Tsado, Y., Gamage, K. A.A. , Lund, D. and Adebisi, B. (2016) Performance
Analysis of Variable Smart Grid Traffic Over Ad Hoc Wireless Mesh
Networks. In: International Conference on Smart Systems and Technologies
(SST), Osijek, Croatia, 12-14 Oct 2016, pp. 81-86. ISBN 9781509037209
(doi:10.1109/SST.2016.7765637)
This is the author’s final accepted version.
There may be differences between this version and the published version.
You are advised to consult the publisher’s version if you wish to cite from
it.
http://eprints.gla.ac.uk/146470/
Deposited on: 24 August 2017
Enlighten – Research publications by members of the University of Glasgow
http://eprints.gla.ac.uk33640
Performance Analysis of Variable Smart grid traffic
over ad hoc Wireless Mesh Networks
Yakubu Tsado
1
and Kelum A. A. Gamage
1
1
Department of Engineering
Lancaster University
Lancaster, UK
y.tsado1@lancaster.ac.uk; k.gamage@lancaster.ac.uk
David Lund
2
, Bamidele Adebisi
3
2
HW Communications Ltd Lancaster, UK,
3
School of Engineering, Manchester Metropolitan
University, Manchester, UK.
d.lund@hwcomms.com; b.adebisi@mmu.ac.uk
Abstract—Recent advances in ad hoc Wireless Mesh
Networks (WMN) has posited it as a strong candidate in Smart
Grid’s Neighbourhood Area Network (NAN) for Advanced
Metering Infrastructure (AMI). However, its abysmal capacity
and poor multi-hoping performance in harsh dynamic
environment will require an improvement to its protocol stacks
in order for it to effectively support the variable requirements of
application traffic in Smart Grid. This paper presents a
classification of Smart Grid traffics and examines the
performance of HWMP (which is the default routing protocol of
the IEEE 802.11s standard) with the Optimised Link State
Routing (OLSR) protocol in a NAN based ad hoc WMN. Results
from simulations in ns-3 show that HWMP does not outperform
OLSR. This indicates that cross layer modifications can be
developed in OLSR protocol to address the routing challenges in
a NAN based ad hoc WMN.
Keywords— HWMP; OLSR; Routing protocol; Traffic
Classification; Smart grid; QoS requirement; NAN.
I. INTRODUCTION
The new and advanced power grid, (also known as Smart
Grid), will extend monitoring and control on the electrical grid
by allowing a bi-directional flow of information and electricity
across the electrical grid network [1]. Amongst several
available communication technologies, the ad hoc Wireless
Mesh Network (WMN) has been acknowledged as a
communication technology well suited to the requirements of
Smart grid’s Neighbourhood Area Networks (NAN). This is
due to its extended coverage (through its multi-hopping
capabilities), low latency, high throughput and Quality of
Service (QoS) functionalities, which can enable data
transportation hop-by-hop from the traffic sources (i.e., Smart
meter on each household) to the backhaul distribution.
However, it is important to highlight that WMN technologies
were only developed to support multimedia applications such
as voice, video, web browsing and node mobility. In contrast,
Smart Grid’s application performance requirements are quite
different; they have strict transport and QoS requirements in
terms of latency, data rate and packet delivery such that it will
allow high reliability of critical functions (up to 99.9999 %
reliability which correspond to total outage period shorter that
one second per year) [2] [3]. Hence the need to undertake a
detailed performance analysis in order to investigate whether a
conventional ad hoc WMN is able to meet these requirements
when deployed in Smart Grid NAN. This will provide a good
understanding of the development areas in the design of an
efficient and reliable NAN based ad hoc WMN for AMI.
Performance of multi-hop ad hoc WMN is hinged on the
ability of the routing protocols choosing reliable paths to a
destination. Normally, paths are selected through the link
metrics used by the routing protocols to estimates the current
network conditions on each path. For this reason, a number of
reliable routing protocols such as: i) Routing Protocol for Low
Power and Lossy Networks (RPL) by Winter et al [4], ii)
Geographic routing, ii) Dynamic Source Routing (DSR)
protocol, and the Hybrid Wireless Mesh Protocol (HWMP);
have been classified for routing in NAN domain [5].
Nonetheless, modifications of these protocols and other routing
protocols are still being carried out to suit Smart Grid’s
application traffic characteristics. For example, performance
evaluation and reliability improvement of HWMP (IEEE
802.11s standard) was carried out for Smart grid in [6] and [7]
which resolves the original problems of HWMP. Given that
HWMP works at the MAC layer, it is worth exploring and
modifying other protocols that work at the network layer. This
is to enable the design of network layer and routing protocols
for smart meters with a network management perspective.
The Optimised Link State Routing protocol (OLSR) is an
established proactive routing protocol that works at the
Network layer. It mostly uses the Extended Transmission
Count (ETX) as its link metric and has been implemented on
several devices despite it deficiencies in certain areas. In order
to carry out modifications on OLSR for Smart grid data
characteristics, a performance analysis of the IEEE 802.11s
standard protocol (HWMP) and OLSR models on ns-3 network
simulator is carried out in a NAN based ad hoc WMN for
Advanced Metering Infrastructure (AMI). The key
contributions of this paper is to classify AMI application traffic
based on delay and reliability requirements as well as evaluate
the performance of each of the traffic class on a NAN based ad
hoc WMN using HWMP and OLSR protocols. The paper is
organized as follows: Section II presents AMI traffic
classification based on delay and reliability requirements.
Section III discusses the background of the routing protocols.
Simulation and performance evaluation of each traffic class is
presented in section IV, while section V highlights the
conclusion.
978-1-5090-3720-9/16/$31.00 ©2016 IEEE
II. CLASSIFICATION OF SMART GRID APPLICATION TRAFFIC
This section explores common traffic scenarios for AMI
applications and categorises them in to four application
classes. This is because it is important to study the traffic
supported by routing protocols, which can be periodic, real-
time or near real-time with strict reliability and low latency.
Reliability and low latency requirement can be a challenging
feat for ad hoc WMN, especially, when it is required for
variability of application traffics. In this section, a
classification of AMI traffics in terms of packet delivery
reliability and delay requirements across a network are
presented.
A. Traffic classification based on network driven
requirements
AMI application traffics can be characterized in terms of
data or network driven performance needs [8]. The packet
delivery performance in time and reliability domain used for
classification of Wireless Sensor Network (WSN) application
in [8] is adopted in this paper to classify Smart Grid AMI
traffic. Performance in time domain (delay sensitive) relates to
the time taken for data to be received at the destination and
reliability domain (loss sensitive) relates to how much data is
received at the destination. Latency requirement of a traffic
type such as delay is used to measure time domain
performance, while reliability domain performance is
dependent on how much data is required to be delivered. The
traffic classification are presented as follows:
Delay-tolerant, Loss-tolerant Class. The AMI
application traffic categorised in this class are those
that are not affected by high traffic delay and losses.
An example of these application types are best effort
traffic types such as periodic AMI data from Home
Area Network (HAN) devices, which are used to
monitor or estimate electricity usage in a household
(sent every 15 seconds and require a latency less than
3 seconds) [9] [10]. These applications can still
function as desired even if data losses are incurred
and/or data delivery time or latency is prolonged.
Delay-tolerant, Loss-intolerant Class. The AMI
applications in this class are those that must be
delivered at the destination but can tolerate delays in
delivering data [8]. Example of this application is the
Power quality data (sent every 3 seconds and has a
latency of less than 3 seconds). Power quality
information must be accurate for better load
estimation and determining the fitness of power to
consumer devices within seconds. High reliability at
the expense of delay must support by the
communication network.
Delay-sensitive, Loss-tolerant Class. Most SG traffic
require very high reliability, a certain amount of loss
rates may be acceptable in this class but data must
arrive in a timely manner (little percentage of Losses
tolerable) [11]. Examples of applications in this class
are Mobile Work Force tracking traffic, video
surveillance and software updates. Support for delay
is critical on this traffic.
Delay-sensitive, Loss-intolerant Class. The
application traffic in this class demand strict time and
reliability performance. Example of applications in
this class include Real Time Pricing (RTP), EV
charging traffic, Distribution Automation (DA), and
Wide Area Measurement (WAM) which involve
monitoring the distribution line and transformers. This
can also apply to event-triggered information
reporting an incident (fault) and/or information from
an actuator to carry out a particular task.
B. Traffic profile for simulation
The application traffic type in each class is modeled using
their expected packet sizes and delay objectives in User
Datagram Protocol (UDP) traffic profiles. Four different traffic
profiles sending variable packet sizes within short intervals
over several hops to the data concentrator as shown in Table 1
are used to represent each traffic class presented in the previous
sub-section. They include: 1) billing/AMI data information sent
every 15 seconds represents Delay-tolerant, Loss tolerant class;
2) power quality measurement sent every 0.5 seconds represent
the Delay-Tolerant, Loss-intolerant class; 3) Software update
sent every one second represents Delay-sensitive, Loss-tolerant
class; and 4) WAM data sent every 0.1 second represents
Delay-sensitive, Loss-intolerant class.
TABLE 1. TRAFFIC CHARACTERISTICS
Traffic Characteristics
Priority
Example
Delay Tolerant
Loss Tolerant
4
UDP IPv4 CBR 123 bytes/15s
Delay Tolerant
Loss Intolerant
3
UDP IPv4 CBR 3000 bytes/3s
Delay Sensitive
Loss Tolerant
2
UDP IPv4 CBR 1024 bytes/1s
Delay Sensitive
Loss Intolerant
1
UDP IPv4 CBR 48 bytes/0.1s
III. BACKGROUND ON OLSR AND HWMP
The category of routing protocols which require nodes
maintaining tables that represent the entire network (proactive)
are known to perform best in static networks. Therefore,
proactive protocols have been mostly proposed for routing in
NAN based WMN since smart meter networks are static nodes.
Performance evaluations of a number of routing protocols have
been carried out in NAN for AMI. The focus in this paper is
HWMP and OLSR, which work at the MAC layer and
Network layer respectively. Both protocols are discussed in the
following sub-sections:
A. Hybrid Wireless Mesh Protocol (HWMP)
The HWMP is the multihop default routing protocol for
IEEE 802.11s WLAN mesh networking. It was developed for
the purpose of allowing interoperability between devices from
different vendors; HWMP serves as a common path selection
protocol for every device that is compliant of IEEE 802.11s
standard. It uses the Air Link Metric (ALM) routing metric for
path selection to enable efficient routing in a dynamic network
environment [12]. The term hybrid is due to the fact that
HWMP allows On-demand (reactive) routing and tree-based
(proactive) routing to run simultaneously. In proactive tree-
based routing, a root node is configured in the mesh network.
A distance vector tree is built from the root node and
maintained for other nodes to avoid unnecessary routing
overhead for path discovery and recovery. In HWMP, when a
node needs a path to a given destination, it broadcasts a
route request message requesting a route to that
destination. The route request message is processed and
forwarded by all mesh points to the originator of the
route discovery. The destination node, or an intermediate
node that owns a path to the destination, answers with a
unicast reply message indicating the route requested. On
receiving this information, the forward path to the
destination from each mesh node is set up using the airtime
cost metric expressed in the equation below [7].
efr
Bt
OpOcCa
1
1
(1)
Where,
Oc
= channel access overhead,
Op
= MAC
protocol overhead
Bt
= size of the transmission frame,
r
=
data rate, and
ef
= error rate.
Due to the requirements of variable application traffic, a
number of improvements and modification have been carried
out in HWMP to support these applications in Smart Grid[6,
13]. They include: i) modifying the route selection mechanism
to reduce route fluctuations, ii) local route recovery mechanism
by using alternative routes, iii) calculation method of the air
cost metric that considers Smart grid’s data characteristics, and
iv) a mechanism to tackle the ARP broadcast storm problem in
802.11s-based NANs by piggybacking the MAC address
resolution in the proactive rote request of HWMP. However,
performance evaluation was carried out on the conventional
HWMP.
B. Optimised Link State Routing protocol
OLSR is an upgrade of the standard link state routing
algorithm for mobile ad hoc networks (MANETS) and it can
also be used for other wireless ad hoc and mesh networks. The
key concept in OLSR protocol is the use of selected nodes
known as Multi Point Relays (MPR) which reduces message
and routing overheads caused by the flooding of broadcast and
control messages in the network. There are several
documentation on OLSR protocol functions and operation
which can be found in [14] [15]. OLSR protocol is also metric
based routing that allows the calculation of link quality by
different link metrics. Several proposed link metrics and cross
layer metrics to improve routing and capture the best paths in
other to increase the performance of WMN have been
integrated with OLSR.
Performance evaluation was carried out on OLSR because
it is a well-known routing protocol for WMN that have been
implemented on several network simulation tools and
Commercial off the shelf Terminal (CoT) devices. This will
enable more research through experimental setup of NAN
scenarios as well as allow the integration and implementation
of modifications to suit application traffic on real test bed for
Smart Grid. The work carried out in this paper, considers
different load and flow rates for AMI application traffic
through simulation and examine the performance of the two
routing protocols.
IV. SIMULATION AND PERFROMANCE EVALUATION
The environmental parameters and mesh topology were set
for both protocols to allow a fair comparison. Each of the
application traffics specified in Table 2 were transmitted over
the network and results of the performance of both protocols
for varying grid sizes of NAN based ad hoc WMN are also
presented.
A. Simulation Setup
The experimental set up used in this study is similar to the
set up used in [16]. The simulation was carried out in ns-3 and
all the evaluation parameters were extracted using the flow
monitor module. A summary of the simulation parameters is
presented in Table 2. The choice of employing the transport
control protocol (TCP) or UDP is a trade-off between
efficiency (throughput and delay) and delivery guarantees with
flow control. Therefore, given that the transmission of metering
information is characterized by short transaction that do not
require persistent connection between the Smart meter nodes
and the data concentrator, it is more suitable to use UDP.
All smart meters on the network simultaneously transmit
their AMI information as a UDP Constant Bit Rate (CBR)
message to the data concentrator. The NAN topology is shown
in Fig. 1 and each of the application traffic profiles presented in
Table 1 were transmitted from all smart meter nodes to the data
concentrator. The grid size was varied from 4-by-4 (16 nodes)
to 11-by-11 (121 nodes) grid sizes. The smart meter nodes
were also configured with a single interface and the simulation
time equivalent of 1 day (86400 seconds) was used for each
grid size to give a representation of an AMI event for a day.
Data
Concentrator
Fig. 1. A 3 by 3 NAN based ad hoc Wireless Mesh Network
B. Performance metric
The metric that were used to assess the performance of
HWMP and OLSR in the network are:
1) Average end-to-end (ETE) delay: These metric
indicates the average ETE delay of each packet that is
successfully delivered to the data concentrator from a smart
meter.
(2)
2) Average Packet Delivery Ratio (PDR): percentage of
the average ratio of successfully received packets at the data
collector to the number of transmitted packets.
(3)
TABLE 2. NODE AND ROUTING PROTOCOL PARAMETERS
Parameters
HWMP
OLSR
Route metric
Air Link Metric (ALM)
Hop count/ETX
Simulation time (s)
86400
86400
Tx Range (meters)
120
120
Distance btw nodes (m)
100
100
Exponent
2.7
2.7
C. Results & Discussion
Fig. 2 depicts the mean and median PDR for smart meters
transmitting to a data concentrator for varying grid size of 16
to 121 nodes. The degradation in PDR as the grid network size
scales is as a result of increased interference, packet drops and
number of hops traversed by the packet. Fig. 2a to 2d also
show PDR performance of OLSR and HWMP for four
different SG applications traffic profile. From PDR results in
Fig. 2a, it is observed that after showing a higher PDR from
16 to 49 grid size, average PDR for HWMP degrades much
more rapidly than OLSR as the size of the grid increases for
the AMI data traffic profile. The steepness in the degradation
of HWMP can also be attributed to the large overheads
generated by HWMP as well as the PREQ travel distance from
the data concentrator (root node) as the network scales. This
demonstrates a clear indication of the advantage of OLSR’s
MPR in achieving better reliability across larger multi-hop
network [17] [18]. Fig. 2b, 2c, 2d, represent the PDR’s of
power quality measurement, software update and WAM
applications traffic profiles respectively. These applications
they have less transmission interval (higher packet generation
rate) between packets and the margin in PDR degradation
between the two protocols is narrowed down. Reliability in
WMN is impacted by both MAC layer factors and non-MAC
layer factors such as packet generation rate, packet sizes, hop
counts, traffic load and number of flows.
a) PDR for AMI data traffic
b) PDR for power quality measurement traffic.
c) PDR for Software updates traffic