合ad-hoc网络协议的次优复制方案来表示信息。本研究的重点是理解，在一个基本的水平上，如何表示，存储和处理信息在分布式存储系统。具体来说，本研究涉及到新的分布式存储代码的设计，使用网络编码理论，以解决现代存储的挑战。众所周知，一般的多源信息理论问题是非常棘手的。本研究主要针对储存网路的架构，并以PI的最新研究成果为基础，探讨：1）编码储存系统的动力学。如何在网络上维护编码信息表示。先前的工作已经证明了网络编码对这些问题的关键作用，但对于几种感兴趣的情况，其基本界限仍然未知。2)用于网络编码的干扰对齐。干扰对齐技术被引入以减轻物理层的无线干扰，但在分布式存储和其他网络编码问题中也具有令人惊讶的适用性。这项研究正在开发一种新的数学框架，通过优化理论解释干扰对准技术。该项目通过在线维基书目和在线教程介绍与教育发展和外联密切相关。