Elsevier

Computer Networks

Volume 126, 24 October 2017, Pages 125-140
Computer Networks

Lightweight On-demand Ad hoc Distance-vector Routing - Next Generation (LOADng): Protocol, extension, and applicability

https://doi.org/10.1016/j.comnet.2017.06.025Get rights and content

Abstract

This paper studies the routing protocol “Lightweight On-demand Ad hoc Distance-vector Routing Protocol – Next Generation (LOADng)”, designed to enable efficient, scalable and secure routing in low power and lossy networks. As a reactive protocol, it does not maintain a routing table for all destinations in the network, but initiates a route discovery to a destination only when there is data to be sent to that destination to reduce routing overhead and memory consumption. Designed with a modular approach, LOADng can be extended with additional components for adapting the protocol to different topologies, traffic, and data-link layer characteristics. This paper studies several such additional components for extending LOADng: support for smart route requests and expanding ring search, an extension permitting maintaining collection trees, a fast rerouting extension. All those extensions are examined from the aspects of specification, interoperability with other mechanisms, security vulnerabilities, performance and applicability. A general framework is also proposed to secure the routing protocol.

Introduction

Low-power and Lossy Networks (LLNs) are composed of Constrained Devices, i.e., devices with strictly limited computational power and storage (1-2 MHz CPUs and a couple of KB of memory), which are communicating over a channel characterised by a high probability of packet losses, typically very small frame sizes, and very limited throughput. Transiting data across such a network, especially when multiple hops are present between the source and the destination, is a challenging task: routing protocols must be frugal in their control traffic and state requirements, as well as in algorithmic complexity. Even once paths have been found, these may be usable only intermittently, or for a very short time, due to changes on the channel such as persistent interference (requiring rediscovery of a usable path). Channel failures, resulting in link failures in a routing path can result from a variety of factors such as heterogeneity of sender and receiver hardware, power supply or power control algorithms, the presence of noise or interferences, or even device failure causing a previously selected intermediary router along a path to no longer be available.

The limitations of the devices and the channel capacity in LLNs suggest a routing protocol of extreme simplicity – yet the fragility and transient nature of links suggest the requirement to be able to quickly discover and establish alternative paths when faced with a link failure. These requirements are, seemingly, contradictory. A “standard” proactive routing protocol, such as OSPF (Open Shortest Path First) [1] or OLSR (Optimised Link State Routing) [2], [3], maintaining a network topology graph, would remove a “broken” link from its graph and re-run a shortest path algorithm – incurring the requirement of each routing device having sufficient memory to store (up to) the complete network topology, as well computational power allowing it to frequently re-run a shortest path algorithm. A “standard” on-demand routing protocol would in the same situation incur path re-discovery, with additional control signals being imposed on the network, as well as additional delays on data packet delivery whilst path re-discovery is ongoing, and either buffering of data packets for that duration or retransmission once a path has been re-discovered.

Since the late 90s, the Internet Engineering Task Force (IETF)1 has embarked upon a path of developing routing protocols for networks with increasingly more fragile and low-capacity links, with less pre-determined connectivity properties, and with increasingly constrained router resources. This, in ’97, by chartering the MANET (Mobile Ad hoc Networks) working group, then subsequently in 2006 and 2008 by chartering the 6LoWPAN (IPv6 over Low power WPAN) and ROLL (Routing Over Low power and Lossy networks) working groups.

The MANET working group converged on the development of two protocol families: reactive protocols, including AODV (Ad hoc On-demand Distance Vector routing [4]) and DSR (Dynamic Source Routing [5]), and proactive protocols, including the OLSR (Optimised Link State Routing [2]) and TBRPF (Topology dissemination based on reverse-path forwarding [6]). Distance vector protocols operate in an on-demand fashion, acquiring and maintaining paths only while needed for carrying data, by way of a Route Request/Route Reply exchange. Proactive protocols are based on periodic control messages exchanges, where each router proactively maintains a routing table with entries for all destinations in the network. A sizeable body of work exists, including [7], studying the performance of these protocols in different scenarios, and justifying their complementarity. For the purpose of this paper, it suffices to observe that proactive provides low delays and predictable, constant control overhead – at expense of requiring memory in each router for maintaining complete network topology. Reactive protocols limit the memory required for routing state to that for actively used paths, at the expense of delays for the Route Request/Route Reply exchange to take place, and control overhead dependent on data flows.

After acquiring operational experiences with DSR, AODV, TBRPF, and OLSR, the MANET working group commenced developing successors to these protocols, denoted OLSRv2 and DYMO. Whereas a relatively large and active community around OLSR thus standardised OLSRv2 [3], the momentum behind DYMO (renamed to AODVv2 in 2013) diminished, and development of reactive routing protocols was abandoned by the MANET working group in 2016.2

The 6LoWPAN working group was chartered for adapting IPv6 for operation over IEEE 802.15.4, accommodating characteristics of that data-link layer, and with a careful eye on resource constrained devices (memory, CPU, energy, ...). Part of the original charter for this working group was to develop protocols for routing in multi-hop topologies among such constrained devices, and over this particular data-link layer. Two initial philosophies to such routing were explored: mesh-under and route-over. The former, mesh-under, would, as part of an adaptation layer between 802.15.4 and IP, provide layer 2.5 multi-hop routing, presenting an underlying mesh-routed multi-hop topology as a single IP link. The latter, route-over, would expose the underlying multi-hop topology to the IP layer, where upon IP routing would build multi-hop connectivity. Several proposals for routing were presented in 6LoWPAN, for each of these philosophies, including LOAD (“6LoWPAN Ad Hoc On-Demand Distance Vector Routing” [8]). LOAD was a derivative of AODV, but adapted for L2-addresses and mesh-under routing, and with some simplifications over AODV (e.g., removal of intermediate Route Replies and of sequence numbers). However, 6LoWPAN was addressing other issues regarding adapting IPv6 for IEEE 802.15.4, such as IP packet header compression, and solving the routing issues was suspended, delegated to a working group ROLL, created in 2008 for this purpose. ROLL produced a routing protocol denoted “Routing Protocol for Low-power lossy networks” (RPL) [9] in 2011.

While LOAD [8] development was suspended by the 6LoWPAN working group, pending the results from ROLL and experiences with RPL, reactive protocol derivatives live on: IEEE 802.11s [10] is based on the principles of Route Request/Route Reply exchanges for Route Discovery, and the ITU-T G3-PLC (Power Line Communication) standard [11], published in 2011, specifies the use of Kim et al. [8] at layer 2 or 2.5, for providing mesh-under routing for utility (electricity) metering networks. Justifications for using a reactive routing protocol in preference to RPL include that such protocols better supports bi-directional data flows such as a request/reply of a meter reading, as well as algorithmic and code complexity [12]. The emergence of LLNs thus triggered a renewed interest in reactive routing protocols for specific scenarios, resulting in work within the IETF [13] for the purpose of standardisation of a successor to LOAD – denoted LOADng (the Lightweight On-demand Ad hoc Distance-vector Routing Protocol – Next Generation). LOADng incorporates the experiences from deploying LOAD – including, but not only, in LLNs – and was, included in a subsequent revision of the G3-PLC ITU-T standard for communication in the “smart grid” [14].

Different routing protocols for LLNs have been proposed and standardised, including RPL [9] and LOADng [14]. While such protocols make different trade-offs and are of vastly different philosophies, they are united in the fact that when a link that has been actively used as part of a routing path fails, it is up to the routing protocol to recover by discovering alternative paths. Data flows are typically either buffered or dropped during this recovery. Dropping data flows while a routing protocol converges is “harmful”, since such traffic will have to be sent again (consuming energy, and creating additional traffic on the links in the network). Unfortunately, so is buffering data flows, as it imposes additional requirements on devices having sufficient memory to hold the buffers. A third alternative is opportunistic forwarding of traffic during route recovery, as proposed in DFF (Depth-First Forwarding in Unreliable Networks” [15]) – as a complement to an LLN routing protocol.

This paper presents, studies, and evaluates a “complete”, yet simple, adaptive, and modular approach to routing in LLNs. Using LOADng as the routing protocol core, this paper explores several extensions for adapting the protocol to different topologies, traffic characteristics, and other conditions: “Smart Route Request” and “Expanding Ring Search” are proposed to improve Route Discovery efficiency ; a “Collection Tree Protocol” is introduced to reduce the routing overhead for building a collection tree; integration of DFF, and extensions to DFF, are studied to allow a LOADng-routed network rapid recovery from data packet forwarding failures.

Preliminary results have been published in Yi et al. [16], Bas et al. [17], Yi et al. [18], Clausen et al. [19], exploring different extensions of LOADng. This paper further extends these results by presenting the protocol components as elements of a modular framework, considering the interoperability and exploring the security vulnerabilities. A generalised security framework for LOADng is also proposed.

The remainder of this paper is organised as follows: Section 2 presents the LOADng routing protocol, its operations, and other characteristics. Next, a set of extensions to the core protocol are presented. Section 3 studies a way of exploiting existing router state, for unicast route requests – with the goal to reduce the overhead of Route Discovery. Section 4 discusses the application of expanding ring search to the LOADng protocol to improve the Route Discovery efficiency by using neighbourhood routing information. Section 5 explores an extension to allow efficient construction of a collection tree for multipoint-to-point traffic, while introducing minimum routing overhead. Section 6 discusses the use of DFF in conjunction with LOADng in lossy network scenarios. These extensions present performance improvements possible and desirable in different scenarios – and are, also, both interoperable with each other, and with the “core” routing protocol: routers with and without these extensions can co-exist in the same network. For each of these extensions, their security characteristics are also evaluated. Section 7 evaluates the performance of different extensions, based on which their applicability is discussed. Section 8 introduces a security framework for LOADng – emphasising the necessary elements for protecting the integrity of the routing infrastructure of a LOADng-routed network. Finally, Section 9 concludes this paper.

Section snippets

LOADng – core protocol

A lightweight reactive distance-vector protocol, LOADng inherits the basic protocol operations of all reactive routing protocols: on-demand generation of Route Requests (RREQs) by a router (originator) for discovering a path to a destination, forwarding of such RREQs until they reach the destination router, generation of Route Replies (RREPs) upon receipt of an RREQ by the indicated destination, and unicast hop-by-hop forwarding of these RREPs towards the originator. If a path is detected

Efficient Route Discovery and Smart Route Request

Reducing the overhead, delay and complexity of the Route Discovery process (RREQ/RREP exchange) is a key to adapte on-demand routing protocols for use in constrained environments. As indicated in Section 2, some reactive routing protocols [4], [5] allow an intermediate router having a path to the destination sought in an RREQ, to respond by generating an “intermediate RREP” to the originator, and a “gratuitous RREP to the sought destination”.

This section discusses the rationale for LOADng not

Expanding ring for LOADng

Expanding Ring flooding is a technique aiming to limit the need for network-wide dissemination of RREQs. A router will at first send an RREQ with a reduced TTL (Time-To-Live) – causing the RREQ to not be flooded through the entire network, but only up to a limited distance. If the destination sought receives the RREQ, an RREP is generated and a network-wide flooding is avoided. For protocols allowing intermediate/gratuitous RREPs, if an intermediate router has a path to the sought destination,

Collection trees for LOADng

LOADng (extended, or not, with SmartRREQ and Expanding Ring) discovers paths between any (orginator,destination) pairs, for carrying point-to-point traffic. In some LLNs, another traffic pattern, called multipoint-to-point, prevails - where one or more devices act as data sink for all traffic – and where and all the other devices in the network communicate with the data sink. Discovering all these paths to the data sing individually may be inefficient, motivating a LOADng extension allowing

Depth-First Forwarding with LOADng

The second “L” in LLN means “lossy”, i.e., communication channels are of low capacity, time-varying and with high loss rates.

Routing protocols for LLNs, such as LOADng, are typically designed to limit the routing overhead imposed to networks as much as possible, and to be adapted to the varying nature of communication media. However, even once paths have been found, these paths may be unusable from time to time due to different reasons: presence of noise or interferences, low power supply in

Simulation and performance study

The performance of LOADng and different extensions described in previous sections, is evaluated by way of network simulations using NS2 (Network Simulator 2).

While network simulations are, at best, an approximation of real-world performance (particularly due to the fidelity of their lower layers to reality), they do provide a baseline for comparison and, generally, best-case results, i.e., real-world performance is expected to be no better than that which is obtained through simulations. The

Securing LOADng

As network devices and networks, emerge in increasingly less controlled environments with less “physical protection” of the infrastructure (e.g., limited access to a building where the network equipment is deployed), security requirements increase: in a wireless network, simply being within radio-range of a router may suffice to launch an attack – and sensor networks are deployed where there’s interesting data to sense, not where it’s easy to prevent physical access to the sensor devices.

Conclusion

This paper presents the protocol design and various optional extensions to, the “Lightweight On-Demand Ad Hoc protocol – Next Generation” (LOADng). A reactive routing protocol, LOADng is part of the ITU-T G3-PLC standard, and was designed with core principles of modularity, extensibility, as well as small footprint – and deployment-tuneable efficiency by way of interoperable extensions.

On the altar of “simple and compact”, LOADng has sacrificed several protocol functions, commonly found in

Acknowledgements

The authors would like to gratefully acknowledge the following people for their precious contributions during the specification and development of LOADng and its extensions: A. Colin de Verdiere, A. Bas (Ecole Polytechnique), C. Lavenu (EDF R&D), T. Lys (ERDF), Justin Dean (NRL, USA), A. Niktash (Maxim Integrated Products), Y. Igarashi, and H. Satoh (Hitachi YRL).

Thomas Clausen a graduate of Aalborg University, Denmark (M.Sc., PhD - civilingeniør, cand.polyt), Thomas Clausen has, since 2004 been on faculty at Ecole Polytechnique, France’s premiere technical and scientific university. At Ecole Polytechnique, Thomas leads the computer networking research group. He has developed, and coordinates, the computer networking curriculum, co-coordinates the Masters program in “Advanced Communication Networks” (ACN) and directs the Gradaute Degree Program

References (29)

  • J. Moy

    OSPF version 2

    (1998)
  • T. Clausen et al.

    Optimized link state routing protocol (OLSR)

    (2003)
  • T. Clausen et al.

    Optimized link state routing protocol version 2 (OLSRv2)

    (2014)
  • C. Perkins et al.

    Ad hoc on-demand distance vector (AODV) routing

    (2003)
  • D. Johnson et al.

    The dynamic source routing protocol (DSR) for mobile ad hoc networks for IPv4

    (2007)
  • R. Ogier et al.

    Topology broadcast based on reverse path forwarding (TBRPF)

    (2004)
  • T. Clausen et al.

    Comparative study of routing protocols for mobile ad-hoc networks

    Proceedings of the IFIP MedHocNet, September, Sardinia, Italy

    (2002)
  • K. Kim et al.

    6LoWPAN ad hoc on-demand distance vector routing

    (2007)
  • T. Winter et al.

    RPL: IPv6 routing protocol for low power and lossy networks

    (2012)
  • G. Hiertz et al.

    Principles of IEEE 802.11s

    Proceedings of WiMAN in conjunction with the 16th ICCCN

    (2007)
  • ITU-T G.9956: narrow-band OFDM power line communication transceivers - data link layer specification,...
  • T. Clausen et al.

    A critical evaluation of the IPv6 routing protocol for low power and lossy networks

    Proceedings of the 5th IEEE International Conference on Wireless & Mobile Computing, Networking & Communication (WiMob)

    (2011)
  • T. Clausen et al.

    The lightweight on-demand ad hoc distance-vector routing protocol – next generation (LOADng)

    (2013)
  • ITU

    ITU-T G.9903: narrow-band orthogonal frequency division multiplexing power line communication transceivers for G3-PLC networks: amendment 1

    (2013)
  • Cited by (0)

    Thomas Clausen a graduate of Aalborg University, Denmark (M.Sc., PhD - civilingeniør, cand.polyt), Thomas Clausen has, since 2004 been on faculty at Ecole Polytechnique, France’s premiere technical and scientific university. At Ecole Polytechnique, Thomas leads the computer networking research group. He has developed, and coordinates, the computer networking curriculum, co-coordinates the Masters program in “Advanced Communication Networks” (ACN) and directs the Gradaute Degree Program Connected Objects for a Digitized Society. Additionally, he coordinates the Cisco “Internet of Everything” academic chaire. Thomas has published more than 70 peer-reviewed academic publications (which have attracted more than10000 citations) and has authored and edited 19 IETF standards, has consulted for the development of IEEE 802.11s, and has contributed the routing portions of the recently ratified ITU-T G.9903 standard for G3-PLC networks upon which, e.g., the current SmartGrid & ConnectedEnergy initiatives are built. He serves on the scientific council of ThinkSmartGrids.

    Jiazi YI is research engineer in the network research group of Ecole Polytechnique, France. He got his Ph.D from University of Nantes, in 2010, a M.Sc. in Electronic System from Polytech’Nantes (France), and a M.Sc. in Computer Science, South China University of Technology. Jiazi’s scientific interests include mobile ad hoc networks, smart grid, routing protocols, sensor networks, long range and low power networks, QoS, etc. His PhD thesis focuses on the multi-path routing protocol for ad hoc networks, in which MP-OLSR was proposed. He has published above 30 peer-reviewed publications and co-authored several international standards. He is also one of the main contributors of LOADng routing protocol, which is the default routing mechanism used in G3-PLC for the future smart grid now being deployed in France and allover the world.

    Dr. Ulrich Herberg is a Senior Java Architect at Verifone, defining software architecture and leading a development team on cloud infrastructure for payment solutions. Before that, he served as Lead Architect at Panasonic, and Research Scientist at Fujitsu Laboratories of America, where he worked on smart grid related and IoT topics, in particular routing for mobile ad hoc networks, Demand Response, and network security. Dr. Herberg served as Working Group (WG) Chair of the 6lo WG and is secretary of the MANET WG at the IETF since 2013. He has published 12 Internet Standards (RFCs) at the IETF. Dr. Herberg is the principal editor of the OpenADR standard for automated Demand Response, and author of more than 25 peer-reviewed conference and journal papers, as well as 6 patents or patent applications. Dr. Herberg has a Ph.D. in Computer Science from Ecole Polytechnique (France), a “Diplom” (equivalent to MSc) in Computer Science from TU Munich (Germany) and an honors degree in Technology Management from CDTM (Germany).

    View full text