BufferBank storage: an economic, scalable and universally usable in-network storage model for streaming data applications

Abstract

Large-scale streaming media distribution services impart unprecedented pressures and challenges to existing Internet architecture. Researchers have proposed CDN, P2P Cache, ICN and other solutions to alleviate the pressure of the core network, as well as improve the user experience (QoE). All these solutions achieve their goals by deploying storage resources close to the end users to cache the hot data. Based on the advantages of P2P service, which takes into account end-users resources, we proposed the BufferBank Storage (BBS). It is a new streaming media distribution-oriented storage model by aggregating end-users resources. This provides a novel approach for implementation of economic, scalable, dynamically deploy streaming media distribution applications. However, the dynamic character of the end-user behavior brings challenges to BBS designation. In our previous work, we have analyzed the basic principle of BBS and its feasibility. There is lack of substantial research on resources management and reliability assessment, which are the core issue of BBS implementation. After carefully analyzing the reliability and security issue in BBS deployment, this paper has proposed the implementation model of BBS and studied the performance of different buffer allocation mechanism through simulation. Our work mainly provides a new way of thinking for the dynamic, universal scalable storage system in the Internet, that suffers “weak reliability”.

中文概要

BufferBank是目前唯一一种聚集互联网环境中端用户分布地、不可靠的存储资源,通过有效的存储管理技术,向不同流媒体应用提供网内存储服务的机制。本文首次对BufferBank服务中BBD的动态管理、可靠性评估和存储管理的核心机制进行研究,设计了互联网环境中基于非受控存储资源的“概率可靠性”存储系统,形式化定义了BufferBank系统中的分布式缓冲区的管理问题,给出了BBM分布式缓冲区管理的实现模型,设计了BufferBank的模拟器,对不同的缓冲区管理策略的性能进行了分析评估,为流媒体分发实现可动态部署的,可扩展的本地缓存提供了新的思路。