【NDN基础】Named Data Networking 学习笔记

弄懂的问题：

（1）Page 4：PIT表的设计考虑：

          <1>支持多播数据交付。

          <2>控制流量负载以实现流平衡。

          <3>PIT条目是路由器负载的指示器，限制PIT表的大小，限制了DDoS攻击的效果。

          <4>而第四点我并不理解：PIT条目超时提供了相对容易的攻击检测，每个PIT条目中兴趣包到达的接口信息支持推回机制。

          原文：The PIT state serves other valuable purposes. First, since it records the set of interfaces over which the Interests for the same data name have arrived, it naturally supports multicast data delivery. Second, since each Interest retrieves at most one Data packet, a router can control the trac load by controlling the number of pending Interests to achieve flow balance. Third, the number of PIT entries is an indicator of router load; constraining its size limits the e ffect of a DDoS attack. Finally, PIT entry timeouts o ffer relatively cheap attack detection, and the arrival interface information in each PIT entry could support a push-back scheme.

（2）Page 4：传输层功能：

          <1>NDN体系结构没有单独的传输层，而是在应用程序、支持库、转发面板的策略模块实现当前IP体系结构的传输层功能。

          <2>当前IP体系结构的端口号和序列号在NDN中不需要。

          <3>所有需要传输的信息都在包名当中。

          原文：The NDN architecture does not have a separate transport layer. It moves the functions of today's transport protocols (demultiplexing, reliable delivery, and congestion control) into applications, supporting libraries, and the strategy module of the forwarding plane. Transport-layer information such as port and sequence numbers are unnecessary; all information required for transport is in the Data names.

（3）Page 4：NDN的流平衡机制：

          <1>通过在每一跳限制待定兴趣包的数量控制流量负载，在整个网络内提供有效的拥塞控制。

          <2>如果因拥塞发生丢包，缓存可以缓解丢包带来的影响。重传的兴趣包可以在丢包节点之前被已经缓存的数据包满足。

          <3>NDN可以避免当包靠近目的地时发生丢失的现象，因为兴趣包的重传会消耗很多带宽。

          原文：NDN's flow balance requirement, together with the ability of nodes to control their own trafic load by limiting the number of pending Interests at each hop, can provide eff ective congestion control throughout the network. If congestion losses occur, caching mitigates the impact since retransmitted Interests can be satis ed by cached Data packets right before the point of packet losses. Thus, NDN can avoid the kind of congestion collapse that can occur in today's Internet when a packet is lost near its destination and repeated retransmissions from the original source host(s) consume most of the bandwidth.

（4）Page 6~Page 7：开源软件支持：

           <1>CCNx。

           <2>NFD。

           <3>ndnSIM

           <4>NDN testbed

存在的问题：

（1）Page 2：NDN中的流平衡与TCP/IP中的TCP ACK有什么相似之处？什么是反馈环路（feedback loop）？

           原文：Data packets always take the reverse path of Interests, and, in the absence of packet losses, one Interest packet results in one Data packet on each link, providing  flow balance. To fetch large content objects that comprise multiple packets, Interests provide a similar role in controlling trafic flow as TCP ACKs in today's Internet: a fine-grained feedback loop controlled by the consumer of the data.

（2）Page 2：什么是flat names？什么是flat routing？

          原文：Flat names can be accommodated as a special case, likely useful in local environments, however hierarchical namespaces are essential both in scaling
the routing system and in providing necessary context for the data. (Even advocates of  flat routing acknowledge that  at names scale by introducing some hierarchy.)

（3）Page 3：Interest selectors不了解。

          原文：(2) Interest selectors in conjunction with longest prefi x matching retrieve the desired data through one or more iterations. Our experience so far suggests that a simple set of selectors can support retrieving data with partially known names.

（4）Page 3：以下这句话不理解，需要学习文献[12]："cognitive dimensions" framework。

          原文：Enabling application developers, and sometimes users, to design their own namespaces for data exchange has several benefi ts: increasing the closeness of mapping between an application's data and its use of the network; reducing the need for secondary notation (record-keeping to map application con figuration to network con gfiuration); and expanding the range of abstractions available to the developers.

（5）Page 3：什么是web-of-trust？什么是pre-agreed trust anchors？

          原文：web-of-trust to enable secure communication without requiring pre-agreed trust anchors.

（6）Page 4：in-network storage和Content Delivery Networks (CDNs)的结合。

           原文：In addition to the Content Store, the architecture now supports a more persistent and larger-volume in-network storage, called a Repository (Repo for short). This type of storage can support services similar to that of today's Content Delivery Networks (CDNs), without having to engineer them as an application layer overlay using creative protocol tricks (e.g., DNS manipulation) to make them work.

（7）Page 5：NDNVideo application不了解。

           原文：The NDNVideo application does not require direct communication between publisher and consumer, enabling publisher independent scalability through NDN's use of in-network storage. Applications that perform on-the-fly assembly of content or selection of video sections, i.e., frame-level random access requirements, are supported directly through namespace design.

（8）Page 5~Page 6：OSPFN路由协议设计不了解。NLSR路由协议不了解。

          原文：Routing Protocol Design部分。

                      Exploring New Routing Paradigms部分。

（9）Page 6：NLSR的信任管理不了解。

          原文：For example, the security needs of NLSR required development of a simple hierarchical trust model in which keys are published with names that reflect their trust relationship. A root key is owned by the network domain's administrator, and below the root are site keys, each owned by a single site's administrator, signed by the root key and published in the next level of the hierarchy. Each site key then signs the site's operator keys, which in turn sign router keys, which in turn sign the key of the NLSR process on that router. Finally, the NLSR key signs the routing data originated by NLSR.