Information-Centric Networking for Machine-to-Machine Data Delivery
– A Case Study in Smart Grid Applications
Konstantinos V. Katsaros
1
, Wei Koong Chai
1
, Ning Wang
2
, George Pavlou
1
, Herman Bontius
3
and
Mario Paolone
4
1
University College London, UK
2
University of Surrey, UK
3
Liandon B.V., Netherlands
4
École Polytechnique Fédérale de Lausanne (EPFL), Switzerland
Abstract
Largely motivated by the proliferation of content-centric applications in the Internet, Information-
Centric Networking (ICN) has attracted the attention of the research community. By tailoring network
operations around named information objects instead of end hosts, ICN yields a series of desirable
features such as the spatiotemporal decoupling of communicating entities and the support of in-
network caching. In this article, we advocate the introduction of such ICN features in a new, fast
transforming communication domain i.e., Smart Grids. With the rapid introduction of multiple new
actors e.g., distributed (renewable) energy resources and electric vehicles, smart grids present a new
networking landscape where a diverse set of multi-party, machine-to-machine applications are
required to enhance the observability of the power grid, often in real-time, and on top of a diverse set
of communication infrastructures. Presenting a generic architectural framework, we show how ICN
can address the emerging smart grid communication challenges. Based on real power grid topologies
from a power distribution network in the Netherlands, we further employ simulations to both
demonstrate the feasibility of an ICN solution for the support of real-time smart grid applications and
further quantify the performance benefits brought by ICN against the current host centric paradigm.
Specifically, we show how ICN can support real-time state estimation in the medium voltage power
grid, where high volume of synchrophasor measurement data from distributed vantage points must be
delivered within a very stringent end-to-end delay constraint, while swiftly overcoming potential
power grid component failures.
1 Introduction
The rationale behind Information-Centric Networking (ICN) [1] is that information consumers are
mainly interested in the information itself rather than the explicit network location of the data/content
source (e.g., the host IP address). As such, the primary concerns of the network will no longer be on
the reachability between specific hosts but on the efficient information dissemination and retrieval.
Accordingly, the ICN design principle has put information/data at the center of the networking
landscape where information is published, resolved, delivered and stored natively based on names
rather than on explicit host locations. This in turn enables a series of desirable features such as the
support of in-network caching and multicast forwarding, as well as native security protection and
mobility support, thanks to the spatiotemporal decoupling of the communicating entities where data
producers and consumers are agnostic to where and when the data will be published/consumed by
their counterparts.
Given that the ICN paradigm has mostly catered for supporting content distribution operations in the
public Internet, the ICN concept has been arguably regarded as a key feature in the design of future
Internet architectures [1]. While it is still debatable whether this will become a reality, proposals have
also been made for applying ICN to alternative application domains such as machine-to-machine
(M2M) smart grid communications [2][3][4]. Similar to the current Internet, today’s power grid
communications are based on the host-centric model for data exchange between specific machines in
the centralized SCADA (Supervisory Control and Data Acquisition) environment. In this context, we
highlight the following challenges in current smart grid communication infrastructures:
• Decentralized large-scale data sharing. The SCADA system is expected to face distinct
challenges with increased participation of new stakeholders (e.g., solar/wind farm owners)
and active power prosumers
1
introduced into the grid operations. It will be very common that
data originated from one device is of interest to multiple entities participating in different
smart grid applications, including existing and future emerging services. Similarly, a single
entity involved in one or multiple applications in the grid may also need to access data
originated from a large number of devices. With the introduction of such 1-to-many, many-to-
1 or even many-to-many communications, the traditional host-centric model will suffer from
increasing complexity stemming from the explicit (usually pairwise) communication between
involved hosts.
• Heterogeneous requirements in distributing smart grid data. Emerging smart grid
applications present diverse requirements on quality of service (QoS) ranging from low data
rate, delay/disruption tolerant (e.g., smart metering and energy pricing) to higher data-rate
ones with stringent delay/disruption requirements (e.g., synchrophasor measurements).
Today’s communication infrastructure is merely a “bit carrier” with awareness of data
delivery requirements being expressed by (static) topological primitives (e.g., VPNs). The
lack of advanced features able to differentiate network behavior on an application/data level,
such as in-network storage and processing, has led to higher complexity/overhead on the
device side, as well as to the inefficient use of underlying network resources.
• Network security requirements. Based on the host-centric model, communication parties need
to know each other’s network location (i.e., IP address) to transmit data. Such exposure of IP
address in mission-critical power grid applications may introduce vulnerability to network
intrusions and denial-of-service (DoS) attacks [4].
In this article, we advocate the introduction of ICN in smart grid applications, given that such
applications are mainly interested in just what – in terms of the power grid data, rather than where i.e.,
the specific network address of the data source. Taking grid measurement applications as a typical
example, though the identity of measurement points is of importance, this is usually indicated by a
standardized identifier code that uniquely identifies the data stream and the physical device but not its
network IP address. This information forms part of the payload (i.e., the what) for the communication
network, regardless the applied networking paradigm. Moreover, in many popular smart grid
applications (e.g., power consumption measurements for the support of demand response), data may
be required in the form of aggregates (e.g., consumption over a particular power grid area), hence
further decoupling the delivered information from the exact location of its origin in the network.
In this context, the introduction of ICN, including its inherent publish/subscribe communication
primitives, enables a higher degree of flexibility in supporting data sharing and smart grid control. In
the long term, an ICN-based approach is expected to facilitate the support of complex and evolving
data delivery structures (e.g., many-to-many communications for data sharing) owing to the
introduction of new applications/devices to the grid, since, in contrast to the current host-centric
model, it does not focus on the establishment of explicit communication sessions. This is expected to
significantly reduce system (re-)configuration complexity with the evolvement of new services. It
further applies in shorter time scales as well i.e., a change in the grid topology due to fault in a line or
maintenance that requires an asset-change will not affect the ongoing data delivery operations. For
instance, a new data consumer is only required to subscribe to the data of interest and the rest of the
communication will be automatically setup. Such automation minimizes manual reconfigurations and
cuts down on possible human errors. Another example is the vision of on-demand islanding
operations by Distribution Network Operators (DNOs) that require dedicated monitoring and control
infrastructures. With ICN, the data can be cached for later retrieval while the island is formed and all
flows redirected to the new subscribers once the islanding maneuver is completed. Data caching,
possibly with local processing, as facilitated by a data/information-aware network, can also contribute
to improving efficiency and facilitating QoS support. For instance, network nodes may adapt rates of
measurement data targeting at different grid operations, rather than having such functions at the end
devices. In addition, upon anomaly events (e.g., failures), faster response times can be achieved by
allowing affected devices to locally fetch recovery instruction/data that is actively cached at nearby
network entities, or by quickly diverting affected data flows at intermediate nodes instead of
1
Entities acting both as producers and consumers of energy.
reconfiguring the data producers themselves. As far as security is concerned, ICN offers intrinsic
support for cyber-security in the power control system, as the identity along with the network and
physical location of machines can be encrypted as part of the payload and are therefore not exposed
[4]. In this case security thus becomes an integral part of the underlying network infrastructure rather
than an a posteriori patch.
In this paper we present an overlay ICN-based communication framework for supporting M2M-
oriented smart grid applications, based on publish/subscribe operations and the notion of location
independent topics. We shed light on the challenges faced in this emerging networking environment
and elaborate on how these can be addressed by ICN communication primitives. Besides qualitative
advantages, we also quantitatively illustrate the benefit of the proposed framework, focusing on the
use case of Phasor Measurement Units (PMUs)-based real time state estimation (RTSE) in medium-
voltage (MV) power distribution networks. Accordingly, our simulations are based on two real
European power grid topologies and our results show that (i) with careful planning and provisioning
of network resources, ICN can successfully support the requirements of such mission-critical
communications, (ii) ICN communication primitives can substantially reduce the complexity of re-
configuration operations in cases of power grid component failures.
2 Machine-to-machine Smart Grid Communication
The evolution of power distribution networks towards the so-called Active Distribution Networks
(ADNs), shown in Figure 1, requires the availability of suitable Energy Management Systems
(EMSs), to achieve specific operation objectives [5], such as:
• Optimal voltage / line-congestion controls;
• Fault detection and location;
• Post-fault management;
• Local load balance;
• Network losses minimization.
Figure 1: Current point-to-point data delivery in smart grid
PDC$B$
PDC$A$
Generator$Governor$
PMU$2$
PMU$1$
PMU$4$
APG$
SCADA$
Ac6ve$power$
generator$
(APG)$
Collocated$at$
common$feeder$
PMU$3$
Data$
These operations are significantly improved if the system state is known. In RTSE, large volumes of
raw synchrophasor measurement data are collected by geographically distributed PMUs, strategically
deployed in the power grid infrastructure to ensure full grid state observability. The UTC-
synchronized data is continuously streamed to phasor data concentrators (PDCs). PDCs collect
synchrophasor data and other quantities (i.e., synchrophasor frequency, rate of change of frequency,
powers, etc.) measured by PMUs and transmit them to other relevant applications.
With this specific real-time monitoring approach, PDCs located at different substations periodically
report state estimation to the central SCADA entity at a lower rate than RTSE. In addition, power
protection relays (that can be collocated with PDC; hence not shown in the figure) are also interested
in such data to detect and react to potential anomalies ensuring the seamless operation of the power
grid. Such anomalies (e.g., cable failures) may result in changes of the power grid topology with the
purpose of restoring operation and avoiding cascading effects (i.e., by opening/closing circuit
breakers), resulting at the same time in changes for the monitoring data flows as well i.e., a PMU
device may need to direct its data to a different PDC. For these types of data receivers, there are
stringent requirements on both data frequency (50 synchrophasor measurements/sec) and end-to-end
data latency (maximum 20msecs).
In the case of renewable power generation units, the synchrophasor data may also be fed to local
active power generators (APGs) for them to adjust power generation operations. The data frequency
required for the APGs is much lower (e.g., 2 measurements/sec) with no strict end-to-end latency
requirement. Meanwhile, individual APGs also transmit information about their locally generated
renewable power to the generation governor.
From Figure 1, it is evident that the current power grid communication is suffering from several
critical deficiencies. Since it is still based on the current host-to-host model, dedicated point-to-point
communication sessions need to be maintained, and this requires complex per machine
configurations. At the same time, network resources may be wasted when different recipients request
different rates of the same data flows in separate communication sessions. Complexity and the
associated power grid control overheads further increase when anomalies call for rapid re-
configurations of data flows i.e., diverting traffic of affected PMUs to alternative PDCs requires each
PMU device to be re-configured individually. Finally, the exposure of IP addresses of individual
mission-critical entities also makes them vulnerable to DoS attacks.
3 ICN Framework for Smart Grids
Based on the aforementioned communication requirements in smart grids, we identify a favorable
match with some major ICN design aspects. We exploit this match and apply ICN as an overlay for
enabling resilient and seamless communication in smart grids. The decoupling of information from
location and time fits the communication patterns of the considered applications, yielding
opportunities for a simplified and efficient management of communication flows. Based on the
inherently supported pub/sub communication primitives, ICN introduces a degree of indirection
between the communicating end-hosts by enabling the network to actively mediate information
delivery similarly to [9]. Apart from previously investigated security related benefits [4], this
practically translates to the ability to:
• Simplify both the establishment and re-configuration of communication flows in the
aforementioned multi-party communications (e.g., delivering synchrophasor measurements
from PMUs at different feeders to PDCs and APGs), including the introduction of new
devices interested in the data produced by legacy elements.
• Facilitate multi-criteria traffic management decisions i.e., selecting one or more indirection
points based on the underlying transmission capabilities, application requirements, network
conditions, topology characteristics, etc. (e.g., selecting only delay sensitive data to forward
on high data rate links).
• Enable in-network management of smart grid data, including caching and processing such as
rate adaptation, aggregation, filtering, etc. (e.g., enabling the rate adaptation of PMU
measurements at an indirection point close to an APG).
• Enhance resilience of information delivery to protect the grid against anomalies/power
failures and subsequently minimize power distribution disruption.
• Enhance security by avoiding the exposure of critical components network locations through
the means of indirection.
In these cases, the information-centrism of the network, expressed by the fundamental role of topics
and their attributes, enables network operations to take place on an information level, bridging the gap
between the application requirements and the underlying technological and topological
characteristics. Concurrently, enabling these ICN features in an overlay fashion facilitates adoption
and deployment of the ICN principles, especially in considerably heterogeneous environment where
smart grid networks are often based on a set of diverse communication technologies.
In the following, we describe an instantiation of a smart grid communication platform based on the
aforementioned ICN concepts that can support heterogeneous smart grid applications, including ones
with stringent real-time requirements. We use common ICN building blocks and primitives and
describe the specific design requirements and challenges of M2M and mission-critical applications
described above.
3.1 Building an ICN-enabled Smart Grid Communication Platform
Figure 2 presents a logical illustration of a topic-based ICN smart grid infrastructure to support data
dissemination across heterogeneous entities. We follow the pub/sub paradigm that is inherent in ICN
schemes for supporting communication between the smart grid entities. For instance, in one form or
another, various ICN projects (e.g., PSIRP, PURSUIT and COMET) employ similar pub/sub
mechanisms to harness the benefits of ICN in terms of flexibility in communication and the added
security features. In our proposal, communication is organized in location-independent topics that
uniquely identify semantically related data. Each topic is associated with a set of attributes such as
spatiotemporal information and reporting rate (where applicable). A topic resolution system handles
data publication and subscription for the interested receivers to access data published to a topic but
without directly contacting the publishers. In this system, which is conceptually similar to resolution
systems in existing ICN schemes [1][8][9][10], we follow the separation of control and data plane
based on homogeneous entities. Specifically, for each topic, the resolution function can be located at a
different node to the one that is responsible for the forwarding function, and each node can flexibly be
responsible for resolution and data forwarding of different topics. In addition to plain forwarding on
the data plane, we further build on the information-centric primitives of the architecture to enable in-
network processing of data subject to the topic they belong to, as well as their attributes. This may
include aggregation of data being disseminated in multi-source based topics, multi-criteria filtering
(e.g., location, time) and rate adaptation according to heterogeneous receivers demanding different
data reporting frequencies [11].
