Self-Managing Named Data Networks

点击翻译按钮获取中文摘要

Named data networks (NDNs) represent a nascent vision for the future Internet that is ideal for military and industrial applications. By design, NDNs break from the conventional end-to-end, host-centric, communication and hence eliminate the thin hourglass IP waist in the current Internet stack and it accompanying security vulnerabilities. NDNs focus on movement of packets uniquely identified by their content names without their source and destination hosts' identities. Hence, frequently requested objects (e.g., web pages, streamed media and scientific data) can be cached in routers and delivered without repeated per-user transmissions. NDNs maintain a simple philosophy of limiting network traffic to content interest and corresponding data packets and are said to democratize the Internet since any entity can publish and consume data. By construction, they avoid many of the current problems including mobility, security and the high cost of content distribution.******Full realization of NDNs poses a number of obstacles. To avoid repeated scenarios of abandoning new technologies (e.g., IPv6), architectures must be gradually deployed and co-exist with the current infrastructure, but eventually installed as a clean slate solution. Furthermore, the newly emerging big-data and Internet-of-things applications along with the futuristic vision of the Internet as a searchable content directory is leading to an unprecedentedly massive size of referenced content namespaces, large routers' forwarding tables and a major scalability issue. Indeed, creating, looking up and managing exponentially increasing routing and caching tables is a major challenge. Finally, analogous to the ever pressing need for service differentiation in IP networks, content differentiation with efficient business models must be developed.******This proposal aims at designing and implementing an efficient self-management architecture for NDNs. It takes advantage of two prominent technologies; the first, virtual network (VN) slicing, is employed to achieve a scalable hierarchy of isolated but collocated and dynamically customizable NDNs. Each slice serves an intelligently and dynamically partitioned subspace of contents providers. NDNs routers' resources will be virtualized and sliced adaptively to react to changes in contents supply and demands. Partitioning and resource optimization techniques will be investigated to achieve functional scalability of these networks. Finally, through the use of the second technology, namely, network software-ization through virtual network functions, the proposed work will realize fair, application agnostic, ultra-low delay self-adaptive routing as well as content and service differentiation. ******The significance of the developed architecture and algorithms will stem from their ability to self-evolve to enhance their performances or to face changes in the problem domains.