Theoretical Foundations of Collaborative Information Storage and Transmission

摘要协同信息存储与传输的理论基础本研究探讨了下一代大规模异构系统中分布式信息存储与通信的理论基础。这项研究的一个关键主题是，在完成这一任务时，系统节点之间的大规模协作，而不是仅仅通过抛出基础设施来解决问题来扩展网络。最根本的挑战是采用一种分布式系统方法，这种方法可以依靠数字的力量或系统自身的规模来构建一个可伸缩的、健壮的、能够满足性能保证的整体网络。这项研究的应用范围很广，包括下一代远程学习和基于点对点协作的多节点视频会议，这些会议有可能触及大众，而不是局限于那些有权使用昂贵的集中基础设施的人。这项研究的基础是理论基础的发展，以及实现大规模可扩展、固有鲁棒性和定量可靠的信息存储和传输系统所需的架构和算法。该研究的一个焦点是分布式存储和协作媒体交付范例，包括有和没有延迟约束。由于认识到知识集中化和全局协调导致的瓶颈，本研究采用了一种涉及局部和随机个体行为的分布式方法，从而形成了一个高度动态的系统拓扑。关键的挑战是利用系统本身的规模来保证这些局部交互的可靠性和全局性能保证。本文的研究重点是基于网络编码的分布式存储系统。

AbstractTheoretical Foundations of Collaborative Information Storage and TransmissionThis research addresses the theoretical foundations for enabling distributed information storage and communication in next-generation large-scale heterogeneous systems. A key theme underlying this investigation is that of massive collaboration among system nodes in accomplishing this, in contrast to scaling the network by solely throwing infrastructure at the problem. The fundamental challenge is to embrace a distributed systems approach that can rely on the strength of numbers or the system scale itself to build an overall network that is scalable, robust, and capable of meeting performance guarantees. The application scope of this research is broad, and includes next-generation distance learning, and multi-node video conferencing based on peer-to-peer collaboration that has the potential to reach the masses rather than being confined to those privileged to have access to expensive centralized infrastructure. The basis for this investigation is the development of both the theoretical foundations and the architectures and algorithms needed to realize a massively scalable, inherently robust, and quantifiably reliable information storage and transmission systems. A focal point of the study is that of distributed storage and collaborative media delivery paradigms, both with and without delay constraints. In recognition of the bottleneck caused by centralized knowledge and global coordination, this study embraces a distributed approach involving local and randomized individual behavior that result in a highly dynamic system topology. The key challenge is to exploit the scale of the system itself in guaranteeing reliability and global performance guarantees out of these local interactions. The focal point of this study will be on distributed storage systems based on network coding.