NeTS: Small: Abstractions and Algorithms for Control of Network Services

点击翻译按钮获取中文摘要

Modern data-center networks use software-defined networking to control the flow of traffic among their systems. Current abstractions for software-defined networking assume that traffic traverses the network without being modified en route. In practice, however, many network operators utilize "middleboxes" to perform complex traffic processing operations, such as deep packet inspection, load balancing, WAN optimization, and caching. The goal of this project is to achieve a transformational shift in software-defined networks from today's limited "data transmission" view to a "data transmission and processing" view that more accurately reflects the reality of current and future networks in order to improve control and operational efficiently of the network infrastructure.The project aims to design and prototype novel mechanisms to describe, implement, and control advanced network services in networks with centralized and hybrid management. Specific objectives are to develop a novel control plane for network services, design novel control plane algorithms, explore dynamics in network traffic that affect the workload on middleboxes, and demonstrate the capabilities of our proposed control plane for network services in a realistic setting based on GENI. The results from this work will provide a novel approach to controlling network services within software-defined networks.The broader impacts of this project aim at improving network infrastructure, which makes up the Internet and is of critical importance to society. While software-defined networking simplifies and enhances the control that is exerted on networks, it falls short of considering middleboxes and network services, which are an integral part of today's network infrastructure. This project addresses this shortcoming and provides solutions that can make software-defined networks more useful in the future. This project will provide training to student in state-of-the-art networking technology, and results will be disseminated in the form of open-source toolsets and publications.