CAREER: A Network Motion Picture Primitive for Network Monitoring and Control

点击翻译按钮获取中文摘要

The Internet is an essential utility for modern life, providing connectivity and access to information any time, anywhere, and to many services that are important for businesses and individuals. Thus, the reliability and availability of the networks that underlie the Internet are critical, and it is important to have a detailed and holistic understanding of their behavior. However, as networks have become larger and more complex, visibility into fine-grained, network-wide behavior that allows network operators to identify and repair problems is increasingly difficult to come by. This project addresses that challenge, with a primary goal of developing tools and techniques that enable unprecedented visibility into the behavior of large and complex computer networks.The goal of this project is the development of a novel primitive that provides whole-network visibility: the Network Motion Picture, a rapid sequence of network-wide snapshots, where each snapshot is a consistent set of local observations of state that provide a view of the network at approximately a single instant in time. In contrast to existing research that focuses on either developing more advanced metrics for individual components/paths or improving inference of network-wide conclusions from those measurements, the proposed primitive makes two key contributions: (1) it treats emergent and network-wide behavior as a first-class citizen, and (2) it develops an abstraction for scalable collection of and adaptation to changes in network-wide behavior. The planned research has three main technical components: i) the development of a practical framework for collecting Network Motion Pictures; ii) novel systems for diagnosing, managing, and repairing networked systems; iii) novel protocols for improving the performance of large, complex networks. Use cases will include data center, edge, and Internet of Things (IoT) networks. The researcher plans evaluation using the NSF=funded Cloudlab platform.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.