2015 IEEE International Conference on Data Science and Data Intensive Systems

Internet of Things for Industrial Automation – Challenges and Technical

Solutions

Hongyu Pei Breivold Kristian Sandström

ABB Corporate Research ABB Corporate Research
[email protected] [email protected]

Abstract

Using internet of things (IoT) to connect things, features/services that increase the capabilities of
service, and people for intelligent operations has been existing products to create new business opportunities.
discussed and deployed in many industry domains such Accordingly, many organizations start to evaluate IoT
as smart city, smart energy, healthcare, food and water as a potential value creator in their business strategies
tracking, logistics and retail, and transportation. [5], through the combination of advanced data analytics
However, scarce information is available for IoT usage and internet of things to connect machines, computers
in industrial automation domain for reliable and and people to enable intelligent industrial operations in
collaborative automation with respect to e.g., enabling many industries, such as industrial domain, smart city
scalable collaboration between heterogeneous devices domain, and health well-being domain [6].
and systems, offering predictable and fault-tolerant Industrial automation is a domain in which Internet-
real-time closed-loop control, and inclusion of of-things can bring with many benefits. For instance,
intelligent service features from edge devices to the remote location of plants would benefit from
cloud. In this paper, we will clarify the specific quality technologies that support remote operation and
attribute constraints within industrial automation, maintenance; autonomous collaboration between
present specific industrial IoT challenges due to these devices so that devices are aware of each other for
constraints, and discuss the potentials of utilizing some information exchange, in this way to reduce engineering
technical solutions to cope with these challenges. costs in terms of manual configurations of all involved
devices; the real-time data collected from a large
Keywords: Internet-of-things, Industrial automation, number of these interconnected physical hardware units
Challenges can be used for developing new intelligent applications,
etc. However, building new intelligent services implies
1. Introduction many challenges, including both general IoT challenges
The Internet of Things (IoT) [1] is “a self-configuring [7] such as security, and industrial automation domain-
and adaptive system consisting of networks of sensors specific requirements such as reliability and availability
and smart objects whose purpose is to interconnect all that remain to be fulfilled in the new IoT context. The
things, including every day and industrial objects, in main objective of this paper is therefore to identify these
such a way as to make them intelligent, programmable challenges, within industrial automation domain in
and more capable of interacting with humans”. The particular, and discuss how some potential technologies
Internet of things extends the cloud computing concept can be deployed to cope with these challenges.
[2] beyond computing and communication to include The remainder of this paper is structured as follows.
everything, i.e., also the physical devices [3]. Industrial Section 2 gives a brief overview of research initiatives
internet [4] uses sensors, software, machine-to-machine and research projects within IoT as well as the status of
collaboration and various technologies to gather and IoT usage in industrial automation. Section 3 presents
analyze data from physical and virtual world for firstly general IoT challenges, specific quality attribute
optimized operations and providing services. In the constraints within industrial automation, and discuss
meanwhile, trends such as fast-changing and then specific industrial IoT challenges due to these
unpredictable market needs and customer requirements, constraints. Section 4 presents briefly some potential
the customer expectation of responsiveness, and the technical solutions that cope with the identified
connectivity capability becoming available to embedded industrial IoT challenges. Finally, Section 5 concludes
systems are triggering many companies to move from the paper.
products to services, and continuously deploy new

978-1-5090-0214-6/15 $31.00 © 2015 IEEE 532

DOI 10.1109/DSDIS.2015.11
2. Internet-of-Things for Industry predictable inclusion of service features vertically from
edge devices to the cloud, as shown in Figure 1.
Many organizations start to see the potential
opportunities of the Internet-of-things and its impacts on
providing solutions that could offer operational
advantages [5]. In line with this, there are several
research initiatives on IoT, such as IERC – European
Research Cluster on the Internet of Things [8], Industrial
Internet Consortium [9], Industri 4.0 [10], and the
creation of the Alliance for Internet of Things
Innovation (AIOTI) by the European Commission [11],
which initiates the development and future deployment
of the IoT technology in Europe. There has been also a Figure 1. Industrial IoT from Edge to Cloud
number of EU-funded research activities in IoT,
covering various aspects such as communication,
3. IoT Challenges in Industrial Automation
hardware technology, identification and network
discovery, security, interoperability, standardization, Consumer IoT and Industrial IoT have similarity in
etc. A few recent examples include IoT@Work [12] many senses, though there are key differences that are
which aimed to develop an IoT-based plug and work reflected in specific real-time and deterministic
concept centered on industrial automation using FIAT requirements of industrial IoT applications as well as
Research Center facilities. The challenge in focus was severity of consequences if failure occurs in the
configuring a highly complex and demanding industrial IoT applications where flawless operation is
communication network as well as an IT security expected due to the huge capital of investment [17]. In
subsystem. The technical approaches used included this section, we will firstly present general IoT
auto-configuration of real-time Ethernet, event challenges. Then we will describe the main quality
dispatching, and capability-based access control. IoT-A attribute constraints within industrial automation, and
[13] focused on creating an architecture reference model discuss specific industrial IoT challenges due to these
for the interoperability of IoT systems and an constraints.
infrastructure for scalable look up and discovery of IoT
3.1. General IoT Challenges
resources and entities. Another example is BUTLER
[14], which focused on developing secure and smart life Data and service security Large-scale applications and
assistant applications using context and location aware services based on IoT are increasingly vulnerable to
and pervasive information system. Another IoT project disruption from attack or information theft [7] because
example is VITAL [15], which aims to develop next having more devices, systems, and technologies
generation smart city applications with integrated IoT connected leads to more decentralized entry points for
services. these security disruptions. As unavailability of service
or data can have severe impact on the customer
The domains and use cases for which most potential
businesses, it is important to provide security
IoT intelligent solutions are being planned or
mechanisms to increase the level of data protection and
implemented cover smart cities, smart energy and smart
security for messages communication between devices
grid, healthcare, food and water tracking, logistics and
as well as from devices (e.g., sensors, actuators, etc.) to
retail, transportation [6, 7, 16]. Within industrial
the cloud platform [7] to ensure service continuity and
domain, most IoT applications address logistic and
required Quality-of-Service (QoS).
product lifetime management, agriculture and breeding,
and industrial processes, in which assets diagnostic is a Trust, data integrity and information privacy The
key application [6]. These data analytics and interconnected devices and the users of the system need
computation applications typically feed data to servers to have confidence that the information and services
for subsequent offline analysis, and do not provide being exchanged can be relied upon [7]. Therefore, trust
direct feedback in real-time back into the actuators of mechanisms need to be able to deal with humans and
devices in a closed-loop control. Moreover, the machines to ensure trustworthy access to data and
application of IoT technologies for industrial proper authorization of service. Moreover, it is essential
automation domain is rarely addressed. Therefore, with assured, proper, and consistent collection,
flexible and cooperative automation needs a holistic processing, communication, use and disposition of
industrial IoT solution that addresses business, sensitive information and personally identifiable
technologies, and architectures aspects in order to (i) information [18] throughout the whole chain from
provide real-time closed-loop control for optimization devices, the edge of the network, and to the cloud. This
and improved uptime; (ii) enable horizontal includes protecting user resources and data from
collaboration between devices; and (iii) offer secure and unauthorized access that may compromise their integrity

533
and confidentiality [19]. Building a trusted and fault- Timing and determinism Industrial control systems are
tolerant system in the IoT context requires taking into real-time systems that have stringent requirements on
account how to protect communications, how to ensure the system's temporal behavior, accuracy, and response.
integrity and confidentiality of data in the system, and Latency and variations in the time an operation takes to
how to manage authentication and access control to perform is of major concern besides average throughput.
resources in systems that consist of thousands of A concrete example is a control system for industrial
devices. robots in the context of welding, where sensors need to
detect the position and direction of seams, the signals
Scalability This challenge is reflected in different
from sensors have to be processed quickly to enable the
aspects [16, 20], including: (i) naming and addressing –
robot controller to change the direction of motion in
the scalability of the device address of the existing
time. The accuracy requirement for the industrial robots
network must be sustainable; (ii) data communication
are related to the timing requirements for the services
and networking – the connection of new networks and
involved, e.g., sampling a measurement at precise and
devices should not jeopardize the performance of
deterministic intervals. For these industrial control
existing networks, devices, and data transmission
systems, there are also timing interactions involved in
despite the high level of interconnection among a large
different processes that are executed in the system.
number of devices and system components; and (iii)
Consequently, an industrial control systems needs to
service provisioning and management – due to the
achieve a level of determinism to ensure its proper
massive number of services/service execution options
operation.
available and the need to handle heterogeneous
resources. Reliability and availability The main indicators for
reliability and availability include (i) Mean Time
Interoperability Industry is dominated by proprietary
Between Failure (MTBF), which is usually a statistical
interfaces and solutions. The amount of devices and
representation of the likelihood of a component, device,
system components from different vendors and different
or system to fail; (ii) Mean Time To Repair (MTTR),
domains poses challenges in terms of multiple
which is the average time to repair a system, or
platforms, numerous protocols and large numbers of
component, that has failed; and (iii) Probability to Fail
APIs. According to [21], the vast variety of devices,
on Demand (PFD), which is a statistical measurement of
applications, and implementations within the industrial
how likely it is that a process, system, or device will be
IoT will result in a massively heterogeneous set of data
operating and ready to serve the function for which it is
with variation in format and interpretation of data,
intended. In industrial automation, production losses
quality, frequency, and timing of the data. Therefore,
due to unplanned stops caused by e.g., a failure or
interoperability is an important aspect that needs to be
component deterioration are unacceptable. A common
considered in IoT solutions so that diverse devices and
practice to avoid these unplanned stops and potential
systems can share information and interact with each
failures is continuously monitoring devices and services
other.
in the event of condition changes and notifying the
3.2. Automation Domain-Specific Constraints system. In this context, industrial communication plays
an important role in industrial measurement and control,
Industrial automation spans over many different
and particularly in process automation due to the trend
types of control systems, e.g., motion control, protection
of decentralization of devices and control systems, as
systems, and digital control. Furthermore, industrial
well as different layers of communication, i.e., between
automation can be found in many different domains that
field devices and controllers, between controllers and
execute different processes, i.e., (i) batch process in
operation stations, and between operator stations and
which the product is created stage by stage over a series
enterprise management services. A challenge to keeping
of workstations, e.g., pharmaceutical, food and
reliability and availability is therefore concerned with
beverages industries; (ii) continuous process in which
guaranteeing the level of quality in all communications
the materials are processed continuously without
when delivering data and services.
interruption, e.g., electric power, oil and gas, mining and
metals, pulp and paper industries; and (iii) discrete Safety and criticality From a functional safety
manufacturing, which is characterized by separate unit perspective, it is difficult to use software technologies
production and requires manufacturing floor space to (such as operating systems) that are developed for
work off orders to build something, e.g., cars and general purpose in a safety context, e.g., protection
industrial robots. Although specific functional functionality that puts the system into a safe state in case
requirements for automation systems may vary for of an error. When integrating different software
various domains, there are similarities on the complexity applications on a common hardware, safety standards
that control systems are physically and logically such as IEC 61508 [22] usually requires that in devices
structured. They also share many requirements related with a mix of software modules of different criticality,
to non-functional properties.

534
all software modules are treated as having the highest on industrial performance predictability, e.g., in terms of
criticality unless strict independence can be shown. timing, jitter, etc.
Interoperability A typical industrial process control Latency With hundreds of thousands of sensors,
system consists of thousands of components including a actuators, programmable logic controllers connected, it
network of many workstations, servers, controllers, and is important to have reliable and deterministic data
up to several hundred thousands of individual I/O transfer and analysis in real time. Due to the use of
signals from various devices. There is thus a huge internet connections and wireless connections for
challenge in managing the large diversity of components industrial IoT solutions, the assumption about the
from different manufacturers. Since the components of predictability of the network latencies of industrial
the system are designed to be integrated into many control system in Local Area Network (LAN)
different types of systems and in different component environment has changed. According to [25], the
constellations, the amount of possible configurations is internet roundtrip delay ranges from tens to a few
large. Therefore, in industrial automation domain, the hundreds of milliseconds. For industrial automation
scale and structure complexity of control systems, the applications in a latency-sensitive environment, how to
heterogeneous issues due to the multitude and diversity distribute the real-time services in the network is
of integrated components make integration and affected by several parameters such as optimal
engineering efforts a costly operation to achieve efficiency, bandwidth availability, location of data, and
interoperability. These challenges of integration have overhead. Due to the required quality attributes
resulted in efforts to standardize for interoperability in constraints in industrial control applications with respect
the automation domain. One notable example of to e.g., timeliness and reliability, a challenge is how to
interoperability standardization is the OPC unified dynamically deploy data and service intelligence at
architecture (OPC-UA) [23], which addresses different levels from the network edge to the cloud to
interoperability between components throughout the achieve optimal performance.
automation hierarchy.
Fault tolerance This is a key consideration in industrial
3.3. Industrial IoT Challenges IoT where many thousands of heterogeneous industrial
devices are connected and exchange information. To
The specific automation domain constraints
ensure industrial constraints on availability and
explained in the previous section imply specific
reliability, it is important to consider how to design
challenges that we need to cope with when deploying
fault-tolerant systems to be adaptive to unforeseen
industrial IoT solutions:
events like failure of underlying services, malicious
Mixed criticality A lot of functions are integrated in an attacks, or changes in the quality of particular services,
industrial automation system. These functions are of as well as faults that are related to internet infrastructure
different criticalities dealing with controller compute and cloud software technologies when providing real-
logics, real-time and non-real-time functionality, and time services. From business perspective, the impact of
security. As availability and reliability are given more an industrial IoT solution is huge, considering their
and more priority in many products, systems, and value propositions with respect to the efficiency of
services, these functions of different criticalities need to industrial plant operation and maintenance as well as
be separated so that functions of low criticalities would potential values brought by disruptive business models.
not interfere with a high-criticality function. The Therefore, it becomes essential to design and deploy
availability of multicore processor hardware may resilient IoT solutions with fault-tolerance and recovery
provide the necessary independence between software mechanisms in order to be able to tolerate and quickly
applications [24]. However, most legacy software is recover from accidental and malicious failures that may
written for being executed on a single core. Therefore, lead to extended service unavailability.
the transition to being able to mix operating systems and
Scalability with respect to data update cycles Many
strictly separate different software partitions becomes
industrial plants could have tens of thousands of control
necessary. This transition will provide an opportunity to
loops and applications to maintain the desired
manage both legacy software running on a traditional
performance in operation [25]. These different industrial
real-time operating systems and new IoT-related
control applications have different cycle-time
functionality of different criticality running on general-
requirements. For motion control applications and fast
purpose operating system. A challenge when making the
control loop applications, the cycle-time requirement
transition to multicore is to prepare the software for both
could reach 1ms and 10ms, respectively. For high speed
imminent and expected future requirements, for
I/O applications, the cycle time required to perform such
instance, basic capabilities to support software
functionality reaches 100ȝs. These high-frequency data
partitioning such that services of mixed criticality can be
sets are useful for studying dynamic behaviour analysis
combined in one device while still keeping constraints
so as to create dynamic models for optimization and

535
monitoring. This implies huge amount of data that need related software therefore makes it challenging or
to be managed through e.g., filtering, pre-processing at infeasible to certify systems that uses general purpose
or close to edge devices before sending to the cloud. technologies lacking proper documentation from the
development process or to depend on remote
Scalable collaboration As shown in Figure 2, for
functionality of which reliability is not possible to
industrial distributed control systems, controllers are
assess. As products and systems are becoming
connected through an Ethernet-based control network.
interconnected, the need for compliance to safety
These controllers may share control data and also
standards to ensure proper functional safety services is
interact with nodes higher in the network hierarchy, e.g.,
paramount.
operator stations. The data connections are usually in
vertical silos, and have to be explicitly engineered, Industry-specific security challenge Industrial control
which do not scale as they are usually point-to-point systems often assume that the network in which they
connections. Today a lot of engineering effort goes into operate are (physically) secure and separated from
achieving interoperability among these heterogeneous outside intervention. However, the introduction of IoT
devices and applications due to the diversity of used solutions that connect industrial machine sensors,
network protocols and communication technologies. actuators, and industrial networks poses a challenge that
Therefore, a challenge here is to enable both horizontal the data collected from these industrial devices and
and vertical systems, platforms and devices to be able to systems need to be stored in a secure and proper manner.
communicate and collaborate. Countermeasures to the constant development of new
attack vectors that lead to numerous security issues are
required. Therefore, the software architecture of an IoT
solution needs to protect the interconnected devices
from intrusions and interference from new attack
vectors (i.e., coming from the communication channels)
from entering the system so as to ensure secure
operations. In addition, in some industrial automation
processes, an outage in milliseconds can be very
harmful. Thus, for continuous operation, a security
challenge of an IoT solution is how to allow for seamless
security upgrades without impairing functional safety or
interference to the provided services in a control
process.
Legacy long-lived industrial systems The software
systems that we have worked with within industry are
Figure 2. Industrial Automation Control Hierarchy often legacy systems that represent valuable software
assets. These industrial systems usually have a lifetime
Functional safety Functional safety is critical for the
of 10-30 years, and are continuously subject to a
overall safety of a system or product that depends on the
substantial amount of evolutionary changes, e.g.,
correct execution of specific commands and functions.
software technology changes and ever-changing
The oil and gas industry, nuclear plants, etc. all rely on
business decisions driven by market situations.
functional safety that offers the necessary risk reduction
Moreover, these systems most likely have gone through
required to achieve safety for the equipment. Safety
many turnovers of the original developers, and thus have
function and safety integrity are two types of
the typical characteristics of legacy systems as described
requirements to achieve functional safety [22]. The
in [26], e.g., increasing complexity, poor
safety concern is reflected in both hardware and
documentation, and lack of understanding by the current
software within process industries, which demand close
developers. In the meanwhile, trends such as fast-
integration of safety and control applications such as
changing and unpredictable market needs and customer
emergency shutdown systems, fire and gas systems,
requirements, the customer expectation of
which are Safety Integrity Level (SIL) compliant
responsiveness, and the connectivity capability
application solutions that prevent inadvertent/accidental
becoming available to embedded systems are triggering
control actions and provide added reliability and uptime
many companies to move from products to services, and
for the concerned processes.
continuously deploy new features/services that increase
Functional safety put very specific requirements on both the capabilities of existing products to create new
how the systems are developed and on the technology business opportunities. Therefore, there is an implied
itself. One such requirement is to provide traceability challenge to retrofit these legacy systems with end-to-
from requirements through design, implementation, end IoT solutions that are deployed for additional
verification, and validation. The requirements for safety operation services.

536
4. Managing Industrial IoT Challenges systems, with several commercially available products
[33]. In virtualization technology a virtual machine
In this section, we will briefly present some potential monitor or hypervisor manages how physical resources,
technical solutions that can help cope with the industrial like RAM memory, CPU, and network interfaces, are
IoT challenges that we described earlier. shared among a number of virtual machines (i.e.,
4.1. Managing Latency and Scalability of Data software partitions). Each virtual machine can run its
through Localization of Computation own operating system independent of other virtual
machines. Thus, it is possible to mix different operating
The term “fog computing” [27] or “edge computing” system on a multicore platform, and offer a high degree
[28] pushes application logic and the underlying data of independence between them. The ability to mix
processing from corporate data centers to the edge of the operating systems and to strictly separate different
network. Running applications at the edge cuts down software partitions provides an opportunity to manage
network latency and produces faster responses to end- both legacy software running a traditional real-time
users [28]. Therefore, the fog/edge computing further operating system and new IoT related functionality of
extends the cloud computing by addressing real-time different criticality running on general purpose
applications that do not fit the paradigm of the cloud at operating systems. From security perspective, system
the edge of the network, e.g., applications that require partitioning allows faster security update cycles on the
very low and predictable latency, and distributed control general purpose operating system and keep the stability
systems. These applications are hosted at the network of the real time operating system.
edge or end devices to reduce service latency and
improve quality of service (QoS). According to [29], In industrial automation domain, most legacy
Fog Computing supports emerging Internet of software is written for being executed on a single core.
Everything (IoE) applications that demand real- In order to manage system partitioning, it becomes
time/predictable latency, such as industrial automation essential to investigate spatial and temporal partitioning
domain, transportation, networks of sensors and of software code on multicore, and investigate how
actuators, and is suitable for real time big data and real software partitioning strategies on multicore will impact
time analytics. Besides the concept of edge computing, product behaviour with respect to the most important
computation offloading [30] is another way to manage industrial-specific quality attributes such as
latency by sending heavy computation to resourceful predictability and performance. For instance, as
servers and receiving results from these servers, based described in [34], it is desirable to run several real-time
on the available resources, such as network bandwidth, applications developed by different organizations and of
storage capacity, and processor performance. different criticality on a single computing board by
exploiting temporal partitioning. To guarantee real-time
Several other studies address latency and data requirements and criticality, this study suggests a
scalability issues as well. For instance, study [31] partition model with a mechanism that can decide each
presents a Smart Gateway, accompanied with a Smart partition’s period and execution time automatically
Network or Fog Computing to pre-process and trim preserving its criticality level.
high-frequency data sets that are useful for studying
dynamic behaviour analysis models for optimization 4.3. Managing Scalable and Secure Real-time
and monitoring before sending the data to the cloud. In Collaboration
this way, unnecessary communication burdens at the Decreasing the engineering effort spent on complicated
core network and the data centre in the cloud can be manual configurations of connected devices is essential
avoided. Another study [32] proposes to utilize a router for a scalable collaboration. Zero-configuration
instead of a central intelligent gateway for distribution networking is a technology that allows communication
of applications that have different latency requirements. between devices and improves network ease-of-use. It is
For instance, basic and real-time applications are built on top of three core technologies [35]: (i) Link-
realized by machine-to-machine communication local addressing – assigning numeric network addresses
between devices in local network. Storage and non-real- for devices in the network; (ii) Multicast DNS name
time applications are managed in cloud network. resolution – automatic distribution of computer
4.2. Managing Mixed Criticality through System hostnames; and (iii) DNS service discovery – automatic
Partitioning location of network services such as printers, cameras,
speakers, gateways, etc.
To enable software with mixed criticality to be
coexisting in one device, virtualization is a technology There are also different communication technologies
that has gained increasing interest in industrial systems. that enable the connection of heterogeneous devices and
Virtualization has been used extensively to manage applications, e.g., AMQP (Advanced Message Queuing
resources in server based systems and it is a technology Protocol) [36], MQTT (Message Queue Telemetry
that is also starting to be applied for embedded real-time Transport) [37], and DDS (Data Distribution Service)

537
[38]. AMQP is a message-centric protocol mostly for different ways to address fault tolerance. For instance,
enabling enterprise messaging. The authentication study [28] applies the edge computing concept to move
and/or encryption is based on Simple Authentication application logic, the underlying data processing, and
Security Layer (SASL) and/or Transport Layer Security services from centralized nodes to edge devices. The
(TLS) with an underlying reliable transport layer dependency on the corporate data enter is therefore
protocol such as Transmission Control Protocol (TCP). minimized. In this way, the single point of failure in the
MQTT is a simple lightweight machine-to-machine infrastructure is removed, thus potential denial of
protocol designed for devices with constrained service attacks is reduced, and service availability is
resources and unreliable networks. Similar to AMQP, improved. Study [41] addresses the issue of task
MQTT runs on networks that use TCP/IP as the migration in mixed-criticality embedded systems to
underlying transport, and use message brokers. DDS is tolerate permanent faults. For hardware faults, fault
designed to support large scale, real-time data sharing prediction using IoT [42] is a key solution that monitors
between distributed devices and embedded systems on a hardware conditions, process and analyse this
network. It can be used with either reliable or unreliable information for preventive maintenance to avoid
networks. DDS for Real-Time Systems is a specification hardware faults.
of a publish/subscribe middleware for distributed
4.5. Managing functional safety
systems created by the Object Management Group
(OMG). This specification defines a data-centric Functional safety in industrial systems have specific
publish-subscribe programming model that provides a requirements and are governed by safety standards and
global data space in distributed systems. For large scale authority regulations. We see three different strategies
applications that require high performance and real-time for functional safety in an industrial IoT context; A first
communications between devices, DDS can be a possible strategy is to completely separate the safety-
suitable candidate. related systems from the IoT systems, which means that
the IoT system is agnostic about the existence of any
Several other studies address the scalability issue in
safety-related system. A second approach is partial
large IoT systems. For instance, in order to invoke
integration where information from safety related
services to extend the cooperation between virtual and
systems are isolated and do not depend on IoT systems
physical worlds, the concepts of service orchestration
but can provide IoT systems with information. This also
and choreography are proposed in [39], describing an
includes cases with a mix of criticality where safety-
architecture that enables efficient integration of services
related software are separated and isolated by certified
by locally orchestrating distributed web-enabled
virtualization solutions (hypervisors) that provide
services in very large scale IoT systems and globally
adequate independence from a functional safety point of
choreographing Web-based applications. Another study
view. Finally, a third strategy, theoretically, would be to
[40] presents a scalable and self-configuring peer-to-
use IoT solutions directly for realizing safety-related
peer (P2P)-based architecture for large scale IoT
systems, which might be in conflict with standards and
networks, aiming at providing automated service and
regulations and thus require additional certifications
resource discovery mechanisms without human
within the foreseeable future.
intervention for configuration.
4.4. Managing Fault Tolerance 5. Summary
There are different ways for achieving fault tolerance With increased pressure to keep up with ever-
requirements. One common design principle is to use changing business needs, Internet-of-things allows
redundancy in the system. Typical redundancy schemes organizations to become more agile and aggressive in
include duplicating servers, controllers, I/O, and pursuing new revenue streams and new business
networks. For instance, study [25] designs a distributed models. It has the potential to change the traditional way
fault tolerance algorithm that is run asynchronously by of spending investment on building, maintaining, and
all redundant controllers so as to guarantee normal upgrading respective infrastructures for multiple
operation under failures. There are also specialized unconnected products. Through delivering new end-
controllers that offer built-in redundancy by applying customer software services that are based on
dual processor modules that are synchronized at information extracted from multiple connected
hardware level. products, Internet-of-things provides a new way of
For industrial automation, due to the importance of realizing business agility and faster pace of innovation.
having reliable and deterministic data transfer and We have started to witness these IoT innovations
analysis in real time, it is important to design local fall- deployed in many industries such as smart city, smart
back solutions, e.g., near the edge devices, in the overall energy, healthcare, logistics and retail, transportation,
edge to cloud software infrastructure when latency etc. However, scarce information is available for IoT
constraints for service requests are violated. There are usage in industrial automation domain for reliable and

538
collaborative automation. In this paper, based on the International Conference on High Performance
specific constraints that we have seen in the industrial Computing and Simulation (HPCS), 2012.
automation domain, we have described relevant [20] D. Miorandi, S. Sicari, F. De Pellegrini, I. Chlamtac,
Internet of Things: Vision, Applications and Research
challenges that we need to cope with when deploying Challenges, Journal of Ad Hoc Networks, 2012.
industrial IoT solutions in automation domain, and [21] R. Quinnell, 10 Top Challenges Industrial IoT Must
identified relevant technical solutions that potentially Overcome in 2015.
address these challenges. [22] IEC 61508, www.iec.ch/functionalsafety.
[23] OPC UA, https://opcfoundation.org/about/opc-
Finally, we see the remaining challenges within long- technologies/opc-ua/.
lived industrial automation systems with huge [24] S. Trujillo, A., Crespo, and A., Alonso, MultiPARTES:
complexity and proprietary solutions. By introducing Multicore Virtualization for Mixed-Criticality Systems,
Euromicro Conference on Digital System Design, 2013.
industrial IoT into these systems, connecting different
[25] T. Hegazy, M., Hefeeda, Industrial Automation as a
long-lived systems, and creating new services based on Cloud Service, IEEE Transactions on Parallel and
fast-paced technologies, additional challenges in Distributed Systems, 2014.
integrating these long lived systems and innovative IoT [26] S. Demeyer, S. Ducasse, O. M. Nierstrasz, Object-
solutions will arise – this has so far been largely Oriented Reengineering Patterns, ISBN 978-3-9523341-
2-6, Morgan Kaufmann, 2003.
unaddressed by the Industrial IoT community, and [27] http://blogs.cisco.com/ioe/cisco-iox-an-application-
remains to be a future research direction. enablement-framework-for-the-internet-of-things/
[28] H. Pang, K.L. Tan, Authenticating Query Results in Edge
References Computing, the 20th International Conference on Data
Engineering, 2004.
[1] IEEE Internet of Things, http://iot.ieee.org/about.html [29] Fog Computing, Ecosystem, Architecture and
[2] B. Sosinsky, Cloud Computing Bible, ISBN 978-0-470- Applications, Research at Cisco, Cisco RFP-2013-078,
90356-8, Wiley Publishing, Inc., 2011. www.cisco.com/web/about/ac50/ac207/crc_new/univers
[3] Y. Chen, H. Hu, Internet of Intelligent Things and Robot ity/RFP/rfp13078.html
as a Service, Journal of Simulation Modelling Practice [30] K. Kumar et al., A Survey of Computation Offloading for
and Theory, 2013. Mobile Systems, Journal of Mobile Networks and
[4] Industrial Internet Consortium, Applications, 2013.
www.industrialinternetconsortium.org/. [31] M. Aazam et al., Fog Computing and Smart Gateway
[5] Industrial Internet Insights Report for 2015, Accenture. Based Communication for Cloud of Things, International
Conference on Future Internet of Things (FiCloud), 2014.
[6] E. Borgia, The Internet of Things Vision: Key Features,
Applications and open Issues, Journal of Computer [32] K. Suzuki, M. Inoue, Home network system with cloud
Communications, 2014. computing and distributed autonomous control, IEEE
16th International Symposium on Consumer Electronics
[7] Internet of Things – From Research and Innovation to (ISCE), 2012.
Market Deployment, Editors: Ovidiu Vermesan, Peter
Friess, River Publishers Series in Communication. [33] K Sandström et al., Virtualization Technologies in
Embedded Real-Time Systems, the 18th International
[8] European Research Cluster on the Internet of Things, Conference on Emerging Technologies and Factory
http://www.internet-of-things-research.eu/ Automation, 2013.
[9] Industrial Internet Consortium, [34] H. W. Jin, S. Han, Temporal Partitioning for Mixed-
http://www.industrialinternetconsortium.org/ Criticality Systems, the 16th International Conference on
[10] Industri 4.0, http://www.plattform-i40.de/ Emerging Technologies and Factury Automation, 2011.
[11] European Commission Internet-of-things, [35] S. Chesire, D.H. Steinberg, Zero Configuration
http://ec.europa.eu/digital-agenda/en/internet-things Networking: The Definite Guide. USA: O’Reilly Media,
[12] IoT@Work, https://www.iot-at-work.eu/ Inc., 2005.
[13] Internet of Things – Architecture, http://www.iot- [36] AMQP, https://www.amqp.org/.
a.eu/public [37] MQTT, http://mqtt.org/.
[14] Ubiquitous, Secure Internet-of-Things with Location and [38] DDS, http://portals.omg.org/dds/.
Context-Awareness (BUTLER), http://www.iot- [39] K. Dar et al., Adaptable service composition for very-
butler.eu/ large-scale Internet of Things systems, The 12th
[15] VITAL, http://vital-iot.eu/project International Middleware Conference, 2011.
[16] J. Gubbi, R. Buyya, S. Marusic, M. Palaniswami, Internet [40] S. Cirani et al. A Scalable and Self-Configuring
of Things (IoT): A Vision, Architectural Elements, and Architecture for Service Discovery in the Internet of
Future Directions, Journal of Future Generation Things, Journal of Internet of Things, 2014.
Computer Systems, 2013. [41] P.K. Saraswat, P. pop, J. Madson, Task Migration for
[17] R. Dolin, Building an IoT for Industrial Control, Fault-tolerance in Mixed-criticaility Embedded Systems,
http://www.embedded.com/design/real-world- ACM SIGBED Review, Special Issue on the 2nd
applications/4426952/Building-an-IoT-for-industrial- International Workshop on Adaptive and Reconfigurable
control--Part-1--What-is-Industrial-IoT- Embedded Systems (APRES), vol. 6, issue 3, 2009.
[18] F. Liu, J. Tong, J. Mao, R. Bohn, J. Messina, L. Badger, [42] X. Xu, T. Chen, M. Minami, Intelligent Fault Prediction
D. Leaf, NIST Cloud Computing Reference Architecture, System Based on Internet of Things, Advanced
NIST Special Publication 500-292, 2011. Technologies in Computer, Consumer and Control, vol.
[19] J.Y. De la Pava Torres, C. Jimenenez-Guarin, 64, Issue 5, 2012.
MagosCloud Secure: A Secure, Highly Scalable Platform
for Services in an Opportunistic Environment,

539