- Email: [email protected]
Decentralised IoT Architecture for Efficient Resources Utilisation
Proceedings, 15th IFAC Conference on Proceedings, Proceedings, 15th 15th IFAC IFAC Conference Conference on on Programmable Devices and Embedded Systems Proceedings, 15th IFAC Conference on Programmable Devices and Embedded Systems Programmable Devices and Embedded Systemsonline at www.sciencedirect.com Available Ostrava, Czech Republic, May 23-25, 2018 Proceedings, 15th IFAC Conference on Programmable Devices and Embedded Systems Ostrava, Czech Czech Republic, Republic, May May 23-25, 23-25, 2018 2018 Ostrava, Programmable Devices and Embedded Systems Ostrava, Czech Republic, May 23-25, 2018 Ostrava, Czech Republic, May 23-25, 2018
ScienceDirect
IFAC PapersOnLine 51-6 (2018) 168–173
Decentralised IoT Architecture for Efficient Decentralised Decentralised IoT IoT Architecture Architecture for for Efficient Efficient Decentralised IoT Architecture Resources Utilisationfor Efficient Resources Utilisation Resources Utilisation ∗∗∗ ∗∗ Jozef Mocnej ∗∗∗ Winston K.G. Seah ∗∗ ∗∗ Adrian Pekar ∗∗∗ ∗∗∗
Jozef Adrian Pekar Jozef Mocnej Mocnej Winston Winston K.G. K.G. Seah Seah ∗∗∗∗ ∗∗ Adrian Pekar ∗∗∗ ∗∗∗∗ Iveta Zolotova Zolotova Jozef Mocnej ∗∗ Winston K.G. Seah ∗∗∗∗ Adrian Pekar Iveta Iveta Zolotova ∗∗ ∗∗∗∗ Adrian Pekar ∗∗∗ Jozef Mocnej Winston K.G. Seah Iveta Zolotova ∗∗∗∗ ∗ Zolotova ∗ Technical UniversityIveta of Kosice (e-mail: [email protected]) ∗ Technical University University of of Kosice Kosice (e-mail: (e-mail: [email protected]) [email protected]) ∗∗ ∗ Technical ∗∗ University Wellington, Zealand Technical University of of Kosice (e-mail:New [email protected]) ∗∗ Victoria Victoria University of Wellington, New Zealand (e-mail: (e-mail: University of Wellington, New Zealand (e-mail: ∗ ∗∗ Victoria Technical University of Kosice (e-mail: [email protected]) [email protected]) Victoria University of Wellington, New Zealand (e-mail: [email protected]) [email protected]) ∗∗ ∗∗∗Victoria University of Wellington, New Zealand (e-mail: ∗∗∗ Victoria University University of Wellington, Wellington, New Zealand Zealand (e-mail: [email protected]) ∗∗∗ Victoria of of Wellington, New New Zealand (e-mail: (e-mail: ∗∗∗ Victoria University [email protected]) [email protected]) Victoria University of Wellington, New Zealand (e-mail: [email protected]) [email protected]) ∗∗∗ ∗∗∗∗ Victoria University of Wellington, New Zealand (e-mail: ∗∗∗∗ of [email protected]) ∗∗∗∗ Technical Technical University University of Kosice Kosice (e-mail: (e-mail: [email protected]) [email protected]) of Kosice (e-mail: [email protected]) ∗∗∗∗ Technical University [email protected]) Technical University of Kosice (e-mail: [email protected]) ∗∗∗∗ Technical University of Kosice (e-mail: [email protected]) Abstract: The exponentially growing number of devices connected to the Internet, the diversity Abstract: number of of devices devices connected connected to to the the Internet, Internet, the the diversity diversity Abstract: The The exponentially exponentially growing growing number of the Internet Internet ofexponentially Things (IoT), (IoT),growing and the thenumber variety of of IoT protocol stackstoyield yield to concerns about IoT Abstract: The of devices connected the Internet, the diversity of the of Things and variety IoT protocol stacks to concerns about IoT of the Internet of Things (IoT), and the variety of IoT protocol stacks yield to concerns about IoT Abstract: TheA exponentially growing number of IoT devices connected toyield thecan Internet, thechallenges diversity sustainability. promising solution is in IoT integration platforms which address of the Internet of Things (IoT), and the variety of protocol stacks to concerns about IoT sustainability. A promising solution is in IoT integration platforms which can address challenges sustainability. A promising solution is in IoT integration platforms which can address challenges of theas Internet of promising Things scalability (IoT), and and the of IoTsurrounding protocol stacks yield to concerns about IoT such interoperability, adaptability IoT. Current implementations sustainability. A solution is invariety IoT integration platforms which can address challenges such as interoperability, scalability and adaptability surrounding IoT. Current implementations such as interoperability, scalability and adaptability surrounding IoT. Current implementations sustainability. A promising solution is inadaptability IoTarchitectures, integration platforms which can address challenges of IoT are based on centralised yet a decentralised IoT architecture such as platforms interoperability, scalability and surrounding IoT. Current implementations of IoT platforms are on architectures, yet IoT architecture of IoT platforms are based based on centralised centralised architectures, yet aa decentralised decentralised IoT architecture such as interoperability, scalability and adaptability surrounding IoT. Current implementations with logic moved to to the (network) edge can can offer offer several several benefits to IoT IoT platforms and devices. of IoT platforms arethe based on centralised yet a decentralised IoT and architecture with logic moved (network) edge to devices. with logic moved to (network) edge can architectures, offer several benefits benefits to IoT platforms platforms and devices. of IoT platforms arethe based on centralised architectures, yet a decentralised IoT Based architecture This paper identifies the feature set that aa decentralised IoT platform should have. on the with logic moved to the (network) edge can offer several benefits to IoT platforms and devices. This paper identifies the feature set that decentralised IoT platform should have. Based on This paper identifies the feature set that a decentralised IoT platform should have. Based on the the with logic moved to the (network) edge can offer several benefits to IoT platforms and devices. determined features, a general decentralised IoT architecture is proposed for efficient resource This paper identifies the feature set that a decentralised IoT platform should have. Based on the determined features, aa general decentralised IoT is for efficient resource determined features, the general decentralised IoT architecture architecture is proposed proposed forhave. efficient resource This paper identifies feature set that a decentralised IoT platform should Based on the utilisation. determined features, a general decentralised IoT architecture is proposed for efficient resource utilisation. utilisation. determined features, a general decentralised IoT architecture is proposed for efficient resource utilisation. © 2018, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. utilisation. Keywords: Internet of Things, Things, decentralised architecture, resource utilisation, efficiency, Keywords: utilisation, efficiency, Keywords: Internet Internet of of Things, decentralised decentralised architecture, architecture, resource resource utilisation, efficiency, optimisation. Keywords: Internet of Things, decentralised architecture, resource utilisation, efficiency, optimisation. optimisation. Keywords: Internet of Things, decentralised architecture, resource utilisation, efficiency, optimisation. optimisation. 1. INTRODUCTION INTRODUCTION 1. 1. INTRODUCTION 1. INTRODUCTION 1. INTRODUCTION Internet of of Things Things (IoT) (IoT) is is a a highly highly discussed discussed paradigm paradigm Internet Internet of Things (IoT) is a highly discussed paradigm aimed at connecting everyday devices to the Internet. It Internet of Things (IoT) is a highly discussed paradigm aimed at connecting everyday devices to the Internet. It aimed at ofconnecting everyday devicesdiscussed to the Internet. It Internet Things (IoT) is a highly paradigm empowers the systems with the ability to sense and conaimed at connecting everyday devices to the Internet. It empowers the the systems systems with with the the ability ability to to sense sense and and conconempowers aimed connecting to Internet. It trol theatenvironment environment around us.devices However, as the the number empowers the systemseveryday with the ability to the sense control the around us. However, as number trol the environment around us. However, as theand number empowers the systems with the ability to sense and conof IoT IoT devices and the the expectations towards the number system trol thedevices environment around us. However, as the of devices and expectations towards the system of and the expectations towards the system trolIoT thedevices environment around us.between However, assolutions the number increase, communication issues IoT will of IoT and the expectations towards the system increase, communication issues between IoT solutions will increase, communication issues between IoT solutions will of IoT devices and the expectations towards the system inevitably emerge. These subsequently lead to concerns increase, communication issues between IoT solutions will inevitably emerge. emerge. These These subsequently subsequently lead lead to to concerns concerns inevitably increase, communication issues between solutions will about sustainability, in particular, particular, issuesIoT related to interinterinevitably emerge. These subsequently lead to concerns (a) Centralised architecture (b) Decentralised architecture about sustainability, in issues related to (a) (b) Decentralised Decentralised architecture architecture about sustainability, in particular, issues related to inter(a) Centralised Centralised architecture architecture (b) inevitably emerge. These subsequently lead to concerns operability, scalability, and adaptability. In addition, as about sustainability, in particular, issues related to inter(a) Centralised architecture (b) Decentralised architecture operability, scalability, and adaptability. In addition, as operability, scalability, and adaptability. In addition, as Fig.(a)1. 1. Types Types of of IoT IoT architectures architectures about sustainability, particular, issues related to inter(b) Decentralised architecture IoT applications are often deployed in constrained envioperability, scalability, and adaptability. In addition, as Fig. IoT applications are in often deployed in constrained constrained enviFig. 1.Centralised Types of architecture IoT architectures IoT applications are often deployed in envioperability, scalability, and adaptability. In addition, as Fig. 1. Types of IoT architectures ronments using low-cost devices with limited computing IoT applications are often deployed constrained environments using low-cost devices withinlimited limited computing ronments using low-cost devices with computing can take place. Since several forwarding devices Fig. 1. Types of IoT architectures making can place. Since IoT applications are often deployed in constrained enviand power powerusing resources, the execution execution of computationally computationally in- making making can take take place. Since several several forwarding forwarding devices devices ronments low-cost devices with limited computing and resources, the of inand power resources, the execution of computationally inmust be involved in the communication between the end making can take place. Since several forwarding devices must be involved in communication between end ronments using low-cost devices with limited computing tensive operations as well as the service life present further must be can involved in the the Since communication between the the end and power resources, the execution of computationally intensive operations as well as the service life present further making take place. several forwarding devices tensive operations as well as the service life present further devices and controller, centralised approaches often lead must be involved in the communication between the end devices and controller, centralised approaches often lead and power resources, the execution of computationally inchallenges. devices and controller, centralised approaches often lead tensive operations as well as the service life present further challenges. must beand involved in thecentralised communication between the load end challenges. to unnecessary unnecessary increase in the traffic traffic utilisation and load devices controller, approaches often lead to increase in the utilisation and tensive operations as well as the service life present further to unnecessary increase centralised in the traffic utilisationoften and load challenges. devices and controller, approaches lead A promising solution to improve IoT sustainability are IoT of resources. A recent study on IoT network architecin the on traffic load A promising solution solution to to improve improve IoT IoT sustainability sustainability are are IoT IoT to of resources. recent IoT network architecchallenges. of unnecessary resources. A Aincrease recent study study on IoTutilisation network and architecA promising unnecessary in the traffic and load integration platforms. IoT platforms provide the foundatures performed by Verma et al. has that of resources. Aincrease recent study IoTutilisation network architecA promisingplatforms. solution toIoT improve IoT sustainability are IoT to integration platforms. IoT platforms provide the the foundafoundatures performed by Verma et on al. (2017) (2017) has shown shown that tures performed by Verma et al. (2017) has shown that integration platforms provide A promising solution to improve IoT sustainability are IoT of resources. A recent study on IoT network architection for connecting devices to the Internet, acquiring the centralised solutions will face challenges in handling tures performed by Verma et al. (2017) has shown that integration platforms. IoT platforms provide the foundation for for connecting connecting devices devices to to the the Internet, Internet, acquiring acquiring the the centralised solutions will challenges in handling the the centralised solutions will face face in the tion integration platforms. IoT platforms provide the foundatures performed by Verma et challenges al. (2017) hashandling shown that generated data, and devices processing them in a meaningful meaningful way, increased number of IoT IoT devices and the growing growing demands centralised solutions will face challenges in handling the tion for connecting to the Internet, acquiring the generated data, and processing them in a way, increased number of devices and the demands increased number of IoT devices and the growing demands generated data, and processing them in a meaningful way, tionthey for connecting devices to the Internet, acquiring the increased centralised solutions willdevices face challenges in handling the i.e. offer a standardised way to manage the connected of users and services. number of IoT and the growing demands generated data, and processing them in a meaningful way, i.e. they offer a standardised way to manage the connected of users and services. of users and services. i.e. they offer a standardised waythem to manage the connected generated data, and processing in a meaningful way, increased number of IoT devices and the growing demands devices. of users and services. i.e. they offer a standardised way to manage the connected devices. devices. Edge computing driven IoT IoT platforms represent represent an alteralterEdge computing driven i.e. they offer a standardised way to manage the connected of users and services. Edge computing driven IoT platforms platforms represent an an alterdevices. The first wave of IoT platforms placed the control, opernative approach which moves part of the decision-making Edge computing driven IoT platforms represent an alterThe first wave of IoT platforms placed the control, opernative approach which moves part of the decision-making devices. approach which moves part of the decision-making The first wave of IoT platforms placed the control, oper- native Edge computing driven IoT platforms represent an ational and computational logic in a geographically cencapabilities from the cloud to the network edge, e.g gatenative approach which moves part of the decision-making The first wave of IoT platforms placed the control, operational and and computational computational logic logic in in aa geographically geographically cencen- capabilities capabilities from from the the cloud cloud to to the the network network edge, edge, e.g e.g altergategateational The first wave of IoT platforms placed the control, opernative approach which moves part of the decision-making tralised location, most commonly in the cloud (Fig. 1(a)). way, intermediate device, etc. (Fig. 1(b)). In contrast to capabilities from the cloud to the network edge, e.g gateational and computational logic in a geographically centralised location, location, most most commonly commonly in in the the cloud cloud (Fig. (Fig. 1(a)). 1(a)). way, way, intermediate intermediate device, device, etc. etc. (Fig. (Fig. 1(b)). 1(b)). In In contrast contrast to to tralised ational and logic Azure, in athe geographically cen- capabilities from the cloud toadvantage the network edge, e.g gateAmazon Webcomputational Services, Microsoft Azure, and IBM Bluemix centralised architectures, the advantage of distributing distributing the way, intermediate device, etc. (Fig. 1(b)). In contrast to tralised location, most commonly in cloud (Fig. 1(a)). Amazon Web Services, Microsoft and IBM Bluemix centralised architectures, the of the centralised architectures, the advantage of distributing the Amazon Web Services, Microsoft Azure, and IBM Bluemix tralised location, most Microsoft commonly in the and cloud (Fig. 1(a)). way, intermediate device,the etc.advantage (Fig. 1(b)). In contrast to are just a few examples of cloud platforms which found logic between the devices throughout the entire network is centralised architectures, of distributing the Amazon Web Services, Azure, IBM Bluemix are just a few examples of cloud platforms which found logic between the devices throughout the entire network is between the devices the throughout theofentire networkthe is are just Web a fewServices, examples of cloudAzure, platforms which found logic Amazon Microsoft and IBM Bluemix centralised architectures, advantage distributing their place soon in production. However, although centhe significant reduction of transferred data, which conlogic between the devices throughout the entire network is are just a few examples of cloud platforms which found their place place soon soon in in production. production. However, However, although although cencen- the significant reduction of transferred data, which conthe significant reduction of transferred data, which contheir are just a few examples of cloud platforms which found logic between the devices throughout the entire network is tralised architectures offer a a simplified simplified control and manmansequently leadsreduction to the the decrease decrease of communication communication delay the significant of transferred data, which contheir place soon in production. However, although centralised architectures offer control and sequently leads to of delay sequently leadsreduction to the decrease of communication delay tralised architectures offer a simplified control and mantheir place soon in production. However, although centhe significant of transferred data, which conagement,architectures in practice, practice, all all thea data data must be be forwarded to aa sequently and utilisation utilisation oftoresources resources including cloud services. services. IoT IoT leads of the decrease of communication tralised offer simplified control and management, in the must forwarded to including cloud and utilisation including cloud services.delay IoT agement, in practice, all the data must be forwarded to a and tralised controller architectures offer athe simplified control and mansequently leads of toresources the decentralised decrease of communication delay central located in cloud before any decision platforms based on the architecture also perand utilisation of resources including cloud services. IoT agement, in practice, all the data must be forwarded to a central controller located in the cloud before any decision platforms based on the decentralised architecture also perplatforms basedof onresources the decentralised architecture also percentral controller located in the cloud before any decision agement, in practice, all the data must be forwarded to a and utilisation including cloud services. IoT central controller located in the cloud before any decision platforms based on the decentralised architecture also percentral controller located in the cloud before any decision platforms based on the decentralised architecture also per2405-8963 © 2018, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Copyright © 2018 IFAC 168 Copyright 2018 168 Copyright © under 2018 IFAC IFAC 168 Control. Peer review© responsibility of International Federation of Automatic Copyright © 2018 IFAC 168 10.1016/j.ifacol.2018.07.148 Copyright © 2018 IFAC 168
2018 IFAC PDES Ostrava, Czech Republic, May 23-25, 2018 Jozef Mocnej et al. / IFAC PapersOnLine 51-6 (2018) 168–173
form better in environments in which a high number of connected devices communicating over an unreliable and constrained network produce a large amount of data. On the contrary, decentralised architectures bring a certain complexity into the management process. In consequence, they create a trade-off between the complexity of management and efficient utilisation of resources. However, as communication complexity challenges can be overcome with an appropriate architecture design, to address sustainability related challenges, we have decided to shift away from traditional approaches and propose a decentralised IoT architecture. This architecture is based on considerations, such as, the method for runtime monitoring including available resources and utilisation, operational requirements and costs, as well as, system adaptation to prolong the service life of battery-powered devices. The contribution of this paper is threefold. First, we identify the optimal set of features required by a generalised form of decentralised IoT platforms. Second, we describe a novel approach for efficient resource utilisation. Third, we specify the structure of the overall architecture. The rest of this paper is organised as follows. Section 2 describes the motivation and related work. Section 3 presents the optimal feature set for decentralised IoT platforms. Section 4 introduces a concept for achieving efficient utilisation of available resources, whereas a proposal of a decentralised IoT architecture is analysed in Section 5. Finally, Section 6 draws a conclusion and provides directions for future work. 2. MOTIVATION AND RELATED WORK In a real-world environment, several factors influence IoT sustainability, including the heterogeneity and relocatability (e.g. static or mobile) of the devices, topology interconnecting the nodes, and condition of the environment. These factors are all present in the most common IoT application domains starting from smart manufacturing and retail (Lojka et al. (2015)), through smart buildings and living (Mekki et al. (2017)), smart agriculture (Paraforos et al. (2016)), up to smart city (Djahel et al. (2015)). An equally important factor is the type of energy source used to power the devices. Currently, either mains, batteries, renewable energy sources or the combination thereof are used. Among them, IoT devices are usually batterypowered (Somov and Giaffreda (2015)). Since resource utilisation, including power consumption, is a major concern in current solutions, it becomes essential to efficiently utilise the power sources in order to extend their lifespan. However, determining the optimal configuration is a challenge itself. Network traffic characteristics as well as the environment can change rapidly over time depending on various factors such as reconfigurations, failures in the topology, the time of the day, whether is there an anomaly or only standard data transmissions are performed. Static or manual configuration becomes unfeasible due to the continuously changing conditions. A decentralised IoT platform with dynamic decision making and intelligent (re)configuration placed at the edge represent a paradigm which could be a step forward in utilising efficiently the available resources. 169
169
However, although improvements can be hypothetically achieved, as the number of the variables that need to be considered rises, the complexity and computing demands of the entire system are increasing as well. As a result, this can similarly lead to efficiency drop in resource utilisation. Considering the above mentioned, our targeted research question can be formulated as follows: What is the (1) optimal decentralised IoT architecture design suitable for (2) obtaining, selecting and using information to (3) achieve improvements in resource utilisation? The architecture design of IoT platforms has been discussed by several works. Sarkar et al. (2014) proposed a scalable distributed architecture tackling characteristics like heterogeneity, scalability, interoperability, automation, zero-configuration and distributiveness. They introduced a three-layered architecture design aimed at supporting multi-application use cases with minimum management effort. This was achieved using distributed intelligence empowered devices which process the data and make decisions at the edge without the need for centralised logic in the cloud. However, as centralised computing can be beneficial in circumstances when, for example, further analysis or accounting is required, abandoning the cloud entirely should be avoided. In addition, the problem of energy efficiency was not addressed. Datta et al. (2014) introduced an IoT architecture with a wireless gateway as the main intermediate device responsible for interactions between the sensors (actuators) and mobile clients. This gateway-centric approach consists of south and north interfaces that use generated metadata to expose the capability of the devices to mobile clients. A similar effort by Vallati et al. (2016) proposed a distributed architecture also utilising gateways. They claim that their platform implementation called BETaaS simplifies the deployment of horizontal solutions by exposing a unified service oriented interface to access things. Although these proposals discuss viable architecture design ideas including distributed logic, scalability, and interoperability, they fall short of addressing the challenges of resource utilisation. Likewise, energy efficiency plays a vital role in the lifespan of IoT devices, and therefore should deserve more attention. Yet, there are only a few efforts targeted at energy efficiency. Rault et al. (2014) developed various mechanisms for improving resource efficiency, and Liu et al. (2014) focused on both ensuring Quality-of-Information (QoI) as well as energy efficiency in sensory environments. However, these consider only a small number of factors (characteristics) influencing the sustainability of IoT. In contrast to the above-mentioned research, energy efficiency and resource utilisation are the key emphasis in this paper. In addition, we also reevaluate the expectations of decentralised IoT platforms and extend the optimal feature set with additional features such as multi-network approach and simplified device management (both central and individual).
2018 IFAC PDES 170 Ostrava, Czech Republic, May 23-25, 2018 Jozef Mocnej et al. / IFAC PapersOnLine 51-6 (2018) 168–173
3. OPTIMAL FEATURE SET FOR DECENTRALISED IOT PLATFORMS Centralised architectures often perform inefficiently in present IoT environments. Efficient data processing, decision making and resource utilisation require new approaches such as a decentralised IoT platform. Obviously, to ensure its sustainability, an appropriate design is essential but itself a challenge, let alone an optimal design. As a step forward, we have identified the network traffic characteristics of the most common IoT services, categorized into eight application domains as per Mocnej et al. (2018). In particular, we examined smart buildings and living, smart healthcare, smart environment, smart city, smart energy, smart transport and mobility, smart manufacturing and retail, and smart agriculture from the perspectives of three user groups, individuals, society, and industry. One observation deduced from this study is a set of features that a decentralised IoT architecture of the future should meet. These features are as follow: Feature 1. Multi-network approach – an IoT platform is destined to connect devices regardless of the underlying network technology in use. IoT is diverse and since different constraints may require different solutions, many network technologies exist designed specifically for different parts of the IoT. Ideally, an IoT platform should enable communication with all heterogeneous devices in a standardised way by hiding the underlying network technologies. Feature 2. Scalable and interoperable implementation – the shift from application-specific to application-independent solutions is highly desired. As the conditions and expectations of IoT application domains may dynamically change over time, an adaptive IoT architecture should recognise and react to these changes in a timely manner. In conclusion, an IoT platform of the future needs to be scalable and non-invasive, and ready to integrate new applications in any circumstances. Feature 3. Low power consumption – low power consumption and efficient utilisation of resources must be a priority to achieve the desired sustainability of the solution. An IoT platform should therefore support mechanisms such as event-driven M2M (Verma et al.) communication and other optimisation methods to achieve improvements in energy utilisation. Feature 4. Intuitive data and device management – to keep pace with the rapid development and innovation, efficient device control and data collection should be present in IoT by design. A platform should support both individual and central management for accessibility, control and modifications. Furthermore, many routines should be automated to make the configuration process simpler and more intuitive. Feature 5. Artificial Intelligence at the edge – Artificial Intelligence (AI) and Machine Learning (ML) are gaining on popularity year by year. Today, they are utilised in the vast majority of technologies and services. IoT is not an exception where AI is most commonly utilised in the cloud to provide a number of improvements. However, in the case of a decentralised architecture, part of the services should be moved from the cloud to the network edge including AI. The most suitable device for hosting AI is an 170
unconstrained gateway which offers reasonable computing resources and sustained energy source. A decentralised IoT platform should be also designed to support contextawareness. These five key features have been carefully incorporated into the architecture of our proposed decentralised solution. 4. A CONCEPT FOR ACHIEVING EFFICIENT RESOURCE UTILISATION Our objective is to achieve improvements in the utilisation of available resources. Available resources denote connected devices providing sensing/actuating services to the IoT platform. How these resources are utilised depend on the applications implemented in the platform and their requirements (expectations). Since the requirements may change over time (which subsequently affects the resource utilisation), the individual changes have to be tracked and recorded over time. This requires: (i) Monitoring the platform in runtime to obtain insight into the system and its processes; (ii) Ensuring the desired quality of the output by taking the appropriate measures; an example output would be optimal subset of active resources for a particular scenario; (iii) Optimising the utilisation of available resources. In the following, these three research topics will be discussed in more details. In general, tasks (i)–(iii) represent our main research topics. 4.1 Monitoring IoT Platforms in Runtime To monitor IoT platforms in runtime, a number of measures can be used: • Quality of Device (QoD) – is responsible for expressing the quality of a particular device and can monitor various attributes, including, but not limited to, battery life, precision, and sending rate (Buchholz and Schiffers (2003)). • Quality of Service (QoS) – uses attributes such as bandwidth, delay, jitter, and packet loss to monitor the performance characteristics of the network (Bhuyan et al. (2010)). • Quality of Information (QoI) – characterises the fitness-for-use of data, and therefore is suitable for measuring the quality of the obtained data from a particular device but can also define the quality of the output by using attributes like accuracy, precision, and freshness (Bisdikian et al. (2013)). Fig. 2 depicts an IoT integration scenario of these measures, i.e. QoD for end device monitoring, QoS for network performance monitoring, and QoI for transferred data quality monitoring is used. One advantage of these measures is that, although they are principally destined to describe a specific part of a system, when used together, they can also provide performance overview of the entire system.
2018 IFAC PDES Ostrava, Czech Republic, May 23-25, 2018 Jozef Mocnej et al. / IFAC PapersOnLine 51-6 (2018) 168–173
171
ticularly highlight the significance of the following three power plan modes to achieve optimal energy efficiency: (1) OFF; (2) ON and sensing; and (3) ON and transmitting. 5. DECENTRALISED IOT ARCHITECTURE
Fig. 2. Implementation of measures in an IoT integration scenario. 4.2 Ensuring the Desired Output Quality
Our decentralised IoT architecture is composed of three main functional blocks: end device, gateway and cloud. We provide a stack diagram for the end device and gateway blocks which outline the modules and their function. 5.1 End Device
The applications determining the requirements (expectations) for the resources are communicating through gateway(s) where each application has its own stream processing unit as per Fig. 2. In general, both the requirements and the capabilities of the resources can be expressed as functions with QoI attributes. To ensure the desired output quality the correlation between these functions have to be identified. To achieve this, we define a further measure: • Value Of Information (VoI) – deals with an assessment of the information utility for a specific use case scenario using attributes such as relevance, integrity, timeliness, and understandability (Bisdikian et al. (2013)). From a decentralised architecture perspective, we can utilise this measure to evaluate the significance of resources for an application. More specifically, by leveraging VoI a subset of optimal resources can be selected for every stream processing module. This provides an option to determine what constitutes to the desired output quality and under what conditions this quality can be achieved. 4.3 Optimising the Utilisation of Available Resources Assuming Section 4.1 and Section 4.2 are achieved, the utilisation of available resources can be optimised. Since the majority of IoT devices is constrained (e.g. by their battery lives), it is preferred to utilise them efficiently. Keeping the transceiver powered on of a device is an energy expensive task, as shown by Harrison et al. (2016). Therefore, one approach to extend the lifespan is to power off those battery-powered devices which generate duplicate data or whose absence is acceptable by any stream processing. In general, this represents an optimisation problem of finding a subset of active devices that would minimise the overall consumption of limited energy resources in the network while satisfying at the same time the requirement of minimal output quality. Another approach is to predict the sensed value of a device rather than powering it on to execute its task(s). This subsequently leads to the minimisation of communication between the devices and the gateways, extending the lifespan. Prediction could be realised as long as QoI of a device satisfies the minimum requirements, which could be specified by the stream processing modules and affected by the dynamics of the environment. Consequently, our approach consists of duty-cycling and adaptive sampling for controlling the power plans and sending rates of devices to conserve their energy. We par171
The computing capabilities of end devices can be diverse. This fact was taken into consideration when the threelayered architecture depicted in Fig. 3 was proposed. The description of the individual layers is as follows: • Connectivity abstraction layer – is responsible for connecting a device to a network. The layer includes a single module called Connectivity SDK that hides the underlying network technology and provides the standardised communication capabilities for the upper layers. • Device services layer – defines services necessary for each device to efficiently utilise its resources. This layer contains four modules: · data filtering – is responsible for a very simple and lightweight selection of important data from all sensed values. · communication service – uses communication capabilities from Connectivity SDK and wraps them to the publicly accessible functions. It also provides a way to remotely control the services of devices. · power plan – manages how a device uses power to achieve Feature 3. By default, it supports at least three power plans - ON, ON with a transmitter turned off, and OFF. · Over-The-Air (OTA) – permits updates over the air to always keep a device’s software up-to-date. • Custom application layer – for a device serving a particular task. It can also enhance the supported functionalities if a device has enough computing resources. An application is installed on a device during its manufacturing process but can be later modified by the OTA module. 5.2 Gateway Since a gateway usually has unconstrained power resources, it can be utilised to support computing-intensive tasks to minimise the load (e.g during decision making)
Fig. 3. End device architecture.
2018 IFAC PDES 172 Ostrava, Czech Republic, May 23-25, 2018 Jozef Mocnej et al. / IFAC PapersOnLine 51-6 (2018) 168–173
of other IoT devices. The architecture showed in Fig. 4 contains three main functionalities: connectivity, end device management, and data management. Connectivity is an essential functionality of the gateway provided by the lower layers of the architecture. To ensure the conformity of our solution with Feature 1, in the connectivity abstraction layer we implement public and well-known IoT specifications such as those defined by OCF 1 or OneM2M 2 . However, our approach is not limited to only these standards. Custom communication protocols can be implemented using a designated custom module. All connectivity abstraction modules are unified by the network manager, which is responsible for the interactions between the services residing at higher layers and connectivity functionalities at the lower layers. End device and data management functionalities are provided by the gateway services layer. In general, this layer is responsible for providing Feature 5. When data from a network arrives at the inbound interface of a gateway, it is processed by the data services module. This module provides functionalities such as QoI measurement and prediction. The obtained data are subsequently transferred to the stream processing module which is the pipeline for various custom applications. Data selection by these applications from the pipeline is performed based on the significance level expressed by VoI. When the appropriate input is determined, applications continue in the execution of their tasks. Application outputs can be further processed in a number of ways. One option is to pass the data to the end devices or another gateway directly. Another option is to store the output in a temporary database and forward it later (if real-time forwarding in not necessary) to, for example, the cloud. The interaction between the individual blocks and modules including the type of communication and measurement method is depicted in Fig. 5. An important component of loosely coupled systems is service orchestration. It provides an efficient way to define functionalities as small independent services which are added to a central list and linked together. This approach is more efficient than presenting all functionalities as one complex application since this way services can be dynamically modified over 1 2
time, ensuring Feature 2. Service orchestration based on service-oriented architecture (SOA) is implemented in our solution at the service orchestration layer. Lastly, we offer the options for gateway communication: (1) via a user-friendly GUI and/or (2) remote connection using the exposed API. This way, both simple as well as advanced gateway control is achieved, ensuring the satisfaction of Feature 4. 5.3 The Cloud Due to the number of services offered by providers, cloud solutions present an advantageous foundation for IoT solutions and cloud-based IoT solutions have already been proposed (Zhou et al. (2013); Hou et al. (2016)). From the perspective of our architecture, one important service we would like add is network management. The network management service module provides control of the entire network from a centralised place. It utilises the acquired metadata to generate the scheme of the network and subsequently provides the user with a way to execute configuration changes and updates without the need for connecting to every single gateway individually. Its main advantage is to deliver a solution using a system which can be configured once and then deployed everywhere. This intuitive data and device management at a centralised place satisfies Feature 4. 6. CONCLUSION This paper addresses the problem of achieving sustainability in diverse IoT platforms. Although the current wave of IoT platforms is based on centralised architectures, we outlined the benefits of decentralised architectures, especially when deployed in a complex real world environment. The main contribution of this paper is the concept of a decentralised IoT architecture. This architecture incorporates five key features a decentralised IoT platform of the future should meet. Based on these features we proposed a solution intended to optimise the resource utilisation of IoT components. Lastly, we also presented the functional blocks of our proposed approach along with some key modules and features. Future work will be aimed at the optimisation process for improving resource utilisation. To achieved this, the individual research topics (see Section 4 have to be investigated in greater details as each of them reveals further open issues and presents new challenges. Future work will also be focused on the development of a prototype based on the proposed decentralised IoT platform that will be used for evaluating the methods and procedures proposed to address IoT challenges.
https://openconnectivity.org http://www.onem2m.org
ACKNOWLEDGEMENTS This publication is prepared within the grant VEGA 1/0663/17, 2017-2020 (100%). Jozef Mocnej would like to thank the European Commission under the Thelxinoe project for the funding that supported his visit to Victoria University of Wellington (VUW), New Zealand. Adrian Pekar and Winston K.G. Seah are supported in part by VUW’s Huawei NZ Research Programme, SoftwareDefined Green Internet of Things project #E2881.
Fig. 4. Gateway architecture 172
2018 IFAC PDES Ostrava, Czech Republic, May 23-25, 2018 Jozef Mocnej et al. / IFAC PapersOnLine 51-6 (2018) 168–173
173
Fig. 5. Interactions between the functional blocks and modules. REFERENCES Bhaskar Bhuyan et al. Quality of service (QoS) provisions in wireless sensor networks and related challenges. Wireless Sensor Network, 2(11):861, 2010. Chatschik Bisdikian, Lance M Kaplan, and Mani B Srivastava. On the quality and value of information in sensor networks. ACM Transactions on Sensor Networks (TOSN), 9(4):48, 2013. Thomas Buchholz and Michael Schiffers. Quality of context: What it is and why we need it. In Proceedings of the 10th Workshop of the OpenView University Association: OVUA’03, 2003. Soumya Kanti Datta, Christian Bonnet, and Navid Nikaein. An IoT gateway centric architecture to provide novel M2M services. In Proceedings of the IEEE World Forum on Internet of Things (WF-IoT), pages 514–519, Seoul, South Korea, 6-8 March 2014. Soufiene Djahel, Ronan Doolan, Gabriel-Miro Muntean, and John Murphy. A communications-oriented perspective on traffic management systems for smart cities: challenges and innovative approaches. IEEE Communications Surveys & Tutorials, 17(1):125–151, 2015. DC Harrison, D Burmester, Winston K. G. Seah, and R Rayudu. Busting myths of energy models for wireless sensor networks. Electronics Letters, 52(16):1412–1414, 2016. L. Hou, S. Zhao, X. Xiong, K. Zheng, P. Chatzimisios, M. S. Hossain, and W. Xiang. Internet of things cloud: Architecture and implementation. IEEE Communications Magazine, 54(12):32–39, December 2016. ISSN 0163-6804. C. H. Liu, J. Fan, J. W. Branch, and K. K. Leung. Toward QoI and Energy-Efficiency in Internet-of-Things Sensory Environments. IEEE Transactions on Emerging Topics in Computing, 2(4):473–487, Dec 2014. ISSN 2168-6750. Tomas Lojka, Marek Bundzel, and Iveta Zolotova. Industrial gateway for data acquisition and remote control. Acta Electrotechnica et Informatica, 15(2):43–48, 2015. Kais Mekki, William Derigent, Eric Rondeau, and Andr´e Thomas. Data lifecycle management in smart building using wireless sensors networks. IFAC-PapersOnLine, 50(1):12944–12949, 2017. Jozef Mocnej, Adrian Pekar, Winston K. G. Seah, and Iveta Zolotova. Network Traffic Characteristics 173
of the IoT Application Use Cases. Technical Report ECSTR18-01, School of Engineering and Computer Science, Victoria University of Wellington, 6 Jan 2018. URL https://ecs.victoria.ac.nz/Main/ TechnicalReportSeries. Dimitris S Paraforos, Vangelis Vassiliadis, Dietrich Kortenbruck, Kostas Stamkopoulos, Vasileios Ziogas, Athanasios A Sapounas, and Hans W Griepentrog. A farm management information system using future internet technologies. IFAC-PapersOnLine, 49(16):324– 329, 2016. Tifenn Rault, Abdelmadjid Bouabdallah, and Yacine Challal. Energy efficiency in wireless sensor networks: A top-down survey. Computer Networks, 67:104–122, 2014. Chayan Sarkar, SN Akshay Uttama Nambi, R Venkatesha Prasad, and Abdur Rahim. A scalable distributed architecture towards unifying iot applications. In Proceedings of the IEEE World Forum on Internet of Things (WFIoT), pages 508–513, Seoul, South Korea, 6-8 March 2014. Andrey Somov and Raffaele Giaffreda. Powering IoT Devices: Technologies and Opportunities. Internet of Things Newsletter, 9 November 2015. Carlo Vallati et al. BETaaS: A platform for development and execution of machine-to-machine applications in the Internet of Things. Wireless Personal Communications, 87(3):1071–1091, 2016. Pawan Kumar Verma et al. Machine-to-Machine (M2M) communications: A survey. Journal of Network and Computer Applications, pages 83 – 105. ISSN 1084-8045. S. Verma, Y. Kawamoto, Z. M. Fadlullah, H. Nishiyama, and N. Kato. A survey on network methodologies for real-time analytics of massive iot data and open research issues. IEEE Communications Surveys & Tutorials, 19 (3):1457–1477, thirdquarter 2017. J. Zhou, T. Lepp¨anen, E. Harjula, M. Ylianttila, T. Ojala, C. Yu, and H. Jin. Cloudthings: A common architecture for integrating the internet of things with cloud computing. In Proceedings of the 2013 IEEE 17th International Conference on Computer Supported Cooperative Work in Design (CSCWD), pages 651–657, Whistler, BC, Canada, 27-29 June 2013.
ScienceDirect
IFAC PapersOnLine 51-6 (2018) 168–173
Decentralised IoT Architecture for Efficient Decentralised Decentralised IoT IoT Architecture Architecture for for Efficient Efficient Decentralised IoT Architecture Resources Utilisationfor Efficient Resources Utilisation Resources Utilisation ∗∗∗ ∗∗ Jozef Mocnej ∗∗∗ Winston K.G. Seah ∗∗ ∗∗ Adrian Pekar ∗∗∗ ∗∗∗
Jozef Adrian Pekar Jozef Mocnej Mocnej Winston Winston K.G. K.G. Seah Seah ∗∗∗∗ ∗∗ Adrian Pekar ∗∗∗ ∗∗∗∗ Iveta Zolotova Zolotova Jozef Mocnej ∗∗ Winston K.G. Seah ∗∗∗∗ Adrian Pekar Iveta Iveta Zolotova ∗∗ ∗∗∗∗ Adrian Pekar ∗∗∗ Jozef Mocnej Winston K.G. Seah Iveta Zolotova ∗∗∗∗ ∗ Zolotova ∗ Technical UniversityIveta of Kosice (e-mail: [email protected]) ∗ Technical University University of of Kosice Kosice (e-mail: (e-mail: [email protected]) [email protected]) ∗∗ ∗ Technical ∗∗ University Wellington, Zealand Technical University of of Kosice (e-mail:New [email protected]) ∗∗ Victoria Victoria University of Wellington, New Zealand (e-mail: (e-mail: University of Wellington, New Zealand (e-mail: ∗ ∗∗ Victoria Technical University of Kosice (e-mail: [email protected]) [email protected]) Victoria University of Wellington, New Zealand (e-mail: [email protected]) [email protected]) ∗∗ ∗∗∗Victoria University of Wellington, New Zealand (e-mail: ∗∗∗ Victoria University University of Wellington, Wellington, New Zealand Zealand (e-mail: [email protected]) ∗∗∗ Victoria of of Wellington, New New Zealand (e-mail: (e-mail: ∗∗∗ Victoria University [email protected]) [email protected]) Victoria University of Wellington, New Zealand (e-mail: [email protected]) [email protected]) ∗∗∗ ∗∗∗∗ Victoria University of Wellington, New Zealand (e-mail: ∗∗∗∗ of [email protected]) ∗∗∗∗ Technical Technical University University of Kosice Kosice (e-mail: (e-mail: [email protected]) [email protected]) of Kosice (e-mail: [email protected]) ∗∗∗∗ Technical University [email protected]) Technical University of Kosice (e-mail: [email protected]) ∗∗∗∗ Technical University of Kosice (e-mail: [email protected]) Abstract: The exponentially growing number of devices connected to the Internet, the diversity Abstract: number of of devices devices connected connected to to the the Internet, Internet, the the diversity diversity Abstract: The The exponentially exponentially growing growing number of the Internet Internet ofexponentially Things (IoT), (IoT),growing and the thenumber variety of of IoT protocol stackstoyield yield to concerns about IoT Abstract: The of devices connected the Internet, the diversity of the of Things and variety IoT protocol stacks to concerns about IoT of the Internet of Things (IoT), and the variety of IoT protocol stacks yield to concerns about IoT Abstract: TheA exponentially growing number of IoT devices connected toyield thecan Internet, thechallenges diversity sustainability. promising solution is in IoT integration platforms which address of the Internet of Things (IoT), and the variety of protocol stacks to concerns about IoT sustainability. A promising solution is in IoT integration platforms which can address challenges sustainability. A promising solution is in IoT integration platforms which can address challenges of theas Internet of promising Things scalability (IoT), and and the of IoTsurrounding protocol stacks yield to concerns about IoT such interoperability, adaptability IoT. Current implementations sustainability. A solution is invariety IoT integration platforms which can address challenges such as interoperability, scalability and adaptability surrounding IoT. Current implementations such as interoperability, scalability and adaptability surrounding IoT. Current implementations sustainability. A promising solution is inadaptability IoTarchitectures, integration platforms which can address challenges of IoT are based on centralised yet a decentralised IoT architecture such as platforms interoperability, scalability and surrounding IoT. Current implementations of IoT platforms are on architectures, yet IoT architecture of IoT platforms are based based on centralised centralised architectures, yet aa decentralised decentralised IoT architecture such as interoperability, scalability and adaptability surrounding IoT. Current implementations with logic moved to to the (network) edge can can offer offer several several benefits to IoT IoT platforms and devices. of IoT platforms arethe based on centralised yet a decentralised IoT and architecture with logic moved (network) edge to devices. with logic moved to (network) edge can architectures, offer several benefits benefits to IoT platforms platforms and devices. of IoT platforms arethe based on centralised architectures, yet a decentralised IoT Based architecture This paper identifies the feature set that aa decentralised IoT platform should have. on the with logic moved to the (network) edge can offer several benefits to IoT platforms and devices. This paper identifies the feature set that decentralised IoT platform should have. Based on This paper identifies the feature set that a decentralised IoT platform should have. Based on the the with logic moved to the (network) edge can offer several benefits to IoT platforms and devices. determined features, a general decentralised IoT architecture is proposed for efficient resource This paper identifies the feature set that a decentralised IoT platform should have. Based on the determined features, aa general decentralised IoT is for efficient resource determined features, the general decentralised IoT architecture architecture is proposed proposed forhave. efficient resource This paper identifies feature set that a decentralised IoT platform should Based on the utilisation. determined features, a general decentralised IoT architecture is proposed for efficient resource utilisation. utilisation. determined features, a general decentralised IoT architecture is proposed for efficient resource utilisation. © 2018, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved. utilisation. Keywords: Internet of Things, Things, decentralised architecture, resource utilisation, efficiency, Keywords: utilisation, efficiency, Keywords: Internet Internet of of Things, decentralised decentralised architecture, architecture, resource resource utilisation, efficiency, optimisation. Keywords: Internet of Things, decentralised architecture, resource utilisation, efficiency, optimisation. optimisation. Keywords: Internet of Things, decentralised architecture, resource utilisation, efficiency, optimisation. optimisation. 1. INTRODUCTION INTRODUCTION 1. 1. INTRODUCTION 1. INTRODUCTION 1. INTRODUCTION Internet of of Things Things (IoT) (IoT) is is a a highly highly discussed discussed paradigm paradigm Internet Internet of Things (IoT) is a highly discussed paradigm aimed at connecting everyday devices to the Internet. It Internet of Things (IoT) is a highly discussed paradigm aimed at connecting everyday devices to the Internet. It aimed at ofconnecting everyday devicesdiscussed to the Internet. It Internet Things (IoT) is a highly paradigm empowers the systems with the ability to sense and conaimed at connecting everyday devices to the Internet. It empowers the the systems systems with with the the ability ability to to sense sense and and conconempowers aimed connecting to Internet. It trol theatenvironment environment around us.devices However, as the the number empowers the systemseveryday with the ability to the sense control the around us. However, as number trol the environment around us. However, as theand number empowers the systems with the ability to sense and conof IoT IoT devices and the the expectations towards the number system trol thedevices environment around us. However, as the of devices and expectations towards the system of and the expectations towards the system trolIoT thedevices environment around us.between However, assolutions the number increase, communication issues IoT will of IoT and the expectations towards the system increase, communication issues between IoT solutions will increase, communication issues between IoT solutions will of IoT devices and the expectations towards the system inevitably emerge. These subsequently lead to concerns increase, communication issues between IoT solutions will inevitably emerge. emerge. These These subsequently subsequently lead lead to to concerns concerns inevitably increase, communication issues between solutions will about sustainability, in particular, particular, issuesIoT related to interinterinevitably emerge. These subsequently lead to concerns (a) Centralised architecture (b) Decentralised architecture about sustainability, in issues related to (a) (b) Decentralised Decentralised architecture architecture about sustainability, in particular, issues related to inter(a) Centralised Centralised architecture architecture (b) inevitably emerge. These subsequently lead to concerns operability, scalability, and adaptability. In addition, as about sustainability, in particular, issues related to inter(a) Centralised architecture (b) Decentralised architecture operability, scalability, and adaptability. In addition, as operability, scalability, and adaptability. In addition, as Fig.(a)1. 1. Types Types of of IoT IoT architectures architectures about sustainability, particular, issues related to inter(b) Decentralised architecture IoT applications are often deployed in constrained envioperability, scalability, and adaptability. In addition, as Fig. IoT applications are in often deployed in constrained constrained enviFig. 1.Centralised Types of architecture IoT architectures IoT applications are often deployed in envioperability, scalability, and adaptability. In addition, as Fig. 1. Types of IoT architectures ronments using low-cost devices with limited computing IoT applications are often deployed constrained environments using low-cost devices withinlimited limited computing ronments using low-cost devices with computing can take place. Since several forwarding devices Fig. 1. Types of IoT architectures making can place. Since IoT applications are often deployed in constrained enviand power powerusing resources, the execution execution of computationally computationally in- making making can take take place. Since several several forwarding forwarding devices devices ronments low-cost devices with limited computing and resources, the of inand power resources, the execution of computationally inmust be involved in the communication between the end making can take place. Since several forwarding devices must be involved in communication between end ronments using low-cost devices with limited computing tensive operations as well as the service life present further must be can involved in the the Since communication between the the end and power resources, the execution of computationally intensive operations as well as the service life present further making take place. several forwarding devices tensive operations as well as the service life present further devices and controller, centralised approaches often lead must be involved in the communication between the end devices and controller, centralised approaches often lead and power resources, the execution of computationally inchallenges. devices and controller, centralised approaches often lead tensive operations as well as the service life present further challenges. must beand involved in thecentralised communication between the load end challenges. to unnecessary unnecessary increase in the traffic traffic utilisation and load devices controller, approaches often lead to increase in the utilisation and tensive operations as well as the service life present further to unnecessary increase centralised in the traffic utilisationoften and load challenges. devices and controller, approaches lead A promising solution to improve IoT sustainability are IoT of resources. A recent study on IoT network architecin the on traffic load A promising solution solution to to improve improve IoT IoT sustainability sustainability are are IoT IoT to of resources. recent IoT network architecchallenges. of unnecessary resources. A Aincrease recent study study on IoTutilisation network and architecA promising unnecessary in the traffic and load integration platforms. IoT platforms provide the foundatures performed by Verma et al. has that of resources. Aincrease recent study IoTutilisation network architecA promisingplatforms. solution toIoT improve IoT sustainability are IoT to integration platforms. IoT platforms provide the the foundafoundatures performed by Verma et on al. (2017) (2017) has shown shown that tures performed by Verma et al. (2017) has shown that integration platforms provide A promising solution to improve IoT sustainability are IoT of resources. A recent study on IoT network architection for connecting devices to the Internet, acquiring the centralised solutions will face challenges in handling tures performed by Verma et al. (2017) has shown that integration platforms. IoT platforms provide the foundation for for connecting connecting devices devices to to the the Internet, Internet, acquiring acquiring the the centralised solutions will challenges in handling the the centralised solutions will face face in the tion integration platforms. IoT platforms provide the foundatures performed by Verma et challenges al. (2017) hashandling shown that generated data, and devices processing them in a meaningful meaningful way, increased number of IoT IoT devices and the growing growing demands centralised solutions will face challenges in handling the tion for connecting to the Internet, acquiring the generated data, and processing them in a way, increased number of devices and the demands increased number of IoT devices and the growing demands generated data, and processing them in a meaningful way, tionthey for connecting devices to the Internet, acquiring the increased centralised solutions willdevices face challenges in handling the i.e. offer a standardised way to manage the connected of users and services. number of IoT and the growing demands generated data, and processing them in a meaningful way, i.e. they offer a standardised way to manage the connected of users and services. of users and services. i.e. they offer a standardised waythem to manage the connected generated data, and processing in a meaningful way, increased number of IoT devices and the growing demands devices. of users and services. i.e. they offer a standardised way to manage the connected devices. devices. Edge computing driven IoT IoT platforms represent represent an alteralterEdge computing driven i.e. they offer a standardised way to manage the connected of users and services. Edge computing driven IoT platforms platforms represent an an alterdevices. The first wave of IoT platforms placed the control, opernative approach which moves part of the decision-making Edge computing driven IoT platforms represent an alterThe first wave of IoT platforms placed the control, opernative approach which moves part of the decision-making devices. approach which moves part of the decision-making The first wave of IoT platforms placed the control, oper- native Edge computing driven IoT platforms represent an ational and computational logic in a geographically cencapabilities from the cloud to the network edge, e.g gatenative approach which moves part of the decision-making The first wave of IoT platforms placed the control, operational and and computational computational logic logic in in aa geographically geographically cencen- capabilities capabilities from from the the cloud cloud to to the the network network edge, edge, e.g e.g altergategateational The first wave of IoT platforms placed the control, opernative approach which moves part of the decision-making tralised location, most commonly in the cloud (Fig. 1(a)). way, intermediate device, etc. (Fig. 1(b)). In contrast to capabilities from the cloud to the network edge, e.g gateational and computational logic in a geographically centralised location, location, most most commonly commonly in in the the cloud cloud (Fig. (Fig. 1(a)). 1(a)). way, way, intermediate intermediate device, device, etc. etc. (Fig. (Fig. 1(b)). 1(b)). In In contrast contrast to to tralised ational and logic Azure, in athe geographically cen- capabilities from the cloud toadvantage the network edge, e.g gateAmazon Webcomputational Services, Microsoft Azure, and IBM Bluemix centralised architectures, the advantage of distributing distributing the way, intermediate device, etc. (Fig. 1(b)). In contrast to tralised location, most commonly in cloud (Fig. 1(a)). Amazon Web Services, Microsoft and IBM Bluemix centralised architectures, the of the centralised architectures, the advantage of distributing the Amazon Web Services, Microsoft Azure, and IBM Bluemix tralised location, most Microsoft commonly in the and cloud (Fig. 1(a)). way, intermediate device,the etc.advantage (Fig. 1(b)). In contrast to are just a few examples of cloud platforms which found logic between the devices throughout the entire network is centralised architectures, of distributing the Amazon Web Services, Azure, IBM Bluemix are just a few examples of cloud platforms which found logic between the devices throughout the entire network is between the devices the throughout theofentire networkthe is are just Web a fewServices, examples of cloudAzure, platforms which found logic Amazon Microsoft and IBM Bluemix centralised architectures, advantage distributing their place soon in production. However, although centhe significant reduction of transferred data, which conlogic between the devices throughout the entire network is are just a few examples of cloud platforms which found their place place soon soon in in production. production. However, However, although although cencen- the significant reduction of transferred data, which conthe significant reduction of transferred data, which contheir are just a few examples of cloud platforms which found logic between the devices throughout the entire network is tralised architectures offer a a simplified simplified control and manmansequently leadsreduction to the the decrease decrease of communication communication delay the significant of transferred data, which contheir place soon in production. However, although centralised architectures offer control and sequently leads to of delay sequently leadsreduction to the decrease of communication delay tralised architectures offer a simplified control and mantheir place soon in production. However, although centhe significant of transferred data, which conagement,architectures in practice, practice, all all thea data data must be be forwarded to aa sequently and utilisation utilisation oftoresources resources including cloud services. services. IoT IoT leads of the decrease of communication tralised offer simplified control and management, in the must forwarded to including cloud and utilisation including cloud services.delay IoT agement, in practice, all the data must be forwarded to a and tralised controller architectures offer athe simplified control and mansequently leads of toresources the decentralised decrease of communication delay central located in cloud before any decision platforms based on the architecture also perand utilisation of resources including cloud services. IoT agement, in practice, all the data must be forwarded to a central controller located in the cloud before any decision platforms based on the decentralised architecture also perplatforms basedof onresources the decentralised architecture also percentral controller located in the cloud before any decision agement, in practice, all the data must be forwarded to a and utilisation including cloud services. IoT central controller located in the cloud before any decision platforms based on the decentralised architecture also percentral controller located in the cloud before any decision platforms based on the decentralised architecture also per2405-8963 © 2018, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Copyright © 2018 IFAC 168 Copyright 2018 168 Copyright © under 2018 IFAC IFAC 168 Control. Peer review© responsibility of International Federation of Automatic Copyright © 2018 IFAC 168 10.1016/j.ifacol.2018.07.148 Copyright © 2018 IFAC 168
2018 IFAC PDES Ostrava, Czech Republic, May 23-25, 2018 Jozef Mocnej et al. / IFAC PapersOnLine 51-6 (2018) 168–173
form better in environments in which a high number of connected devices communicating over an unreliable and constrained network produce a large amount of data. On the contrary, decentralised architectures bring a certain complexity into the management process. In consequence, they create a trade-off between the complexity of management and efficient utilisation of resources. However, as communication complexity challenges can be overcome with an appropriate architecture design, to address sustainability related challenges, we have decided to shift away from traditional approaches and propose a decentralised IoT architecture. This architecture is based on considerations, such as, the method for runtime monitoring including available resources and utilisation, operational requirements and costs, as well as, system adaptation to prolong the service life of battery-powered devices. The contribution of this paper is threefold. First, we identify the optimal set of features required by a generalised form of decentralised IoT platforms. Second, we describe a novel approach for efficient resource utilisation. Third, we specify the structure of the overall architecture. The rest of this paper is organised as follows. Section 2 describes the motivation and related work. Section 3 presents the optimal feature set for decentralised IoT platforms. Section 4 introduces a concept for achieving efficient utilisation of available resources, whereas a proposal of a decentralised IoT architecture is analysed in Section 5. Finally, Section 6 draws a conclusion and provides directions for future work. 2. MOTIVATION AND RELATED WORK In a real-world environment, several factors influence IoT sustainability, including the heterogeneity and relocatability (e.g. static or mobile) of the devices, topology interconnecting the nodes, and condition of the environment. These factors are all present in the most common IoT application domains starting from smart manufacturing and retail (Lojka et al. (2015)), through smart buildings and living (Mekki et al. (2017)), smart agriculture (Paraforos et al. (2016)), up to smart city (Djahel et al. (2015)). An equally important factor is the type of energy source used to power the devices. Currently, either mains, batteries, renewable energy sources or the combination thereof are used. Among them, IoT devices are usually batterypowered (Somov and Giaffreda (2015)). Since resource utilisation, including power consumption, is a major concern in current solutions, it becomes essential to efficiently utilise the power sources in order to extend their lifespan. However, determining the optimal configuration is a challenge itself. Network traffic characteristics as well as the environment can change rapidly over time depending on various factors such as reconfigurations, failures in the topology, the time of the day, whether is there an anomaly or only standard data transmissions are performed. Static or manual configuration becomes unfeasible due to the continuously changing conditions. A decentralised IoT platform with dynamic decision making and intelligent (re)configuration placed at the edge represent a paradigm which could be a step forward in utilising efficiently the available resources. 169
169
However, although improvements can be hypothetically achieved, as the number of the variables that need to be considered rises, the complexity and computing demands of the entire system are increasing as well. As a result, this can similarly lead to efficiency drop in resource utilisation. Considering the above mentioned, our targeted research question can be formulated as follows: What is the (1) optimal decentralised IoT architecture design suitable for (2) obtaining, selecting and using information to (3) achieve improvements in resource utilisation? The architecture design of IoT platforms has been discussed by several works. Sarkar et al. (2014) proposed a scalable distributed architecture tackling characteristics like heterogeneity, scalability, interoperability, automation, zero-configuration and distributiveness. They introduced a three-layered architecture design aimed at supporting multi-application use cases with minimum management effort. This was achieved using distributed intelligence empowered devices which process the data and make decisions at the edge without the need for centralised logic in the cloud. However, as centralised computing can be beneficial in circumstances when, for example, further analysis or accounting is required, abandoning the cloud entirely should be avoided. In addition, the problem of energy efficiency was not addressed. Datta et al. (2014) introduced an IoT architecture with a wireless gateway as the main intermediate device responsible for interactions between the sensors (actuators) and mobile clients. This gateway-centric approach consists of south and north interfaces that use generated metadata to expose the capability of the devices to mobile clients. A similar effort by Vallati et al. (2016) proposed a distributed architecture also utilising gateways. They claim that their platform implementation called BETaaS simplifies the deployment of horizontal solutions by exposing a unified service oriented interface to access things. Although these proposals discuss viable architecture design ideas including distributed logic, scalability, and interoperability, they fall short of addressing the challenges of resource utilisation. Likewise, energy efficiency plays a vital role in the lifespan of IoT devices, and therefore should deserve more attention. Yet, there are only a few efforts targeted at energy efficiency. Rault et al. (2014) developed various mechanisms for improving resource efficiency, and Liu et al. (2014) focused on both ensuring Quality-of-Information (QoI) as well as energy efficiency in sensory environments. However, these consider only a small number of factors (characteristics) influencing the sustainability of IoT. In contrast to the above-mentioned research, energy efficiency and resource utilisation are the key emphasis in this paper. In addition, we also reevaluate the expectations of decentralised IoT platforms and extend the optimal feature set with additional features such as multi-network approach and simplified device management (both central and individual).
2018 IFAC PDES 170 Ostrava, Czech Republic, May 23-25, 2018 Jozef Mocnej et al. / IFAC PapersOnLine 51-6 (2018) 168–173
3. OPTIMAL FEATURE SET FOR DECENTRALISED IOT PLATFORMS Centralised architectures often perform inefficiently in present IoT environments. Efficient data processing, decision making and resource utilisation require new approaches such as a decentralised IoT platform. Obviously, to ensure its sustainability, an appropriate design is essential but itself a challenge, let alone an optimal design. As a step forward, we have identified the network traffic characteristics of the most common IoT services, categorized into eight application domains as per Mocnej et al. (2018). In particular, we examined smart buildings and living, smart healthcare, smart environment, smart city, smart energy, smart transport and mobility, smart manufacturing and retail, and smart agriculture from the perspectives of three user groups, individuals, society, and industry. One observation deduced from this study is a set of features that a decentralised IoT architecture of the future should meet. These features are as follow: Feature 1. Multi-network approach – an IoT platform is destined to connect devices regardless of the underlying network technology in use. IoT is diverse and since different constraints may require different solutions, many network technologies exist designed specifically for different parts of the IoT. Ideally, an IoT platform should enable communication with all heterogeneous devices in a standardised way by hiding the underlying network technologies. Feature 2. Scalable and interoperable implementation – the shift from application-specific to application-independent solutions is highly desired. As the conditions and expectations of IoT application domains may dynamically change over time, an adaptive IoT architecture should recognise and react to these changes in a timely manner. In conclusion, an IoT platform of the future needs to be scalable and non-invasive, and ready to integrate new applications in any circumstances. Feature 3. Low power consumption – low power consumption and efficient utilisation of resources must be a priority to achieve the desired sustainability of the solution. An IoT platform should therefore support mechanisms such as event-driven M2M (Verma et al.) communication and other optimisation methods to achieve improvements in energy utilisation. Feature 4. Intuitive data and device management – to keep pace with the rapid development and innovation, efficient device control and data collection should be present in IoT by design. A platform should support both individual and central management for accessibility, control and modifications. Furthermore, many routines should be automated to make the configuration process simpler and more intuitive. Feature 5. Artificial Intelligence at the edge – Artificial Intelligence (AI) and Machine Learning (ML) are gaining on popularity year by year. Today, they are utilised in the vast majority of technologies and services. IoT is not an exception where AI is most commonly utilised in the cloud to provide a number of improvements. However, in the case of a decentralised architecture, part of the services should be moved from the cloud to the network edge including AI. The most suitable device for hosting AI is an 170
unconstrained gateway which offers reasonable computing resources and sustained energy source. A decentralised IoT platform should be also designed to support contextawareness. These five key features have been carefully incorporated into the architecture of our proposed decentralised solution. 4. A CONCEPT FOR ACHIEVING EFFICIENT RESOURCE UTILISATION Our objective is to achieve improvements in the utilisation of available resources. Available resources denote connected devices providing sensing/actuating services to the IoT platform. How these resources are utilised depend on the applications implemented in the platform and their requirements (expectations). Since the requirements may change over time (which subsequently affects the resource utilisation), the individual changes have to be tracked and recorded over time. This requires: (i) Monitoring the platform in runtime to obtain insight into the system and its processes; (ii) Ensuring the desired quality of the output by taking the appropriate measures; an example output would be optimal subset of active resources for a particular scenario; (iii) Optimising the utilisation of available resources. In the following, these three research topics will be discussed in more details. In general, tasks (i)–(iii) represent our main research topics. 4.1 Monitoring IoT Platforms in Runtime To monitor IoT platforms in runtime, a number of measures can be used: • Quality of Device (QoD) – is responsible for expressing the quality of a particular device and can monitor various attributes, including, but not limited to, battery life, precision, and sending rate (Buchholz and Schiffers (2003)). • Quality of Service (QoS) – uses attributes such as bandwidth, delay, jitter, and packet loss to monitor the performance characteristics of the network (Bhuyan et al. (2010)). • Quality of Information (QoI) – characterises the fitness-for-use of data, and therefore is suitable for measuring the quality of the obtained data from a particular device but can also define the quality of the output by using attributes like accuracy, precision, and freshness (Bisdikian et al. (2013)). Fig. 2 depicts an IoT integration scenario of these measures, i.e. QoD for end device monitoring, QoS for network performance monitoring, and QoI for transferred data quality monitoring is used. One advantage of these measures is that, although they are principally destined to describe a specific part of a system, when used together, they can also provide performance overview of the entire system.
2018 IFAC PDES Ostrava, Czech Republic, May 23-25, 2018 Jozef Mocnej et al. / IFAC PapersOnLine 51-6 (2018) 168–173
171
ticularly highlight the significance of the following three power plan modes to achieve optimal energy efficiency: (1) OFF; (2) ON and sensing; and (3) ON and transmitting. 5. DECENTRALISED IOT ARCHITECTURE
Fig. 2. Implementation of measures in an IoT integration scenario. 4.2 Ensuring the Desired Output Quality
Our decentralised IoT architecture is composed of three main functional blocks: end device, gateway and cloud. We provide a stack diagram for the end device and gateway blocks which outline the modules and their function. 5.1 End Device
The applications determining the requirements (expectations) for the resources are communicating through gateway(s) where each application has its own stream processing unit as per Fig. 2. In general, both the requirements and the capabilities of the resources can be expressed as functions with QoI attributes. To ensure the desired output quality the correlation between these functions have to be identified. To achieve this, we define a further measure: • Value Of Information (VoI) – deals with an assessment of the information utility for a specific use case scenario using attributes such as relevance, integrity, timeliness, and understandability (Bisdikian et al. (2013)). From a decentralised architecture perspective, we can utilise this measure to evaluate the significance of resources for an application. More specifically, by leveraging VoI a subset of optimal resources can be selected for every stream processing module. This provides an option to determine what constitutes to the desired output quality and under what conditions this quality can be achieved. 4.3 Optimising the Utilisation of Available Resources Assuming Section 4.1 and Section 4.2 are achieved, the utilisation of available resources can be optimised. Since the majority of IoT devices is constrained (e.g. by their battery lives), it is preferred to utilise them efficiently. Keeping the transceiver powered on of a device is an energy expensive task, as shown by Harrison et al. (2016). Therefore, one approach to extend the lifespan is to power off those battery-powered devices which generate duplicate data or whose absence is acceptable by any stream processing. In general, this represents an optimisation problem of finding a subset of active devices that would minimise the overall consumption of limited energy resources in the network while satisfying at the same time the requirement of minimal output quality. Another approach is to predict the sensed value of a device rather than powering it on to execute its task(s). This subsequently leads to the minimisation of communication between the devices and the gateways, extending the lifespan. Prediction could be realised as long as QoI of a device satisfies the minimum requirements, which could be specified by the stream processing modules and affected by the dynamics of the environment. Consequently, our approach consists of duty-cycling and adaptive sampling for controlling the power plans and sending rates of devices to conserve their energy. We par171
The computing capabilities of end devices can be diverse. This fact was taken into consideration when the threelayered architecture depicted in Fig. 3 was proposed. The description of the individual layers is as follows: • Connectivity abstraction layer – is responsible for connecting a device to a network. The layer includes a single module called Connectivity SDK that hides the underlying network technology and provides the standardised communication capabilities for the upper layers. • Device services layer – defines services necessary for each device to efficiently utilise its resources. This layer contains four modules: · data filtering – is responsible for a very simple and lightweight selection of important data from all sensed values. · communication service – uses communication capabilities from Connectivity SDK and wraps them to the publicly accessible functions. It also provides a way to remotely control the services of devices. · power plan – manages how a device uses power to achieve Feature 3. By default, it supports at least three power plans - ON, ON with a transmitter turned off, and OFF. · Over-The-Air (OTA) – permits updates over the air to always keep a device’s software up-to-date. • Custom application layer – for a device serving a particular task. It can also enhance the supported functionalities if a device has enough computing resources. An application is installed on a device during its manufacturing process but can be later modified by the OTA module. 5.2 Gateway Since a gateway usually has unconstrained power resources, it can be utilised to support computing-intensive tasks to minimise the load (e.g during decision making)
Fig. 3. End device architecture.
2018 IFAC PDES 172 Ostrava, Czech Republic, May 23-25, 2018 Jozef Mocnej et al. / IFAC PapersOnLine 51-6 (2018) 168–173
of other IoT devices. The architecture showed in Fig. 4 contains three main functionalities: connectivity, end device management, and data management. Connectivity is an essential functionality of the gateway provided by the lower layers of the architecture. To ensure the conformity of our solution with Feature 1, in the connectivity abstraction layer we implement public and well-known IoT specifications such as those defined by OCF 1 or OneM2M 2 . However, our approach is not limited to only these standards. Custom communication protocols can be implemented using a designated custom module. All connectivity abstraction modules are unified by the network manager, which is responsible for the interactions between the services residing at higher layers and connectivity functionalities at the lower layers. End device and data management functionalities are provided by the gateway services layer. In general, this layer is responsible for providing Feature 5. When data from a network arrives at the inbound interface of a gateway, it is processed by the data services module. This module provides functionalities such as QoI measurement and prediction. The obtained data are subsequently transferred to the stream processing module which is the pipeline for various custom applications. Data selection by these applications from the pipeline is performed based on the significance level expressed by VoI. When the appropriate input is determined, applications continue in the execution of their tasks. Application outputs can be further processed in a number of ways. One option is to pass the data to the end devices or another gateway directly. Another option is to store the output in a temporary database and forward it later (if real-time forwarding in not necessary) to, for example, the cloud. The interaction between the individual blocks and modules including the type of communication and measurement method is depicted in Fig. 5. An important component of loosely coupled systems is service orchestration. It provides an efficient way to define functionalities as small independent services which are added to a central list and linked together. This approach is more efficient than presenting all functionalities as one complex application since this way services can be dynamically modified over 1 2
time, ensuring Feature 2. Service orchestration based on service-oriented architecture (SOA) is implemented in our solution at the service orchestration layer. Lastly, we offer the options for gateway communication: (1) via a user-friendly GUI and/or (2) remote connection using the exposed API. This way, both simple as well as advanced gateway control is achieved, ensuring the satisfaction of Feature 4. 5.3 The Cloud Due to the number of services offered by providers, cloud solutions present an advantageous foundation for IoT solutions and cloud-based IoT solutions have already been proposed (Zhou et al. (2013); Hou et al. (2016)). From the perspective of our architecture, one important service we would like add is network management. The network management service module provides control of the entire network from a centralised place. It utilises the acquired metadata to generate the scheme of the network and subsequently provides the user with a way to execute configuration changes and updates without the need for connecting to every single gateway individually. Its main advantage is to deliver a solution using a system which can be configured once and then deployed everywhere. This intuitive data and device management at a centralised place satisfies Feature 4. 6. CONCLUSION This paper addresses the problem of achieving sustainability in diverse IoT platforms. Although the current wave of IoT platforms is based on centralised architectures, we outlined the benefits of decentralised architectures, especially when deployed in a complex real world environment. The main contribution of this paper is the concept of a decentralised IoT architecture. This architecture incorporates five key features a decentralised IoT platform of the future should meet. Based on these features we proposed a solution intended to optimise the resource utilisation of IoT components. Lastly, we also presented the functional blocks of our proposed approach along with some key modules and features. Future work will be aimed at the optimisation process for improving resource utilisation. To achieved this, the individual research topics (see Section 4 have to be investigated in greater details as each of them reveals further open issues and presents new challenges. Future work will also be focused on the development of a prototype based on the proposed decentralised IoT platform that will be used for evaluating the methods and procedures proposed to address IoT challenges.
https://openconnectivity.org http://www.onem2m.org
ACKNOWLEDGEMENTS This publication is prepared within the grant VEGA 1/0663/17, 2017-2020 (100%). Jozef Mocnej would like to thank the European Commission under the Thelxinoe project for the funding that supported his visit to Victoria University of Wellington (VUW), New Zealand. Adrian Pekar and Winston K.G. Seah are supported in part by VUW’s Huawei NZ Research Programme, SoftwareDefined Green Internet of Things project #E2881.
Fig. 4. Gateway architecture 172
2018 IFAC PDES Ostrava, Czech Republic, May 23-25, 2018 Jozef Mocnej et al. / IFAC PapersOnLine 51-6 (2018) 168–173
173
Fig. 5. Interactions between the functional blocks and modules. REFERENCES Bhaskar Bhuyan et al. Quality of service (QoS) provisions in wireless sensor networks and related challenges. Wireless Sensor Network, 2(11):861, 2010. Chatschik Bisdikian, Lance M Kaplan, and Mani B Srivastava. On the quality and value of information in sensor networks. ACM Transactions on Sensor Networks (TOSN), 9(4):48, 2013. Thomas Buchholz and Michael Schiffers. Quality of context: What it is and why we need it. In Proceedings of the 10th Workshop of the OpenView University Association: OVUA’03, 2003. Soumya Kanti Datta, Christian Bonnet, and Navid Nikaein. An IoT gateway centric architecture to provide novel M2M services. In Proceedings of the IEEE World Forum on Internet of Things (WF-IoT), pages 514–519, Seoul, South Korea, 6-8 March 2014. Soufiene Djahel, Ronan Doolan, Gabriel-Miro Muntean, and John Murphy. A communications-oriented perspective on traffic management systems for smart cities: challenges and innovative approaches. IEEE Communications Surveys & Tutorials, 17(1):125–151, 2015. DC Harrison, D Burmester, Winston K. G. Seah, and R Rayudu. Busting myths of energy models for wireless sensor networks. Electronics Letters, 52(16):1412–1414, 2016. L. Hou, S. Zhao, X. Xiong, K. Zheng, P. Chatzimisios, M. S. Hossain, and W. Xiang. Internet of things cloud: Architecture and implementation. IEEE Communications Magazine, 54(12):32–39, December 2016. ISSN 0163-6804. C. H. Liu, J. Fan, J. W. Branch, and K. K. Leung. Toward QoI and Energy-Efficiency in Internet-of-Things Sensory Environments. IEEE Transactions on Emerging Topics in Computing, 2(4):473–487, Dec 2014. ISSN 2168-6750. Tomas Lojka, Marek Bundzel, and Iveta Zolotova. Industrial gateway for data acquisition and remote control. Acta Electrotechnica et Informatica, 15(2):43–48, 2015. Kais Mekki, William Derigent, Eric Rondeau, and Andr´e Thomas. Data lifecycle management in smart building using wireless sensors networks. IFAC-PapersOnLine, 50(1):12944–12949, 2017. Jozef Mocnej, Adrian Pekar, Winston K. G. Seah, and Iveta Zolotova. Network Traffic Characteristics 173
of the IoT Application Use Cases. Technical Report ECSTR18-01, School of Engineering and Computer Science, Victoria University of Wellington, 6 Jan 2018. URL https://ecs.victoria.ac.nz/Main/ TechnicalReportSeries. Dimitris S Paraforos, Vangelis Vassiliadis, Dietrich Kortenbruck, Kostas Stamkopoulos, Vasileios Ziogas, Athanasios A Sapounas, and Hans W Griepentrog. A farm management information system using future internet technologies. IFAC-PapersOnLine, 49(16):324– 329, 2016. Tifenn Rault, Abdelmadjid Bouabdallah, and Yacine Challal. Energy efficiency in wireless sensor networks: A top-down survey. Computer Networks, 67:104–122, 2014. Chayan Sarkar, SN Akshay Uttama Nambi, R Venkatesha Prasad, and Abdur Rahim. A scalable distributed architecture towards unifying iot applications. In Proceedings of the IEEE World Forum on Internet of Things (WFIoT), pages 508–513, Seoul, South Korea, 6-8 March 2014. Andrey Somov and Raffaele Giaffreda. Powering IoT Devices: Technologies and Opportunities. Internet of Things Newsletter, 9 November 2015. Carlo Vallati et al. BETaaS: A platform for development and execution of machine-to-machine applications in the Internet of Things. Wireless Personal Communications, 87(3):1071–1091, 2016. Pawan Kumar Verma et al. Machine-to-Machine (M2M) communications: A survey. Journal of Network and Computer Applications, pages 83 – 105. ISSN 1084-8045. S. Verma, Y. Kawamoto, Z. M. Fadlullah, H. Nishiyama, and N. Kato. A survey on network methodologies for real-time analytics of massive iot data and open research issues. IEEE Communications Surveys & Tutorials, 19 (3):1457–1477, thirdquarter 2017. J. Zhou, T. Lepp¨anen, E. Harjula, M. Ylianttila, T. Ojala, C. Yu, and H. Jin. Cloudthings: A common architecture for integrating the internet of things with cloud computing. In Proceedings of the 2013 IEEE 17th International Conference on Computer Supported Cooperative Work in Design (CSCWD), pages 651–657, Whistler, BC, Canada, 27-29 June 2013.