CC* Integration-Large: In-Network Distributed Infrastructure for Advanced Network Applications

这个项目让互联网“更智能”。互联网设计的“端到端”原则指出，组成网络的路由器、交换机、WiFi基站等应该只转发数据包。在许多情况下，中间阶段的少量处理可以使Internet应用程序更高效。云计算服务对社会产生了重大影响。许多人设想一个更强大的互联网，像云一样的服务分布在整个网络，一直到网络的边缘。这个项目正在探索一个智能网络，在互联网的结构中嵌入小型服务器。该项目结合了一种新颖的轻量级执行模型，适用于高效的网络内微服务，以及拓扑感知的协调和控制层，以及用于表达所需应用程序服务和需求的策略语言。称为InLocus的执行模型旨在弥合网络数据包处理器和有状态网络服务之间的差距。InLocus可以在传统的微处理器和微控制器上运行，也可以在现场可编程门阵列（fpga）等加速器上运行。这样可以用最少的资源实现必要的功能。协调框架监控网络（通过对PerfSONAR网络监控系统的扩展），并可以做出智能的服务放置和实例化决策。该项目将整合和评估两个重要的应用环境，一个用于地球科学，另一个用于分布式机器学习。这些都将与NSF网络基础设施资源相结合。该项目将对科学和商业应用及其用户产生影响。正如互联网在过去几年触动了许多人的生活一样，一个更好的互联网将继续推动创新和用户体验的质量。我们正在探索的技术被广泛认为是充分实现自动驾驶汽车和智能城市以及增强现实等面向用户的能力所必需的。此外，我们在不同层次的课堂和研讨会上教授这个项目的许多原则，因为它们体现了互联网的历史和未来。这个项目的网站是http://indiana.open.luddy.indiana.edu/，在那里可以找到项目概述、组件描述和源代码库的链接。该项目构建并集成了以前资助的项目。这些先前的项目即使在资助期结束后仍然是开源的。此外，核心功能得到了维护，因为其他工作都依赖于它。这个项目的数据和代码将尽可能长时间维护。该奖项反映了美国国家科学基金会的法定使命，并通过使用基金会的知识价值和更广泛的影响审查标准进行评估，被认为值得支持。

This project makes the Internet "smarter". The "end to end" principle of Internet design states that the routers, switches, WiFi base stations, etc. that comprise the network should only forward packets. There are many instances in which small amounts of processing at intermediate stages can make Internet applications more efficient. Cloud computing services have had a significant impact on society. Many envision an even more powerful Internet with cloud-like services distributed throughout and all the way to the edge of the network. This project is exploring a smart network with small servers embedded in the fabric of the Internet.This project brings together a novel lightweight execution model suitable for efficient in-network microservices, and a topology aware coordination and control layer, with a policy language to express desired application services and requirements. The execution model, called InLocus, is designed to bridge the gap between network packet processors and stateful network services. InLocus runs on conventional microprocessors and microcontrollers, as well as on accelerators like Field Programmable Gate Arrays (FPGAs). This enables necessary functionality with minimal resources. The coordination framework monitors the network (with extensions to the PerfSONAR network monitoring system) and can make intelligent service placement and instantiation decisions. This project will integrate and evaluate two important application environments, one in Earth Sciences and one applicable to distributed machine learning. Each of these will work in conjunction with NSF cyberinfrastructure resources.This project stands to impact scientific and commercial applications and their users. Just as the Internet has touched many lives in the past years, a better Internet will continue to drive innovation and users' qualities of experience. Techniques like the ones we are exploring are widely regarded as necessary to fully realize autonomous vehicles and smart cities, as well as user-oriented capabilities like augmented reality. Additionally, we teach many of the tenets of this project in classes and seminars at various levels, as they embody the history and the future of the Internet.This project's web site is at http://indiana.open.luddy.indiana.edu/ and a project overview, component descriptions, and links to source code repositories can be found there. This project builds, and integrates, previously funded projects. These previous projects remained open source even after their period of funding ended. Further, the core functionality has been maintained as other work depends on it. This project's data and code will be maintained as long as possible.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.