A Lightweight Forwarding Strategy for Named Data Networking in Low-end IoT

https://doi.org/10.1016/j.jnca.2019.102445Get rights and content

Abstract

The proliferation of Internet of Things (IoT) devices has given a new dimension to networking operations. In fact, IoT issues in terms of complex communication patterns, dynamic topologies and security force to re-think basic networking operations in a more adapted way. Moreover, the IoT needs contribute to show the limitations of the IP model to support content-based applications, and the emergence of the Information Centric Networking (ICN) paradigm seems to be more compliant with the IoT vision. In this context, to enable the ICN model in IoT devices, we present a lightweight forwarding strategy for Named Data Networking (NDN) over IEEE 802.15.4. NDN is an ICN architecture with a great potential for the IoT and future Internet. This paper presents a forwarding strategy that reduces network overhead to the bare minimum, while keeping satisfactory performance in different IoT application scenarios. To forward named contents without node addresses, the strategy is based on a reinforcement learning technique that provides an accurate broadcast-based forwarding with a reduced overhead.

Introduction

Massive technological innovations towards electronic miniaturization, associated with the deployment of manifold wireless technologies, have fostered the emergence of billions of small devices connected over the Internet creating the omnipresent Internet of Things (IoT) (Yaqoob et al., 2017). IoT systems are currently deployed on the TCP/IP protocol suite. However, the IP architecture shows its limitations to support the IoT needs. For example, security is still focused on communication channels when the data itself needs to be secured. Moreover, IoT systems need efficient support for resource naming and discovery, which is not easy to deploy with IP in constrained infrastructures (Shang et al., 2016a).

Internet Engineering Task Force (IETF) Working Groups are adapting the TCP/IP stack to suit IoT systems through solutions such as 6LoWPAN (Transmission ofv, 2007) to support IPv6 communication over low-power wireless technologies, and Constrained Application Protocol (CoAP) (The Constrained App, 2014) to handle data exchange in constrained environments. In practice, most IP solutions support the IoT at application level using the Representational State Transfer (REST) architecture, which indicates that the TCP/IP stack has reached its limit to support these new requirements. Moreover, most of those solutions are converging to the principles of the Information Centric Networking (ICN) paradigm (Ahlgren et al., 2012).

Unlike IP networking, ICN operates with human-readable names that identify every piece of content regardless of host locations. Applications use these names to request and fetch data. Location independent names, combined with digital signatures carried in every content item, provides self-secured and reusable data packets that may be cached to satisfy multiple requests. Consequently, ICN natively provides the mechanisms needed to support IoT requirements such as resource identification with human-readable names (Nour et al., 2017), security and access control through self-certified content (Zhang et al., 2018), and mobility by identifying content regardless of host location.

One interesting embodiment of the ICN model is Named Data Networking (NDN) (Zhang et al., 2014). NDN consumer applications request contents from the network by sending Interest packets that pull back Data packets. Native features come along with this principle such as: communication without establishing end-to-end connections and without name-to-address resolution. In addition, it is not necessary to maintain consumer-provider paths, which provides a native support of connection disruption resulting from node mobility.

So far, the suitability of NDN for the IoT has been investigated to some extent (Shang et al., 2016b), but the lightweight aspect of NDN should be improved in order to enable it in current IoT systems. It has been observed in (Baccelli et al., 2014) that an NDN stack implementation in IoT devices (RIOT and Contiki) can save up to 60% of ROM and 80% of RAM compared to the RPL/6LoWPAN stack. Moreover, according to the same study, a simple flooding mechanism in an NDN wireless network generates three times fewer packets than RPL/6LoWPAN routing. Even if the superiority of NDN in terms of performance remains to be demonstrated, the data-centric approach of NDN can be an alternative to the 6LoWAPN stack in future IoT solutions (Shang et al., 2016c).

To deploy NDN in IoT devices, one of the first challenges is to support the NDN forwarding in low-rate wireless networks, such as IEEE 802.15.4 (The IEEE, 2019). NDN wireless forwarding strategies are generally based on a broadcast-and-learn mechanism. In other words, nodes exploit broadcast communications to learn and take efficient forwarding decisions. Two examples of forwarding strategies can be given to illustrate the broadcast-and-learn approach.

The first one is Blind Flooding (BF) (Amadeo et al., 2015) which has been proposed for NDN over IEEE 802.11. In BF, forwarders defer each Interest and Data transmission for a random delay, with Data packet delays always shorter than the delays of Interest packets. While waiting the delay, a forwarder listens to the channel if a packet with the same name is transmitted. This way, any forwarder can cancel its pending transmission if a neighbor node is sending a similar Interest/Data packet. This mechanism significantly reduces overhead and packet redundancy, but still too cumbersome for constrained IoT environments such as IEEE 802.15.4 networks.

The second example is Reactive Optimistic Name-based Routing (RONR, detailed in Section 3) in which forwarders broadcast Interests to discover a content only when no knowledge is available about the content name. After the Data packet is received, each node keeps a mapping between the prefix-name of the content and the MAC address of the node that sent the Data packet. The mapping will be used to find the next-hop MAC address to forward subsequent Interests with unicast. Interest broadcast is triggered periodically or when the mapping is no longer correct. Although this lightweight mechanism is designed for constrained devices, we show in this paper that unicast is not always the best approach, especially with mobile devices, is the sens that the reusable nature of NDN packets is limited by unicast addresses.

In short, nodes in the first forwarding strategy example use broadcast to take forwarding decisions on-the-fly, whereas in the second example nodes use broadcast to maintain forwarding information. Both approaches have their advantages and inconvenients as mentioned.

To get the best of both worlds, low-end IoT environments need a lightweight forwarding strategy that combines the data dissemination/discovery efficiency of BF with the minimum overhead generated by RONR. That is to say, we need to take advantage of the broadcast while reducing transmission overhead. To that purpose, we design and evaluate in this paper a Reinforcement-based Lightweight Forwarding (R-LF) strategy for NDN over constrained networks.

The idea of R-LF is to make the broadcast as accurate as the unicast, in terms of transmitted frames and forwarding decisions.

As mentioned, in NDN wireless forwarding approaches, nodes have to get knowledge about the content (routing) and propagate packets (forwarding) in tightly coupled operations, taking advantage of the broadcast. Thus, these mechanisms, including R-LF, can be considered as routing and forwarding solution at once, but we follow related literature which widely adopts the term forwarding or forwarding strategy.

Instead of targeting optimal performance for one configuration, which can be achieved more efficiently with a proactive mechanism, our objective is to design a lightweight tradeoff technique that can keep satisfactory performance in different communication patterns such as multiple data flows, multi-consumer, multi-source, and dynamic topology changes as shown in the evaluation.

Our contribution through this paper can be summarized as follows:

  • We study the importance of the broadcast mechanism and we draw up guidelines for designing a forwarding strategy over low-rate wireless networks.

  • We propose the R-LF strategy based only on content names and broadcast communications.

  • We show by simulation that our strategy can fit different small-scale IoT scenarios in comparison to all-unicast and all-broadcast forwarding approaches mentioned above.

  • We deploy NDN with R-LF on Arduino devices and we give measures on its lightness aspect.

The rest of this paper is organized as follows: Section 2 gives an overview of NDN, focusing on the forwarding aspect. Section 3 presents relevant forwarding approaches for NDN in wireless networks. In Section 4, we show the importance of the broadcast in wireless NDN forwarding, followed by our guidelines for a lightweight forwarding strategy design. Section 5 details the proposed strategy and its relevant steps, operations and parameters. Section 6 reports the results of the strategy evaluation in different communication scenarios. Section 7 concludes the paper with a brief discussion and future improvements.

Section snippets

Named Data Networking

NDN is an ICN protocol in which the communication is completely based on content names, while the IP protocol uses source and destination addresses to deliver contents.

In NDN, content chunks are decoupled from the producer location, and each data packet must be signed by its producer. Thereby, routers can reuse the same packet to satisfy other consumer requests for the same content, providing NDN with a native caching mechanism. The connectionless communication model of NDN gives it a native

Related Work

Even if most of the following wireless forwarding solutions are designed for MANETs and do not consider constrained IoT devices, they represent an interesting starting point to design a wireless NDN forwarding strategy.

In (Wang et al., 2012), a broadcasting protocol has been designed for data dissemination in vehicular networks. Timers are used to defer packet forwarding similarly to BF. During the waiting time, if a node overhears a packet with the same name, the delayed packet is dropped,

NDN-MAC Mapping and Design Guidelines

In this section, we simulate a simple IoT scenario to show the impact of NDN-MAC mapping on forwarding performance. Using the unicast approach described above (RONR), other mappings can be envisioned and evaluated. Therefore, instead of using a unicast address to send both Interest and Data packets, nodes can use one of the following mappings:

  • 1.

    Interest Broadcast, Data Broadcast (IBDB). All nodes forward Interests by broadcast when a matching prefix is found in the FIB. When a node receives a

Overview

Routing and forwarding operations are significantly different in NDN and IP. In IP, only the routing operation is smart in the sense that different routing protocols can be envisioned. The forwarding operation always consists in finding the longest match available in the routing table and sending the packet to the corresponding next hop. In NDN however, in addition to the routing operation that can be smart as in IP, multiple approaches are possible to handle packet forwarding with more or less

Evaluation

For evaluation purpose, we implemented the NDN module in the OMNeT++ simulator (OMNet++) including R-LF, BF and RONR forwarding strategies.2 We first study the impact of the learning rate (α) on R-LF performance, and we fix the values for parameters M, m and Th. After that, we evaluate R-LF in comparison with RONR, a completely unicast approach, and BF which is a broadcast-only mechanism that uses delayed transmissions.

Discussion and Conclusion

In the three simulated scenarios, the R-LF performance is always close the best one among RONR and BF for all the metrics. This shows the adaptability of R-LF to different communication scenarios. The only exception to this observation is the RTT in the multi-consumer scenario, which certainly requires more sophisticated techniques for forwarding cached Data packets.

In our strategy, the random Interest flooding occurs rarely, as the delta adjustment step can be used to compute a waiting delay

Acknowledgments

This work is supported by the 17MDU988 project which is a PHC (Partenariat Hubert Curien) Tassili, a French-Algerian cooperation program. This Tassili project is funded by the French ministries (MEAE and MERI) and the Algerian ministry MESRS.

Samia Bouzefrane is Associate Professor and has the accreditation to conduct research (HDR) at the Conservatoire National des Arts et Métiers (CNAM) of Paris. She received her PhD in Computer Science from the University of Poitiers (France) in 1998. After four years at the University of Le Havre (France), she joined in 2002 the CEDRIC Lab of CNAM. She is the co-author of many books (Operating Systems, Smart Cards, and Identity Management Systems). Her current research areas cover security in

References (22)

  • M. Amadeo et al.

    Forwarding strategies in named data wireless ad hoc networks: design and evaluation

    J. Netw. Comput. Appl.

    (2015)
  • D. Saxena et al.

    Named data networking: a survey

    Computer Science Review

    (2016)
  • B. Ahlgren et al.

    A survey of information-centric networking

    IEEE Commun. Mag.

    (2012)
  • E. Baccelli et al.

    Information centric networking in the iot: experiments with ndn in the wild

  • J.A. Boyan et al.

    Packet routing in dynamically changing networks: a reinforcement learning approach

  • R. Chiocchetti et al.

    Inform: a dynamic interest forwarding mechanism for information centric networking

  • P. Kietzmann et al.

    The need for a name to MAC address mapping in NDN: towards quantifying the resource gain

  • M. Michael et al.

    Listen first, broadcast later: topology-agnostic forwarding under high dynamics

  • B. Nour et al.

    M2hav: a standardized icn naming scheme for wireless devices in internet of things

  • OMNet++

    OMNeT++ discrete event simulator

  • W. Shang et al.

    Challenges in IoT Networking via TCP/IP Architecture

    (February 2016)
  • Cited by (28)

    • Content-centric data and computation offloading in AI-supported fog networks for next generation IoT

      2022, Pervasive and Mobile Computing
      Citation Excerpt :

      This increase also raises problems at different levels such as data transfer and networking, data processing and management, context awareness and service provisioning [9]. From data and scalability perspectives, unlike FC, traditional IP-based networking architecture is not designed to fulfill the requirements of future IoT systems [10]. IP-based networking is inherently designed for host-to-host communication where host addresses and locations play a vital role [11].

    • Machine learning based solutions for security of Internet of Things (IoT): A survey

      2020, Journal of Network and Computer Applications
      Citation Excerpt :

      The Internet of Things (IoT) interlink electrical devices with a server and exchanges information without any human intervention (Li et al., 2011; Abane et al., 2019; Sheng et al., 2013; HaddadPajouhet al., 2019).

    • On estimating the interest satisfaction ratio in IEEE 802.15.4-based named-data networks

      2024, Annales des Telecommunications/Annals of Telecommunications
    View all citing articles on Scopus

    Samia Bouzefrane is Associate Professor and has the accreditation to conduct research (HDR) at the Conservatoire National des Arts et Métiers (CNAM) of Paris. She received her PhD in Computer Science from the University of Poitiers (France) in 1998. After four years at the University of Le Havre (France), she joined in 2002 the CEDRIC Lab of CNAM. She is the co-author of many books (Operating Systems, Smart Cards, and Identity Management Systems). Her current research areas cover security in cloud computing and internet of things, resource allocation in Cloud computing, and new paradigms for networking such as NFV and NDN.

    View full text