A Fog-enabled Smart Home Analytics Platform

Theo Zschörnig

, Robert Wehlitz

and Bogdan Franczyk

2,3

Institute for Applied Informatics (InfAI), Leipzig University, Goerdelerring 9, 04109 Leipzig, Germany

Information Systems Institute, Leipzig University, Grimmaische Str. 12, 04109 Leipzig, Germany

Business Informatics Institute, Wrocław University of Economics, ul. Komandorska 118-120, 53-345 Wrocław, Poland

Keywords: Smart Home, Fog Computing, Internet of Things, Analytics Architecture.

Abstract: Although the usage of smart home devices such as smart speakers, light bulbs and thermostats has increased

rapidly in the past years, their added value, compared to conventional devices, is mostly limited to simple

control and automation logic. In order to provide adaptive smart home environments, it is necessary to gain

deeper insights into the data generated by these devices and use it in sophisticated data processing pipelines.

Providing such analytics to a multitude of consumers requires specialised architectures, which are able to

overcome various challenges identified by scientific literature. Currently available smart home analytics

architectures are not designed to tackle all of these issues, specifically fault-tolerance, network-usage, latency

and external regulations. In this paper, we propose an architectural solution to address these challenges based

on the concept of Fog computing. Furthermore, we provide insight into the motivation for this research as

well as an overview of the current state of the art in this field.

1 INTRODUCTION

In the past years, Internet of Things (IoT)

technologies and solutions have been adopted in a

variety of domains for personal and business use. In

the year 2018, the number of connected Internet of

Things devices worldwide has already risen to 16.8

billion in communications, 5.4 billion in commercial

and industrial electronics as well as 5.9 billion in

terms of consumer devices (IHS, 2018). All these

devices are sources of data, which provide their users,

but also businesses, industry and researchers, with the

opportunity to gain valuable insights into everyday

life and industrial value creation. Furthermore, state

of the art analytical algorithms and methods play a

key role in achieving an even deeper understanding

of the use of IoT devices and their surroundings,

hence further increasing their added value.

In this regard, a major challenge for businesses is

to provide their customers with technical solutions to

utilize the full potential of their IoT devices in terms

of data insight and value-added information. In this

context, the solutions, offered by businesses and

researchers alike, mostly involve Big Data

technologies, embedded into cloud platforms, which

offer an abundance of processing and storage

resources. While this seems appropriate for a variety

of scenarios, especially IoT environments

characterized by sensitive data, such as smart home

ecosystems, expose additional requirements

regarding data security, fault tolerance and latency,

but at same time ease of use, therefore creating the

need for new solutions regarding analytical

architectures. In order to tackle these challenges, we

present an IoT analytics platform designed for smart

home environments based around the Fog computing

paradigm.

In this paper, we motivate our research as well as

the technical challenges to be addressed by our

solution proposal and the opportunities it provides

(Section 2). We give an overview of the state of the

art in IoT analytics regarding technologies and

architectures. In addition, we show that these are not

fully suitable in terms of smart home analytics

(Section 3). The main contribution of this paper, an

approach to build an analytics platform architecture

to be used in smart home environments, is described

in Section 4. In conclusion, we provide ideas to

further the research in this field as well as our own

(Section 5).

2 MOTIVATION

It is estimated that the number of annually sold smart

home devices will grow to 939 million in the year

616

Zschörnig, T., Wehlitz, R. and Franczyk, B.

A Fog-enabled Smart Home Analytics Platform.

DOI: 10.5220/0007750006160622

In Proceedings of the 21st International Conference on Enterprise Information Systems (ICEIS 2019), pages 616-622

ISBN: 978-989-758-372-8

2022 (IDC, 2018). Consumers anticipate benefits

from the use of these smart home technologies mainly

in the areas of energy management, home automation

and control as well as home security (Wilson et al.,

2017). In this context, a major part in the creation of

added value from smart devices is the ability to

generate insights from the data they are collecting.

While research focuses on developing new and

applying existing methods of data processing on

smart home environments (Brush et al., 2018),

current data processing and analytics architectures

evolve around cloud-based Big Data solutions.

These architectures aim to resolve a variety of

different challenges for IoT analytics in general,

already identified in scientific literature:

 Limited resources at IoT devices

(Stolpe et al., 2016)

 High data volume, heterogeneity and velocity

(Chen et al., 2015; Stolpe et al., 2016; Marjani

et al., 2017)

 Occasional connection loss of IoT devices

(Chen et al., 2015; Rozik et al., 2016)

 Real-time data processing

(Chen et al., 2015; Zaslavsky et al., 2015;

Stolpe et al., 2016)

 Personalized analytics

(Biswas et al., 2014; Auger et al., 2017)

 Data security and privacy

(Zaslavsky et al., 2015; Stolpe et al., 2016)

Analytical architectures in common smart home

scenarios need to be able to address these

requirements. In contrast to other fields of application

of the IoT, they also need to offer capabilities to

handle long-running, rather static, Big Data problems

as well as smaller, more intimate analytical problems,

which are often changed in terms of data sources and

requirements. For example, the training of outlier

detection or non-intrusive load monitoring algorithms

based on large datasets for energy management, are

as important as small processing tasks of a few data

sources, e.g. temperature tracking of a single

thermostat sensor. At a technical level, this causes the

necessity for an analytics architecture, which is

scalable and elastic while at the same time providing

fast and flexible ways to change and extend analytical

processing. In addition, because of possible

connection loses, data processing needs to be fault-

tolerant at both the cloud and the local level to ensure

the ongoing operation of the smart home. With time

critical use cases such as home security and disaster

detection and prevention, being an essential

component of smart home expectations by end-

consumers (Brush et al., 2018), low latency is key in

terms of data processing. Another important aspect of

analytics architectures is data security and privacy.

Especially with the introduction of the General Data

Protection Regulation (GDPR) in the European

Union, processing of sensitive data, as it is the case in

most smart home environments, needs to be in

accordance to legislative regulation.

Looking at current smart home analytics

architectures, these requirements

seem to have only been insufficiently met. Especially

in terms of fault-tolerance, latency and data privacy

cloud-only data processing solutions seem to be ill

equipped for the aforementioned requirements. In

contrast, fog computing provides a promising

approach to address these issues. The term “fog

computing” was first used by CISCO in 2012 and

describes a concept where data processing

capabilities are moved from the center of the cloud to

the edge of the network. It is therefore an extension

of the cloud computing paradigm and fog components

cannot stand alone. (Mouradian et al., 2018)

In general, the fog computing paradigm offers

several advantages over cloud computing in terms of

reduced latency and network usage, higher fault-

tolerance, resource availability at the source of data

and data processing in compliance with specific

legislation (Byers, 2017; Klonoff, 2017; Ravindra et

al., 2017; Velasquez et al., 2017). In terms of smart

home analytics, enabling local networks of smart

devices to process their data cloud-independent

ensures the continuous operation of all processes and

tasks in case of connectivity issues to cloud services.

Moreover, offloading processing and analytics tasks

to local processing nodes further reduces processing

latency. Furthermore, reducing the volume of data

send in-between and from IoT devices to cloud

services will decrease latency issues even more

(Stojkoska & Trivodaliev, 2017). In addition, local

data processing addresses concerns regarding data

privacy and security. This involves anonymization at

the location of data generation as well as reduced

transmissions of sensitive user data to cloud services.

3 STATE OF THE ART

Smart home networks typically include various IoT

devices such as smart thermostats, light bulbs,

speakers, locks, but also voice control devices and

cameras. These devices are connected either directly

to their respective cloud backend or via a central

entity, a so-called “gateway”, utilizing IoT-based

communication protocols, such as Z-Wave, ZigBee,

etc. The cloud backend services are used to control

and access IoT devices remotely and, furthermore,

A Fog-enabled Smart Home Analytics Platform

617

most of the data processing of IoT devices is done via

the underlying cloud infrastructure after transferring

the required data. Especially large internet companies

like Amazon rely solely on cloud-based services for

data processing (Amazon, 2019).

Looking at the scientific literature regarding IoT

analytics architectures, dedicated smart home

solutions are rare. They focus on specific topics such

as device security (Haddadi et al., 2018) or energy

management (Al-Ali et al., 2017) and offer cloud-

based solutions for data processing with data

collection nodes at a local level. The area of

application of these works is limited to specific use

cases, hence missing the needed flexibility to react to

changing requirements of smart home environments

as described in Section 2.

Other domains of application have yielded

additional approaches to IoT analytics architectures.

Several publications in the field of industrial IoT

(IIoT) propose fog computing-based architectures

(Rehman et al., 2018; Alexopoulos et al., 2018) for

data processing and analytics. Although, these

approaches are able to address most of the

requirements of IoT analytics in general, they are

specifically designed around the processes and roles

of industrial manufacturing. It is therefore

questionable if their deployment in smart home

environments is possible without extensive

adjustments.

There has also been research into analytics

architectures in Smart City scenarios. These works

usually offer cloud-based Big Data solutions (Cheng

et al., 2015; Ta-Shma, 2018) and are therefore not

well suited for smart home scenarios because of the

requirements described in Section 2.

General fog computing architectures for IoT use

cases are described in Alturki et al. (2017) and

Ravindra et al. (2017). They offer valuable insights

into fog architectures regarding important

components and data flow modelling. Nevertheless,

their experimental setups are rather static and are

hence missing flexibility in data processing locations

as well as use case adoption.

None of the analysed related works is able to

provide full coverage of all challenges for analytics

architectures in smart home environments. Hence,

this paper aims to

provide an architectural approach to fill this gap.

4 SOLUTION PROPOSAL

In order to provide sophisticated data manipulation,

analytics and persistence in a flexible manner for

smart home environments, we propose the following

architectural solution. This proposal is based on the

previous works of the authors which has been adapted

to reflect the changed requirements in smart home

environments as mentioned before. In this regard, the

formerly presented architectural proposal was

extended to include a fog layer, which aims to address

the challenges described in section 2. The main

purpose of this layer is to allow for offloading data

processing tasks from the cloud layer. The complete

architecture is shown in Figure 1.

4.1 Cloud Layer

The cloud layer is based on the concept of the Kappa

architecture, which treats all data as a stream and

drops the batch layer of the more traditional Lambda

architecture in favour of only a speed and serving

layer (Kreps, 2014). This allows for flexible data

processing while still providing capabilities to handle

large, high velocity amounts of data from various

sources. The data processing and analytics are

executed by the stream processing system. This

system includes a log data store for handling data

streams and additionally container-based

microservices, which represent individual processing

tasks of analytics pipelines.

The core component to orchestrate the processing

jobs is the flow engine, which executes analytics

flows. These flows are typically designed by power

users and may be accessed and deployed by regular

users as well. The execution of analytics flows is

triggered directly at the flow engine by either API

access, frontend or mobile applications.

Surrounding the flow engine are auxiliary services

providing data and metadata for analytics flow

execution. In this regard, the flow repository is the

central point for storing analytics flows. These flows

employ a simple flow-chart methodology with nodes

and edges, in which nodes represent tasks of data

manipulation or analytics, so called operators, and the

edges represent the data flow between nodes. The

capabilities of operators as well as additional

metadata are stored in the operator repository. This

information is used to design analytics flows, but also

by the flow parser to check analytics flow validity

before execution. A running instance of an analytics

flow is an analytics pipeline and registered in the

pipeline registry by the flow engine.

In addition to these, already by Zschörnig et al.

(2017) established, components of the cloud layer, we

introduced a reasoner. Its main purpose is to decide

whether analytics operators may be offloaded to the

fog layer of the overall architecture. Therefore, a set

ICEIS 2019 - 21st International Conference on Enterprise Information Systems

618

Figure 1: Proposed architecture with data and control flows, adapted from Zschörnig et al. (2017).

of rules and conditions needs to be in place to support

these decisions. The conditions have to be designed

in a way that allows to conclude hard (MUST and

MUST NOT) and soft (SHOULD and SHOULD

NOT) scheduling decisions for the flow engine. An

example for a condition could be: “Sensitive data

used  MUST be deployed locally”. Following the

scheduling rules, the flow engine either requests

operator deployment in the cloud stream processing

system or issues an operator request to the fog layer.

To ensure fault-tolerance requests to the fog layer are

decoupled using a message queue.

The cloud layer by itself already provides a wide

range of capabilities to address the challenges of IoT

analytics architectures. In terms of smart home

analytics, fault-tolerance, low latency and data

privacy issues remain. Therefore, the fog layer

specifically addresses these issues.

4.2 Fog Layer

The design of the fog layer components and its

integration in the existing architectural proposal is

based on the challenges to be addressed by fog

computing architectures as described in scientific

literature.

Fog resources are mobile as well as added and

removed in an unpredictable manner (Dastjerdi &

Buyya, 2016; Byers, 2017; Velasquez et al., 2017),

therefore, a fog architecture has to provide

mechanisms to cope with the resulting insecurities

concerning connection stability and data flow

volatility. In addition, the processing of data is

supposed to be done in the proximity of its

generation, requiring optimal orchestration of

processing tasks (Byers, 2017; Ravindra et al., 2017)

while taking into account the limited resources of fog

devices (Dastjerdi & Buyya, 2016). Beyond that, fog

architectures need to handle a large variety of

different use cases (Byers, 2017) and provide the

means for secure data transmission (Dastjerdi &

Buyya, 2016; Velasquez et al., 2017). Literature

regarding this topic also mentions a high number of

fog nodes, which need to be managed by the overall

architecture (Mouradian et al., 2018), which requires

scalable components for orchestration and

Fog layer

Cloud Layer

Fog Node

Stream Processing System

Log Data Store

Serving layer

IoT data stream

Serving DB

queried

data

job version

n+1

(operator)

job

version n

(operator)

processed

stream

data

Sensoren

IoT Device

Pipeline Registry Flow Engine

Operator

Repository

request operator

deployment

IoT Service

raw

IoT

data

Data Lake

job version

(operator)

stream data

Data API

Query

Visualize

job version n

(operator)

job version

n+1

(operator)

IoT stream data

Fog Master

Fog Agent

Registry

Job Registry

Fog Agent

Ressource

Manager

Monitoring

Service

Operator

Manager

Flow

Repository

request operator

metadata

pipeline

request flow

metadata

pull operator request

start

operator

raw

IoT

data

job data

send operator request

check agent

status

Flow Parser

parse flow

Reasoner

request deployment priorities

Operator Request

Message Queue

push operator request

A Fog-enabled Smart Home Analytics Platform

619

communication. Probably the most important

challenge is the support of real-time data processing

and analytics under the assumption of external

constraints. These include limited and heterogeneous

resources of fog devices along with organisational or

judicial regulations (Dastjerdi & Buyya, 2016;

Ravindra et al., 2017; Velasquez et al., 2017;

Mouradian et al., 2018).

The fog layer comprises all components of the

architecture, which are not deployed in the cloud but

rather at a local hardware level such as IoT gateways.

These fog nodes are not as mobile as edge devices and

therefore more reliable in terms of connectivity and

availability. Still, architectural components need to be

able to run independently from the cloud layer to

ensure the ongoing usefulness of smart devices in

smart home scenarios. In order to achieve this, the

proposed solution is based on similar approaches like

Brito et al. (2017) and employs two main

components, the fog master (FM) and the fog agent

(FA). Every fog node, which is used to execute

analytics operators, needs to have a FA deployed.

This component is able to manage and monitor

hardware resources as well as analytics operators. In

addition, at least one fog node has to deploy the FM

component. Its main tasks are to register all available

FAs and to orchestrate operator deployments at the

fog layer. The FM component may be deployed at

multiple fog nodes thus providing additional fault

tolerance. In this scenario, all decisions are made via

a quorum of all FMs. During the registration process

of a FA, their available resources and location are

registered as well. This is used by the FM to

determine if an operator request may be executed,

taking into account the external constraints of an

analytics flow.

The FM constantly pulls operator deployment

requests from the corresponding message queue of

the flow engine. When an operator request is received,

the FM checks its own FA registry, if the operator

request can be satisfied using the available FAs. The

status of all known FAs is continuously checked by

the FM. This includes their overall health as well as

available resources.

If a FA is available for operator deployment, the

FM sends it a request containing the necessary

metadata to start the operator. The FM prioritizes

FAs, which are near the source of the data to be

processed. The information regarding the data source

location is relayed along with the operator request

from the flow engine. In addition, a new entry in the

job registry of the FM is created. This ensures that in

the case of an offline FA, all its jobs may be

reassigned to other FAs to ensure fault-tolerance. In

the case of no available FA for operator deployment,

the FM informs the flow engine, which either cancels

the flow execution or tries to deploy the operator at

the cloud layer.

A FA comprises three components to handle

operator deployment. The resource manager checks

for available hardware resources and the current load

of the device. This information is constantly sent to

the FM. The operator manager checks, if an operator

to be executed is available or even possible to deploy

with regards to available resources. FA operators are

container-based comparable to their cloud

counterparts. This allows for processing isolation. In

addition, missing operators are pulled conveniently

using the mechanisms current container software

solutions offer. The monitoring service checks if all

deployed operators are running and sends FA health

data to the FM.

5 CONCLUSIONS AND

OUTLOOK

In this paper, we presented and architectural approach

for IoT analytics in smart home environments. It is

based on the concept of fog computing to enable low

latency, fault-tolerant and privacy observing data

processing, all of which are requirements for smart

home analytics architectures identified by scientific

literature.

We provided insights into the current state of the

art in IoT analytics architectures research and showed

that already existing solutions are not sufficient to

address all architectural challenges identified for

smart home analytics. We found that fault-tolerance,

low-latency data processing as well as external

regulations are key aspects when designing an

analytics architecture. The presented approach

utilizes several concepts from system and software

engineering, such as fog computing, Kappa

architecture, microservices and container

virtualisation to solve the surrounding problems.

Moreover, the solution architecture comprises

components to allow for optimal orchestration of data

processing along individual analytics pipelines.

Future research in this field needs to focus on

identifying conditions and requirements for

deployment rules of analytics pipeline tasks. It seems

plausible to derive the resulting rules from “static”

sources such as resource availability, legislation, etc.,

but also from human behaviour using already

gathered IoT data. Understanding the usage of smart

devices by consumers may lead to the use of machine

ICEIS 2019 - 21st International Conference on Enterprise Information Systems

620

learning algorithms to establish optimal distribution

of analytics pipeline jobs between cloud and fog

nodes. Finally, the research concerning

choreography-based scheduling approaches of

analytics tasks needs to be furthered in order to

provide increased fault-tolerance of the overall

architecture.

A prototypical implementation of the solution

proposal has already been developed. The resulting

software prototype needs to be evaluated in future

research with regards to the challenges for IoT

analytics as well as fog computing architectures, but

also regarding performance compared to different

architectural concepts.

ACKNOWLEDGEMENTS

The work presented in this paper is partly funded by

the European Regional Development Fund (ERDF)

and the Free State of Saxony (Sächsische Aufbaubank

- SAB)

REFERENCES

Alturki, B., Reiff-Marganiec, S., and Perera, C. (2017). A

hybrid approach for data analytics for internet of things.

In S. Mayer (Ed.), ICPS: ACM international conference

proceeding series, Proceedings of the Seventh

International Conference on the Internet of Things

(pp. 1–8). New York, NY, USA: ACM.

Amazon. (2019). Alexa Voice Service. Retrieved from

https://developer.amazon.com/de/alexa-voice-service

Auger, A., Exposito, E., and Lochin, E. (2017). Sensor

observation streams within cloud-based IoT platforms:

Challenges and directions. In 2017 20th Conference on

Innovations in Clouds, Internet and Networks (ICIN)

(pp. 177–184). IEEE.

Biswas, A. R., and Giaffreda, R. (2014). IoT and cloud

convergence: Opportunities and challenges. In 2014

IEEE World Forum on Internet of Things (WF-IoT)

(pp. 375–376). IEEE.

Brito, M. S. de, Hoque, S., Magedanz, T., Steinke, R.,

Willner, A., Nehls, D., . . . Schreiner, F. (2017). A

service orchestration architecture for Fog-enabled

infrastructures. In 2017 Second International

Conference on Fog and Mobile Edge Computing

(FMEC) (pp. 127–132). IEEE.

Brush, A. J., Hazas, M., and Albrecht, J. (2018). Smart

Homes: Undeniable Reality or Always Just around the

Corner? IEEE Pervasive Computing, 17(1), 82–86.

Byers, C. C. (2017). Architectural Imperatives for Fog

Computing: Use Cases, Requirements, and

Architectural Techniques for Fog-Enabled IoT

Networks. IEEE Communications Magazine, 55, 14–

20.

Cheng, B., Longo, S., Cirillo, F., Bauer, M., and Kovacs, E.

(2015). Building a Big Data Platform for Smart Cities:

Experience and Lessons from Santander. In B.

Carminati (Ed.), 2015 IEEE International Congress on

Big Data (BigData Congress): June 27, 2015 - July 2,

2015, New York, New York, USA (pp. 592–599).

Piscataway, NJ: IEEE.

Dastjerdi, A. V., and Buyya, R. (2016). Fog Computing:

Helping the Internet of Things Realize Its Potential.

Computer, 49, 112–116.

Haddadi, H., Christophides, V., Teixeira, R., Cho, K.,

Suzuki, S., and Perrig, A. (2018). SIOTOME: An Edge-

ISP Collaborative Architecture for IoT Security.

IDC (2018). New IDC Smart Home Device Tracker

Forecasts Solid Growth for Connected Devices in Key

Smart Home Categories. Retrieved from

https://www.idc.com/getdoc.jsp?containerId=prUS437

01518

IHS. (2018). IoT Trend Watch 2018. Retrieved from

https://cdn.ihs.com/www/pdf/IoT-Trend-Watch-

eBook.pdf?utm_campaign=PC10273-

2_MD_eT1_MT_TMT_GLOBAL_IoT-Theme_3rd-

IoT-

eBook_2018_customers&utm_medium=email&utm_s

ource=Eloqua

Klonoff, D. C. (2017). Fog Computing and Edge

Computing Architectures for Processing Data From

Diabetes Devices Connected to the Medical Internet of

Things. Journal of Diabetes Science and Technology,

11, 647–652.

Kreps, J. (2014). Questioning the Lambda Architecture:

The Lambda Architecture has its merits, but

alternatives are worth exploring. Retrieved from

https://www.oreilly.com/ideas/questioning-the-

lambda-architecture

Marjani, M., Nasaruddin, F., and Gani, A. (2017). Big IoT

Data Analytics: Architecture, Opportunities, and Open

Research Challenges. IEEE Access, 5, 5247–5261.

Mouradian, C., Naboulsi, D., Yangui, S., Glitho, R. H.,

Morrow, M. J., and Polakos, P. A. (2018). A

Comprehensive Survey on Fog Computing: State-of-

the-Art and Research Challenges. IEEE

Communications Surveys & Tutorials, 20(1), 416–464.

Ravindra, P., Khochare, A., Reddy, S. P., Sharma, S.,

Varshney, P., and Simmhan, Y. (2017). ECHO: An

Adaptive Orchestration Platform for Hybrid Dataflows

across Cloud and Edge.

Rehman, M. H. u., Ahmed, E., Yaqoob, I., Hashem, I. A.

T., Imran, M., and Ahmad, S. (2018). Big Data

Analytics in Industrial IoT Using a Concentric

Computing Model. IEEE Communications Magazine,

56, 37–43.

Rozik, A. S., Tolba, A. S., and El-Dosuky, M. A. (2016).

Design and Implementation of the Sense Egypt

Platform for Real-Time Analysis of IoT Data Streams.

Advances in Internet of Things, 06, 65–91.

Stojkoska, B. L. R., and Trivodaliev, K. V. (2017). A

review of Internet of Things for smart home:

A Fog-enabled Smart Home Analytics Platform

621

Challenges and solutions. Journal of Cleaner

Production, 140, Part 3, 1454–1464.

Stolpe, M. (2016). The Internet of Things: Opportunities

and Challenges for Distributed Data Analysis. ACM

SIGKDD Explorations Newsletter, 18, 15–34.

Ta-Shma, P., Akbar, A., Gerson-Golan, G., Hadash, G.,

Carrez, F., and Moessner, K. (2018). An Ingestion and

Analytics Architecture for IoT Applied to Smart City

Use Cases.IEEE Internet of Things Journal,5, 765–774.

Velasquez, K., Abreu, D. P., Goncalves, D., Bittencourt, L.,

Curado, M., Monteiro, E., and Madeira, E. (2017).

Service Orchestration in Fog Environments. In M.

Younas, M. Aleksy, and J. Bentahar (Eds.), 2017 IEEE

5th International Conference on Future Internet of

Things and Cloud: FiCloud 2017: Prague, Czech

Republic, 21-23 August 2017 : proceedings (pp. 329–

336). Piscataway, NJ: IEEE.

Wilson, C., Hargreaves, T., and Hauxwell-Baldwin, R.

(2017). Benefits and risks of smart home technologies.

Energy Policy, 103, 72–83.

Zaslavsky, A., and Georgakopoulos, D. (2015). Internet of

Things: Challenges and State-of-the-Art Solutions in

Internet-Scale Sensor Information Management and

Mobile Analytics. In 2015 16th IEEE International

Conference on Mobile Data Management (MDM)

(pp. 3–6).

Zschörnig, T., Wehlitz, R., and Franczyk, B. (2017). A

Personal Analytics Platform for the Internet of Things:

Implementing Kappa Architecture with Microservice-

based Stream Processing. In Proceedings of the 19th

International Conference on Enterprise Information

Systems (pp. 733–738). SCITEPRESS - Science and

Technology Publications.

ICEIS 2019 - 21st International Conference on Enterprise Information Systems

622