SOFTWARE: Collaborative Research: Scalable Software Systems for Large Internet Servers

Internet上的几个应用程序，如脚本和密码处理、数据库管理和多媒体处理，涉及处理安全、计算密集型、多媒体和高带宽信息。这些应用需要大规模的科学计算、可靠性和服务器节点的高带宽。加州大学戴维斯分校和河滨分校的这个联合项目开发了一个可扩展的分布式软件系统架构，其中对互联网服务器所需的主要功能进行了划分，并根据它们的需求分配资源。这些应用程序的计算量和体系结构要求通过服务器上的实际执行来确定。开发了新的技术来处理系统级问题，如准入控制、功能分发、内容管理、流量管理、资源的动态重新配置、进程间通信和轻量级I/O。这些新技术在原型系统上实现，该原型系统使用可用的现成组件和特殊用途设备的组合构建。该项目的成果将影响未来大型互联网服务器的开发和管理。该项目是与英特尔公司的研究人员合作进行的，将有可能在未来进行技术转移。

Several applications over the Internet, such as script and cryptographic processing, database management, and multimedia processing, involve processing of secure, computation-intensive, multimedia, and high-bandwidth information. These applications require large-scale scientific computing, reliability, and high-bandwidth at the server nodes. This joint project between two campuses of the University of California (Davis and Riverside) develops a scalable distributed software system architecture, where the major functionalities required in the Internet servers are partitioned, and resources are allocated on the basis of their needs. The amount of computation and architectural requirements of these applications are determined through actual execution on servers. New techniques are developed to handle system-level issues like admission control, functionality distribution, content management, traffic management, and dynamic reconfiguration of resources, inter-process communication, and lightweight I/O. They are implemented on a prototype system, which is built using a combination of the available off-the-shelf components and special-purpose devices. The outcome of the project will impact the development and management of future large-scale Internet servers. The project is conducted in collaboration with researchers at Intel Corporation, and would have a potential for future technology transfer.