Study on large-scale scalable P2P grid infrastructure for large-capacity distributed computing

P2P网格的概念是利用办公室和实验室中的PC作为网格计算的计算资源，使用P2P技术，而网格计算主要是为了共享各个研究机构的计算资源。我们的目标是研究和建立面向大容量计算的P2P网格技术和基础设施，并在几个应用中展示其有效性。为了利用办公室中的PC等潜在计算资源作为网格计算的计算源，我们开发了一个名为BEE的系统，它使Linux二进制程序能够在Windows下运行网格工作者，并开发了一个网络层，允许通信通过UDP打洞穿越防火墙。在此基础上，提出了一种支持匿名和恶意用户识别的P2P认证方法AUBReX(以交叉证书为代表的伙伴关系认证方法)和一个并行程序…框架通过远程过程调用实现了更多的冲撞模型，连接了由多个支持网格的作业调度系统管理的大规模计算资源池。提出了一种将数据管理层的数据传输从RPC编程中分离出来的模型。我们设计并实现了一个原型数据传输层OmniStorage，它允许高效地传输大量数据。提出了一种支持全球分布式文件系统大容量计算的框架和编程模型。提出了一种基于可扩展哈希的存储负载均衡方法和分布式哈希表相结合的XML数据存储和检索方案。提出了一种主从式并行求解广义特征值问题的方法，适用于P2P网格中的主从式规划模型。我们还开发了适合于像Cell处理器这样的异构型多核系统的数值算法，它可以被认为是P2P网格中一种强大的计算资源。较少

The concept of P2P grid is to make use of PCs in office and laboratory as a computing resource of Grid computing using P2P technology, while the Grid computing is mainly for sharing computing resource in each research organization. Our objective is to investigate and establish the technology and infrastructure of P2P grid for large-capacity computing and demonstrate its effectiveness in several applications.1. To make use of potential computing resources such as PCs in office as a computing source of grid computing, we have developed a system, called BEE, which enables the Linux binary to run a Grid worker in Windows, and a network-layer which allows the communication traversing the firewall by using UDP-hole punching. And, we also proposed an authentication method called AUBReX (Authentication method Using Buddy-buddy relationship Represented by Cross certificate), which supports both anonymity and malicious user identification, for P2P environment, and a framework for a parallel prog … More ramming model by remote procedure calls, which bridge large-scale computing resource pools managed by multiple Grid-enabled job scheduling systems.2. We proposed a model to separate the data transfer by a data management layer from the RPC programming. We have designed and implemented a prototype data transfer layer, OmniStorage, which allows efficient data transmission of a large amount of data. And we proposed a framework and programming model to support large capacity computing on global distributed file system. We have proposed a scheme for storage and retrieval of XML data using DHT with storage-load balancing method based on extendible hashing.3. We have developed a master-worker type parallel method for finding several eigenvalues and eigenvectors of a generalized eigenvalue problem, which suitable for master-worker programming models in P2P grid. We have also developed numerical algorithms suited to the heterogeneous multi-core systems such as Cell processor which can be recognized as a powerful computing resource in P2P grid. Less