This item was submitted to Loughborough’s Institutional Repository
(https://dspace.lboro.ac.uk/) by the author and is made available under the
following Creative Commons Licence conditions.
For the full text of this licence, please go to:
http://creativecommons.org/licenses/by-nc-nd/2.5/
Wireless Pers Commun manuscript No.
(will be inserted by the editor)
IPv6 Multicast Forwarding in RPL-Based Wireless Sensor
Networks
George Oikonomou · Iain Phillips ·
Theo Tryfonas
Received: date / Accepted: date
Abstract In wireless sensor deployments, network layer multicast can be used to
improve the bandwidth and energy efficiency for a variety of applications, such as
service discovery or network management. However, despite efforts to adopt IPv6
in networks of constrained devices, multicast has been somewhat overlooked. The
Multicast Forwarding Using Trickle (Trickle Multicast) internet draft is one of the
most noteworthy efforts. The specification of the IPv6 Routing Protocol for Low
power and Lossy Networks (RPL) also attempts to address the area but leaves many
questions unanswered. In this paper we highlight our concerns about both these
approaches. Subsequently, we present our alternative mechanism, called Stateless
Multicast RPL Forwarding algorithm (SMRF), which addresses the aforementioned
drawbacks. Having extended the TCP/IP engine of the Contiki embedded operating
system to support both Trickle Multicast (TM) and SMRF, we present an in-depth
comparison, backed by simulated evaluation as well as by experiments conducted on
a multi-hop hardware testbed. Results demonstrate that SMRF achieves significant
delay and energy efficiency improvements at the cost of a small increase in packet
loss. The outcome of our hardware experiments show that simulation results were re-
alistic. Lastly, we evaluate both algorithms in terms of code size and memory require-
ments, highlighting SMRF’s low implementation complexity. Both implementations
have been made available to the community for adoption.
George Oikonomou (
B
)
University of Bristol, Faculty of Engineering, Merchant Venturers Building, Woodland Road, Clifton, BS8
1UB, UK
E-mail: g.oikonomou@bristol.ac.uk
Previous address: Loughborough University, Computer Science, Loughborough, LE11 3TU, UK
Iain Phillips
Loughborough University, Computer Science, Loughborough, LE11 3TU, UK
E-mail: i.w.phillips@lboro.ac.uk
Theo Tryfonas
University of Bristol, Faculty of Engineering, Queen’s Building, University Walk, Clifton, BS8 1TR, UK
E-mail: theo.tryfonas@bristol.ac.uk
2 George Oikonomou et al.
Keywords 6LoWPAN · Wireless Sensor Networks · IPv6 Multicast · Trickle
1 Introduction and the Need for Multicast in Wireless Sensor Networks
Over the past years, the research community has invested considerable efforts to-
wards the seamless integration of wireless sensor networks (WSNs) with the Internet.
Previous work has demonstrated that pure IPv6-based WSN architectures are not only
viable but can also outperform application-centric designs [19]. Significant standard-
isation efforts have contributed to mature, interoperable implementations of embed-
ded IPv6 stacks, such as uIPv6 which is distributed as part of the Contiki embedded
Operating System. Among those standards are RFC 4944 [31] and RFC 6282 [20]. Pub-
lished by IETF’s 6LoWPAN work group, they discuss techniques for IPv6 datagram
fragmentation and header compression, in order to achieve their efficient transmis-
sion within IEEE 802.15.4 low power radio frames. For those networks (6LoWPANs),
the most widely adopted standard for routing is the “IPv6 Routing Protocol for Low-
Power and Lossy Networks” (RPL), which is specified in RFC 6550 [44].
The importance of network layer multicast forwarding in 6LoWPANs stems from
its ability to improve the efficiency of applications adopting a one-to-many communi-
cation paradigm. Examples of services which can benefit by multicast include service
discovery [2,21], network management and publish/subscribe schemes. For instance,
Extended Multicast DNS (xmDNS) builds on the mDNS specification [6] and expands
it “to site-local scope in order to support multi-hop LANs that forward multicast pack-
ets but do not provide a unicast DNS service” [29]. uBonjour [21] is a service discov-
ery scheme for resource-constrained devices in the wireless embedded internet. It is
essentially Bonjour’s lightweight variant and is based on multicast DNS (mDNS) and
DNS service discovery (DNS-SD) [5], which have recently had their message sizes
optimised for 6LoWPANs [22]. Additionally, the Contstrained Application Protocol
(COAP) is an emerging standard aiming at the integration of WSN devices with the
web. It operates over UDP (mandatory) or TCP (optionally) and has been designed so
that its messages can be easily translated to HTTP. It targets embedded devices with
severe memory and power supply restrictions. The current version of its specification
provides an extensive discussion on its operation over IPv6 multicast [40].
Previous research efforts in the area of multicast for WSNs focus on bespoke net-
work stack designs and do not investigate IPv6-specific challenges. The majority of
multicast forwarding algorithms encountered in current literature are based on geo-
graphic routing [39,38,3,14, 23, 42]. However, most of those approaches have certain
characteristics and make assumptions which render them unsuitable for IPv6-based
WSN deployments. For instance, many of them assume that, for every multicast mes-
sage, the sender is aware of the addresses or IDs of all intended destinations. Addi-
tionally, some efforts suffer from poor scalability while others rely on unrealistically
large network packets. Lastly, they are only applicable in situations where the source
as well as all destinations are within the WSN boundaries.
Despite the number of existing efforts, IPv6-based multicast has been somewhat
overlooked by the 6LoWPAN research community, as we discuss further in section 2.
The “Multicast Forwarding Using Trickle” internet draft (Trickle Multicast - TM) [17]
IPv6 Multicast Forwarding in RPL-Based Wireless Sensor Networks 3
discusses an algorithm which poses among the most suitable candidates. The RPL
RFC also briefly mentions multicast, but the discussion focuses on group management
without providing a sufficient level of detail in terms of forwarding.
The open issues outlined above have motivated us to design and implement SMRF,
a lightweight Stateless Multicast RPL Forwarding algorithm. In this paper, we dis-
close its design in depth and we highlight how it addresses current open issues while
maintaining high speed, energy efficiency and low complexity. Compared to geo-
graphic multicast algorithms, SMRF does not require geolocation information (nei-
ther explicit nor via a location service) and does not suffer from the aforementioned
scalability and datagram size issues. Moreover, by recommending a forwarding al-
gorithm, SMRF fills the gaps left open by the RPL RFC. In this context, this paper’s
contributions are the following:
– We compare the performance of SMRF against TM. For the evaluation, we con-
sider four metrics: i) Packet delivery ratio, ii) End-to-end delay, iii) Out-of-order
datagram delivery ratio and iv) Energy consumption. Evaluation is performed
through simulations as well as on a multi-hop hardware testbed. We also inves-
tigate the complexity of the two algorithms and we compare their code size and
memory requirements. Simulation and testbed results demonstrate that SMRF is
less complex and that it outperforms TM on three of the four metrics.
– Based on the outcome of the comparative evaluation, we present a host of criteria,
which can be used by network engineers and designers in order to select the more
suitable between SMRF and TM, depending on their deployment’s specific needs.
– We have extended Contiki’s TCP/IP stack to support both algorithms. Both im-
plementations have been released
1
to the community for adoption and further
scrutiny as a part of our port of the Contiki OS
2,3
[32].
This paper extends our previous work [33], providing the following additional contri-
butions: i) Extended design details for SMRF, ii) Evaluation of an additional metric:
the ratio of datagrams delivered out of order by TM (sec 5.4), iii) Additional simu-
lation experiments in a different topology for the evaluation of both algorithms on a
hop-by-hop basis (sections 5.2.1, 5.3.1 and 5.5.1), iv) Results from the evaluation of
both algorithms on a hardware testbed (sec. 6) and lastly v) Discussion on the code
size and memory requirements for both algorithms (sec. 7).
2 Related Work
2.1 Multicast in Traditional WSNs
Previous research efforts in the area of multicast for WSNs have primarily been fo-
cusing on traditional, application-centric network designs and as such do not ad-
dress IPv6-specific challenges. Multicast forwarding algorithms based on geographic
routing are dominant in existing bibliography and can be broadly classified as either
1
https://github.com/g-oikonomou/contiki-sensinode/tree/mcast-forward
2
https://github.com/g-oikonomou/contiki-sensinode/wiki
3
http://nets-www.lboro.ac.uk/george/contiki-sensinode/
4 George Oikonomou et al.
purely geographic [39, 38,3,14] or hybrid [23,42], whereby the geographic compo-
nent is complemented by features of other approaches, such as hierarchical routing.
The Geographic Multicast Routing (GMR) algorithm builds on existing unicast
geographic routing approaches. By adapting them, it aims to achieve multicast mes-
sage delivery to all intended destinations while maintaining minimum bandwidth
consumption. Nodes exchange position information with their neighbours through
periodic beacons. GMR is characterised by low computational complexity and a small
memory footprint [39]. According to subsequent works, GMR scales better than some
of its predecessors [38] but still suffers from scalability issues when dealing with
large deployments [23].
Using periodic beacons can have negative side-effects such as collisions and
increased energy consumption [38]. Based on this observation, BRUMA attempts
geographic multicasting without beacons, whereby neighbour positions are discov-
ered reactively. Next hop selection happens opportunistically through a mechanism
which only requires a low number of control messages. The authors demonstrate that
BRUMA is more efficient than GMR [38].
Carzaniga et al. propose a compact and completely decentralised multicast scheme
with asymptotically optimal network congestion properties [3]. It operates by build-
ing a multicast forwarding tree over an underlying geographic unicast routing ser-
vice, which allows nodes to send messages to a destination defined by a coordinate
pair (x, y).
Receiver-Based Multicast (RBMulticast) [14] is another purely geographic ap-
proach. Its principal novelty lies in the next-hop determination phase: potential next
hops contend for the channel based on their contribution towards delivering a packet
to its destination. Nodes offering the highest forward progress have higher probability
of getting selected as next hop. By adopting this approach, RBMulticast can operate
without routing tables and without maintaining a forwarding tree. To achieve this,
RBMulticast embeds the geographic location of all destinations in the packet header.
This raises questions regarding its scalability in large deployments.
The Hierarchical Geographic Multicast Routing (HGMR) [23] is a hybrid algo-
rithm combining the key design concepts of GMR [39] and the Hierarchical Ren-
dezvous Point Multicast (HRPM) protocol [8]. The resulting HGMR algorithm is fur-
ther optimised to be more energy efficient and scalable. HGMR divides multicast
groups into subgroups by using HRPM’s geolocation hashing. It takes advantage of
layer 2 reliability mechanisms by using HRPM’s unicast forwarding approach for
long, sparse paths and reverts to layer 2 broadcasting in areas of high density in order
to reduce the number of transmissions. When unicast forwarding is in use, HRPM (and
therefore HGMR) uses source routing along the branches of an overlay tree generated
by the traffic source. In this work the authors conducted a performance evaluation
of the three algorithms in a simulated IEEE 802.11 network. It is therefore difficult to
understand how the algorithms would behave under the frame size and bandwidth
limitations or the loss characteristics related to IEEE 802.15.4 networks.
The Multicast Routing with Branch Information Nodes (MR.BIN) [42] protocol is
a hybrid approach, combining geographic unicast routing with state-based multicast.
It maintains multicast states only on branch nodes of the forwarding tree. Commu-
nication between non-branching nodes takes place with geographic unicast. In order
