KCN: Knowledge Centric Networking

最近出现的杀手级应用，如内容分发、云计算和物联网(IoT)，都要求底层网络能够理解特定的服务上下文。在这个项目中，我们提出了以知识为中心的网络(KCN)范式，在这个范式中，知识处于网络格局的中心。其目标是实现网络内知识的生成和分发，以便开发必要的网络控制智能，以处理复杂性和不确定性。为了实现这一目标，将制定具体的算法和机制/协议，用于在互联网的单一管理域内和跨同类/异质管理域的知识获取、处理、传播和组织。该项目将研究基于软件定义网络(SDN)原则的三种类型的知识交换：知识作为工具(KAAT)、知识作为服务(KAAS)和知识作为云(KAAC)。KAAT将在动态网络环境中实现智能网络运营，该环境由在不同优势点收集的知识驱动。我们倡导分层的知识框架，在这种框架中，知识和控制功能分布在网络中的正确位置，以完成特定的控制任务。此外，我们还将探讨互联网市场中不同参与者之间的知识共享。这可以通过从知识提供者到知识消费者(CAAS)的显式知识转移来实现，也可以基于开放的知识云来实现，在开放的知识云中，知识消费者可以通过开放但受控的知识生态系统(KAAS)发布或订阅信息。KCN驱动的内容流量在未来移动互联网的不同无线电接入技术之间的分流旨在通过考虑与内容、用户和网络条件相关的各种知识来实现自适应资源控制。此外，KCN驱动的能量管理根据获取的知识和动态变化的上下文信息，针对IP层和物理光层的跨层节能技术，直接为Ti3子挑战1、2、3和4做出贡献。首先，基于KCN的知识生态系统将为下一代互联网配备必要的智能，以应对动态条件下的复杂需求。这样一个生态系统，与SDN架构无缝结合，将能够优雅地支持未来联网服务和众多用户日益增长的复杂性和异构性。这两个互补的用例展示了拟议的KCN框架将如何在两个不同的应用领域实例化，即移动/无线接入网络中的内容流量分流和IP/光传输网络中的能源效率。用例1有助于第三个子挑战，即基于知识的内容缓存和流量分流技术，用于未来面向内容的移动互联网。另一方面，使用案例2通过IP层和光纤层的智能节能机制完成了第2个子挑战。最后，通过基于原始上下文信息的网络内知识推理和学习，该项目还解决了从数据中提取理解的第四个子挑战。总而言之，在网络/服务运营期间捕获的上下文信息将被用于获取系统的网内知识和情报，以便自适应地处理复杂性和不确定性，并实现接近最佳的网络运营。

The recent advent of killer applications such as content distribution, cloud computing and Internet of things (IoT), all require for the underlying network to be able to understand specific service contexts. In this project we propose the Knowledge Centric Networking (KCN) paradigm, in which knowledge is positioned at the centre of the networking landscape. The objective is to enable in-network knowledge generation and distribution in order to develop necessary network control intelligence for handling complexity and uncertainty. In order to achieve this, specific algorithms and mechanisms/protocols will be developed for knowledge acquisition, processing, dissemination and organisation both within single and across homogeneous/heterogeneous administrative domains in the Internet.The project will investigate three styles of knowledge exchange based on Software Defined Networking (SDN) principles: Knowledge as a Tool (KaaT), Knowledge as a Service (KaaS) and Knowledge as a Cloud (KaaC). KaaT will enable intelligent network operations in dynamic network environments driven by knowledge gathered at different vantage points. We advocate a hierarchical knowledge framework in which knowledge and control functions are distributed at the right places within the network for fulfilling specific control tasks. In addition, we will invetigate knowledge sharing between different players in the Internet marketplace. This can be achieved either through explicit knowledge transfer from a knowledge provider to a knowledge consumer (KaaS), or based on open knowledge clouds where knowledge prosumers may publish or subscribe to information through an open but controlled knowledge ecosystem (KaaS).The proposed KCN architecture will be validated through two complementary use cases. KCN-driven content traffic offloading between heterogeneous radio access technologies for the future mobile Internet aims to achieve adaptive resource control by taking into account a wide variety of knowledge associated with content, users and network conditions. In addition, KCN-driven energy management targets cross-layer energy saving techniques at both the IP and the physical optical layer according to the derived knowledge and dynamically changing context information.The project provides direct contributions to the TI3 sub-challenges 1, 2, 3 and 4. First of all, the KCN-based knowledge ecosystem will equip the next generation Internet with necessary intelligence for handling complex requirements under dynamic conditions. Such an ecosystem, seamlessly coupled with the SDN architecture, will be able to gracefully support the ever increasing complexity and heterogeneity of future networked services and multitude of users. The two complementary use cases demonstrate how the proposed KCN framework will be instantiated in two different application domains, content traffic offloading in mobile/wireless access networks and energy efficiency in IP/optical transport networks. Use case 1 contributes to the 3rd sub-challenge, with knowledge-based content caching and traffic offloading techniques for the future content-oriented mobile Internet. Use case 2, on the other hand, contributes to the 2nd sub-challenge with intelligent energy saving mechanisms at both the IP and optical layer. Finally, with in-network knowledge inference and learning based on raw context information, the project also addresses the 4th sub-challenge of extracting understanding from data. In summary, context information captured during network/service operation will be used to derive systematic in-network knowledge and intelligence in order to deal adaptively with both complexity and uncertainty and enable near-optimal network operation.