NDNrel: A mechanism based on relations among objects to improve the performance of NDN

Abstract

Named-Data Networking (NDN) is a promising architectural approach that focuses on the efficient distribution of data objects. Each of these objects represents an individual piece of content that is uniquely named and can be cached by network nodes. Recent work on Information-Centric Networking (ICN) explores new uses of the data object concept to enable advanced applications based on content distribution. Such work includes the use of metadata and manifests to describe contents as sets of multiple correlated objects. In this paper, we extend our previous work on a model and a backward-compatible mechanism, named NDNrel, that allows publishers to distribute contents as related objects. The new version of NDNrel reduces the cost of distributing relation information to clients by leveraging the usage of metadata and manifests on ICN architectures. Also, it considers additional aspects such as the authentication of relations and data that comprise the content. The evaluation of NDNrel is revisited and extended to analyze the performance and overhead caused by the usage of the proposed relation model in two scenarios: distribution of multimedia and Web content. Results demonstrate that using NDNrel to structure the contents as objects and relations outperforms default NDN, reducing the average client download time and the global network traffic in at least 28% and 34% respectively.