Session persistence for dynamic web applications in Named Data Networking

Abstract

Named Data Networking (NDN) is a promising future network architecture. It greatly facilitates static Web content distribution due to its information-centric communication paradigm, novel in-network caching, and name-based routing features. However, NDN still lacks efficient support for dynamic Web applications. Specifically, the inefficient communications are attributed to inherent features of NDN, such as its being multisource and multicast, as well as its decoupling of content from location. A dynamic Web request generally includes dozens of Interest packets. Since NDN lacks efficient support for the session persistence of dynamic Web applications, existing name-based forwarding strategies cannot effectively forward those Interest packets that belong to the same session to the specific server in cases where multiple server clusters are deployed by the Application Service Provider (ASP). This, therefore, leads to the packet retransmission problem: retransmitted Interest packets waste network resources and degrade the performance of dynamic Web applications. How to effectively maintain session persistence is worth considering for an NDN network. To this end, we design and implement a system framework that aims to enhance dynamic Web applications over NDN through a session-oriented approach. First, we propose a session-based forwarding model that maintains a virtual connection during an online interaction. Second, we design a robust forwarding scheme that can balance the dynamic content traffic with a fault recovery capability. Finally, we implement our approach and conduct experiments on the ndnSIM, network topology in the simulation is designed based on the China Education and Research Network 2. Experimental results show that the proposed approach significantly improves network utilization and reduces the service delay, i.e., 90.36%, 50.23% utilization improvement and 84.83%, 45.68% service delay reduction, respectively, compared with two existing approaches.