Athens Journal of Technology and Engineering - Volume 10, Issue 4,

December 2023 – Pages 215-226

A Data Streaming Architecture for Air Quality

Monitoring in Smart Cities
By Aleksa Miletić, Petar Lukovac, Tamara Naumović,
Danijela Stojanović & Aleksandra Labus
This paper aims to present a modeling approach for the seamless data streaming
process from smart IoT systems to Apache Kafka, leveraging the MQTT
protocol. The paper begins by discussing the concept of real-time data streaming,
emphasizing the need to transfer data from IoT/edge devices and sensors to
Apache Kafka in a timely manner. The second part consists of a literature
overview that shows the analysis and systematization of different types of
architectures in the broad sense of crowdsensing, followed by specific
architectures regarding edge and cloud computing. The methodology section
will propose an infrastructure and data streaming architecture for smart
environment services, such as air quality monitoring. Lastly, a discussion about
results and future development will be shown in the last two sections. The
proposed integration approach offers several advantages, including efficient
and scalable data streaming, real-time analytics, and enhanced data processing
capabilities.

Keywords: real-time data streaming, smart healthcare, Apache Kafka, data

integration, stream processing

Introduction

Efficient message distribution systems have gained significant attention in

recent years, as they contribute to optimizing architectural frameworks. These
optimizations became a necessity having in mind that the Internet of Things (IoT)
paradigm, a leader in Industry 4.0, increased the number of devices connected to
the Internet (Akbar et al. 2017, Froiz-Míguez et al. 2018, Khriji et al. 2022),
creating a climate for the successful development of numerous services in the
domain of smart cities (Cheng et al. 2018, Khriji et al. 2022, Samizadeh Nikoui et
al. 2021). Moreover, nearly all services of smart cities are in demand for ultralow
latency and fast response time, further proving the need for efficient message
distribution systems.
Different architectures require different approaches in the modelling and
development of message distribution systems:


Teaching Associate, University of Belgrade Faculty of Organizational Sciences, Serbia.

Teaching Associate, University of Belgrade Faculty of Organizational Sciences, Serbia.

Teaching Assistant, University of Belgrade Faculty of Organizational Sciences, Serbia.

Research Associate, Institute of Economic Sciences, Serbia.

Full Professor, University of Belgrade Faculty of Organizational Sciences, Serbia.

https://doi.org/10.30958/ajte.10-4-2 doi=10.30958/ajte.10-4-2
Vol. 10, No. 4 Miletic et al.: A Data Streaming Architecture for Air Quality Monitoring…

- (Hugo et al. 2020) propose the use of Apache Kafka as a centralized

message distribution system to streamline the architecture of monolithic
applications.
- (Fridelin Panduman et al. 2019) also propose the use of Apache Kafka
server, alongside MQTT protocol, for SEMAR (Smart Environment
Monitoring and Analytics in Real-time), a cloud computing platform based
on microservice architecture.
- (Khriji et al. 2022) proposed AWS (Amazon Web Services) cloud-based
architecture with AWS message broker - REDA (Real-time Event-Driven
Architecture).

Smart city services are relying on infrastructures based on IoT, edge, and
cloud computing technologies, whose interoperability requires an efficient message
distribution system. This paper proposes a methodology that enables continuous
monitoring and collection of air quality data in real-time, from sensors deployed at
different locations in smart cities. The methodology introduces a two-layer
architecture: Edge – air quality data collection, and Cloud – receiving and
processing data in a stream.
This paper is organized as follows. Next section presents a review of relevant
literature related to smart city IoT architectures and data processing. Then comes
the methodology and design of the proposed data streaming architecture. While
insights about the implementation and discussion about results are provided
afterwards and finally the paper is concluded with a discussion of the applicability
of the proposed architecture and future development.

Literature Review

Based on the rapid growth and complex requests of modern systems, for
cloud and edge computing it is important to study different methods and
architectures which have proven to be successful. This analysis will allow us to
better understand best practices and address the challenges that bring us this
dynamic field of computing.
Edge computing is a decentralized computing infrastructure that allows
remote devices to process data closer to the edge of the network, near the source
(Mitrović et al. 2023). Several analyses have been conducted on an edge
computing platform that proves edge computing is a good solution for cooperation
with cloud, network communication, and edge equipment (Chen et al. 2018,
Martin Fernandez et al. 2018, Raza et al. 2019). This approach offers several
advantages, including reduced latency, bandwidth optimization, enhanced privacy
and security, offline operation (Hassan et al. 2019, Shi et al. 2016, Varghese et al.
2016). Also, one of the most important things in edge is data privacy, reduced
attack surface, local threats, communication security, trustworthiness of edge
devices. Some of these problems are addressed in the following papers (Ali et al.
2021, Markham and Payne 2001, Xiao et al. 2019).

216
Athens Journal of Technology & Engineering December 2023

Cloud-based Data Stream Processing is a type of data processing system that

executes a set of continuous queries over a potentially unbounded data stream
(Heinze et al. 2014). It constantly outputs new results and is designed to
dynamically scale to hundreds of computing nodes and automatically cope with
varying workloads. This streaming approach enables real-time or near-real-time
data ingestion, allowing organizations to gain valuable insights, make data-driven
decisions, and take timely actions based on the incoming data (De Souza et al.
2020, Sahni and Vidyarthi 2021). The main thing about cloud computing is that it
allows users to quickly deploy their applications in an elastic setting through on-
demand acquisition/release of resources at a low cost (Runsewe and Samaan
2021).
In recent years, there has been a growing interest in employing efficient
message distribution systems to optimize architectural frameworks. The authors
(Hugo et al. 2020) propose utilizing Apache Kafka as a centralized message
distribution system to streamline the architecture for monolithic applications. The
integration of MQTT and Apache Kafka is achieved through the development of a
specialized Kafka Connect connector, enabling scalable and resilient data
collection from MQTT sources and its transmission to Kafka server. This
integration facilitates fast and reliable real-time data transfer between various
applications, particularly for handling large data volumes. The platform called
SEMAR (Smart Environment Monitoring and Analytical in Real-time) (Fridelin
Panduman et al. 2019) presents a cloud computing platform based on microservice
architecture. The system consists of a device that is placed in the car and sends
data via MQTT protocol to the server on which Apache Kafka is installed. They
found that the average delay in sending information is 0.09ms, which is enough
for the system to be considered to work in real-time. REDA is a cloud-based
event-driven architecture proposed for real-time data processing in wireless sensor
networks and IoT devices (Khriji et al. 2022). This architecture offers flexibility,
high availability, and distributed computing while achieving high throughput and
minimizing latency. By utilizing open-source frameworks and components such as
a sensing unit, gateway unit, lightweight messaging protocol, event-stream
processing unit, and distributed document database, REDA presents a cost-
effective solution for real-time data processing. Authors (Mitrović et al. 2023)
presented an IoT based framework for air quality measurements, that is suitable
for implementation into different smart environments.
The article of Javed et al. (2018) presents a fault-tolerant architecture for IoT
applications, which combines cloud and edge computing. This architecture, named
CEFIoT, allows flexible compute placement on edge or cloud without any
changes in source code. Also, the article shows a case study of a security camera
system to show fault tolerance possibility in architecture. Cao et al. (n.d.) explain
edge computing as a computing paradigm that does a computation on the edge of a
network. Proximity to users and the provision of better intelligent services are
highlighted by better performance of data transfer and decreasing data latency.
Edge computing aims to provide services at the edge of the network and meet the
needs of the IT industry in high-speed connectivity, real-time processing, and data
security. Also, one interesting topic is EMMA which serves as a middleware

217
Vol. 10, No. 4 Miletic et al.: A Data Streaming Architecture for Air Quality Monitoring…

solution that enhances the efficiency, reliability, and performance of MQTT-based

communication in edge computing environments, enabling more effective and
responsive IoT applications. The authors (Rausch et al. 2018) consider how edge
computing can improve the performances of just cloud-based architectures. The
opinion of Koziolek et al. (n.d.) is that MQTT Broker is a good choice for
distributed IoT Edge Computing for a few reasons. First, MQTT brokers such as
Eclipse Mosquitto, EMQX, HiveMQ i VerneMQ support the MQTT protocol
which is lightweight, effective, and good for IoT communication. Therefore, these
brokers provide support for clustering, enabling scalability and high system
availability. As open-source code, they are customizable and available in different
variants, providing opportunities to adapt to specific project requirements. These
features make MQTT Brokers a popular choice for distributed IoT Edge Computing
scenarios.
Sotskov et al. (2023) presented IRONEDGE architectural framework that can
be used in different edge Stream Processing solutions. The first layer of this
architecture is the Data Source layer which is used for processing and sending
data. Data that is collected and processed in the Data Source layer is ready for
publishing and called Data Reports. The second layer consists of a few services.
The first one being Data Collection service which receives data from corresponding
Data Sources sending it to the Stream processing service which is used to derive
knowledge from the accumulated Data Reports, which creates a base for event
creation. Data transfer is enabled by AMM (Asynchronous Messaging Middleware).
Local data, events, and reports are stored in Local Storage for further analysis.
Data stream processing can be divided into more phases: receiving data, mapping
in a particular object, decomposition of data according to certain criteria, analysis
of grouped data, and transferring to the cloud. As AMM (Sotskov et al. 2023)
chose Apache Kafka because of easy connectivity with external systems. The
authors have observed the performances of the system in two cases: with Kafka
Connect (K0-WC) and without Kafka Connect (K1-NC). They concluded that for
loss rates and latency, K1-NC offered lower event loss and lower processing
latency compared to K0-WC. For actual throughput and log time, K1-NC showed
an increase in throughput but never matched the input rate for the Data Reports,
while K0-WC clearly showed low throughput resulting from high losses.
As it is shown in previous examples of architectures, to send data in
constantly changing environments and at the same time in real-time, MQTT is
particularly useful, while Apache Kafka is used to store large amounts of data or to
integrate different applications or data centers. Because of their similarities, they
can be considered a great couple. Both have topics and publish/subscribe patterns.
There are several ways to connect MQTT and Kafka (MQTT and Kafka. How to
Combine Two Complementary… | by Techletters | Python Point | Medium n.d.):

1. Sending messages from edge devices, using MQTT and Kafka broker. The
problem arises in the fact that the device has to send data using two
protocols, and it has to be sure that the data either arrive on both or does
not arrive on either of them.

218
Athens Journal of Technology & Engineering December 2023

2. Creating an application that will represent a bridge between MQTT and

Kafka.
3. Connecting to Kafka via MQTT proxy. It is used if there is a need to send
messages that will for sure arrive on the other side.
4. Connection of MQTT broker and Kafka cluster via Kafka Connect – a
good solution that uses connectors as an extension to connect.
5. Connection between MQTT broker and Kafka via MQTT broker
extension - this extension enables the injection of messages from IoT
devices into the Kafka cluster.

Most authors include mobile phones in their systems to gather data from
mobile device sensors. In the research of Kraft et al. (2020), a mobile collective
sensing system is proposed that enables the implementation of a noise level map
for tinnitus patients. After the requirements analysis and design phase, the system
is decomposed into bounded contexts to achieve a clear and shared definition of
consistency between team members. Five bounded contexts were identified,
including user identity, social aspects, measurement, incentives, and communication.
This architecture uses a cloud-native approach, which enables efficient and
scalable processing of concurrent noise measurement requests, using microservices
and containerization technologies such as Docker and Kubernetes. In addition, the
system uses in-stream processing and the Apache Kafka platform for real-time
data processing and enabling decoupled processing of incoming geospatial data.
To display polluted areas on the map, the authors decided to use geospatial data
partitioning techniques. Hierarchical partitioning of geospatial data, such as the
implementation of the Discrete Global Grid System (DGGS), allows data to be
divided into different levels of detail. DGGS enables the representation of data in
different partition sizes, which enables aggregation and visualization at different
scales. The map has Hexagonal Hierarchical Spatial Index (H3) system. H3
systems allow precise positioning of geospatial data in appropriate partitions based
on their coordinates. This technique facilitates analyzing and visualizing polluted
parts on the map.
The development of the mobile crowdsensing system for monitoring noise
pollution for decision-making purposes in smart cities by collecting, storing, and
visualizing data on noise pollution in real time has been described in the article
(Jezdović et al. 2021). These authors also point out the experiment results
conducted in Belgrade, Serbia, and recommendations on how this system can be
applied in other cities. The application presented in this paper is a mobile
crowdsensing system for detecting noise in smart cities. The system contains a
crowdsensing mobile application, cloud, and big data infrastructure. The mobile
application enables noise recording using a microphone on mobile devices,
recording the location of detected noise using a GPS device, performing spectral
analysis on audio data, and storing transformed data and location data in a cloud
database. The web application allows the view of polluted data on Google Maps.
The type of the map is a heatmap on which is presented red, orange, and green
areas for the high, medium, and low levels of noise respectively. In the last two

219
Vol. 10, No. 4 Miletic et al.: A Data Streaming Architecture for Air Quality Monitoring…

papers we can see that the authors have used two different approaches to represent
their data on the map.

Methodology

This paper presents a methodology that enables continuous monitoring and

collection of air quality data in real time, from sensors deployed at different
locations in smart cities. With this architecture, data is transferred from Edge
devices to Cloud servers, where it can be further analysed and used for decision-
making and air quality management in smart cities. The methodology of this
research includes the use of a two-layer architecture: Edge and Cloud. In the Edge
layer, air quality data is collected from sensors connected to a Raspberry Pi (RPi)
device. According to the Environmental Protection Agency, the data that needs to
be measured in Belgrade are SO2, PM10, NO2, CO, NO, PM2.5 (Агенција За
Заштиту Животне Средине - Министарство Заштите Животне Средине
n.d.). Such data is sent via the MQTT protocol to the cloud (see Figure 1).

Figure 1. Data Streaming Architecture for Air Quality Monitoring in Smart Cities

To enable data transfer, the MQTT protocol is used and the Mosquitto MQTT
broker, which is installed on the RPi device. Mosquito was chosen because it is the
MQTT protocol which provides a lightweight method of carrying out messaging
using a publish/subscribe model. This makes it suitable for IoT messaging such as
with low power sensors or mobile devices such as phones, embedded computers
or microcontrollers (What Is Mosquitto MQTT? n.d.). These tools allow sending
data to Apache Kafka, which resides in the Cloud layer of the architecture. Apache
Kafka is used as a central mechanism for receiving and processing data in a
stream. Also, it is chosen as the tool because it is highly scalable for processing
and streaming data in real time. Kafka enables data replication and data availability
even in case of network failures or interruptions, which is important for continuous
monitoring of air quality (Korab n.d.). Also, Kafka supports simultaneous sending
and receiving of data, which enables real-time data analysis and processing. This is

220
Athens Journal of Technology & Engineering December 2023

a very important feature in smart cities because it provides quick detection and
reaction to changes in air quality. The data coming to Kafka would be processed
and saved in the database and sent in real time via the access series to end user
applications.

Figure 1. Data Streaming Architecture for Air Quality Monitoring in Smart Cities
with Data Flow

This paper proposes a detail architecture which is presented on Figure 2. The

system is designed to connect to devices across the city. A Raspberry Pi device is
set up to which a group of sensors for measurement is attached SO2, PM10, NO2,
CO, NO, CO2, NH3, PM2.5. The image shows the flow of data from the carbon
monoxide sensor. Data is prepared for sending by creating an object with all the
necessary information and metadata (see Figure 3). From this point data will be
sent to corresponding MQTT topic and from there streamed to the Apache Kafka.
Specifically, carbon monoxide data will be sent to the topic named carbon-
monoxide. Connection between MQTT and Apache Kafka using Kafka Connect
was chosen. In order to connect MQTT with Kafka, there is a special Kafka
connect that collects data and writes it to the Kafka topic (carbon-monoxide) –
MQTT Source Connector. Before the data reaches the Kafka Cluster, it is
necessary to check if data is valid. Kafka Schema Registry serves to validate the
object that arrives to Kafka. If the structure of the objects does not match, that data
will not be processed. When the data reaches the Kafka Cluster, i.e. on the carbon-
monoxide topic, it is forwarded to Kafka Stream Analytics, where data analysis is
performed, such as geographic analysis based on location, time analysis for
tracking data history etc. The processed data is forwarded to the carbon-monoxide
analytics topic. From this topic, data is collected by the Access Service and
forwarded to all users for monitoring pollution in real time. The processed data is
also stored in the database to monitor the history of air quality. The data is first
pulled from the topic via a Kafka Connect called MongoDB Sink, which is used to
write data to MongoDB.

221
Vol. 10, No. 4 Miletic et al.: A Data Streaming Architecture for Air Quality Monitoring…

Figure 2. Example of Object Sent to Apache Kafka

Results and Discussion

A prototype data streaming system was developed to display measurements

of air quality, focusing on monitoring Sulfur Dioxide (SO2), Particulate Matter
(PM10, PM2.5), Nitrogen Dioxide (NO2), Carbon Monoxide (CO) and Nitric
Oxide (NO), Ammonia (NH3), carbon dioxide (CO2) levels. The system was
based on a Raspberry Pi microcomputer with various sensors connected,
including the MQ-7, MQ-137, MQ-135, MQ-136, MQ-165, MG-811, WSP-
1110, and SDS011. Moreover, instances where a sensor malfunctioned or
produced default values that significantly deviated from previous measurements
resulted in the exclusion of data from being sent to the cloud for processing.
To evaluate the quality of data, it was necessary to take into consideration
that sensors might exhibit a certain degree of imprecision. To ensure reliable
values, averaging calculations were performed using data from multiple
sensors. This approach allowed for a more accurate assessment of data quality
and minimized the impact of potential inaccuracies from individual sensors.
The collected raw data was streamed via Apache Kafka. Every type of sensor
has its topic in the MQTT broker which sends data to the Kafka topic for that
specific measured value. Apache Kafka performed real-time analyses of
collected data, stored them to the MongoDB and streamed live. NoSQL
database has been chosen because large amounts of data are expected from IoT
systems with numerous writing operations (Tudorica and Bucur 2011). By
leveraging the power of Apache Kafka, the system facilitated efficient data
transmission, allowing for real-time monitoring and analysis of air quality
metrics. To derive meaningful insights, calculations were performed to determine
the average pollution levels for each specific hour (see Figure 4) and specific
periods of the day (see Figure 5). All measurements are shown on different

222
Athens Journal of Technology & Engineering December 2023

diagrams due to the sensors different output value ranges. These metrics
provided valuable information for assessing air quality trends and identifying
potential pollution patterns during different periods of the day in Belgrade.

Figure 3. Line Chart Based on Average Values for Every Hour in the Day

Figure 4. Bar Chart Based on Average Values for Specific Times of the Day

223
Vol. 10, No. 4 Miletic et al.: A Data Streaming Architecture for Air Quality Monitoring…

Conclusions

The main goal of this paper is to present the design and development of the
robust architecture for crowdsensing systems in smart cities. This architecture
provides the base for effective collection and analysis of data in real-time, opening
new possibilities for measuring and monitoring level of pollution and other
parameters in local environment. Even though the architecture shows easy
processing and analyses of data, some shortcomings have also been identified.
Setting optimal number of topics and replication factor of topics, and how many
brokers is it needed. For now the replication number is set to three, because it is
the golden rule (Ibryam n.d.), but with the further analysis and testing that can be
changed because of a number of sensors that will be sending data to the Apache
Kafka.
Implementation of the developed architecture has applicability in practice for
creating crowdsensing systems in smart cities, especially for measuring air quality
levels. This enables fast and efficient data collection from various sensors and
devices, providing valuable information for management and monitoring of
environmental quality. Future steps may include implementing the developed
architecture on the Docker platform, scaling the system via Kubernetes, and
displaying measurements via a map. This would enable better resource management
and scaling, as well as greater flexibility and fault tolerance in the system
environment. Through this work, it was observed that creating an efficient
architecture for data streaming in crowdsensing systems is essential for successful
real-time data collection and analysis. These results can serve as guidelines for
other researchers dealing with similar problems in the field of smart cities and air
quality meters. Future research will be focused on exploring the possibility of
applying this architecture in wider contexts of smart cities, as well as optimizing
the number of topics, replications, and brokers to achieve maximum efficiency and
scalability. Also, research into the integration of this architecture with other
relevant technologies and platforms opens the door for further improvement of
crowdsensing systems in smart cities.

References

Akbar A, Khan A, Carrez F, Moessner K (2017) Predictive analytics for complex IoT data
streams. IEEE Internet of Things Journal 4(5): 1571–1582.
Ali B, Gregory MA, Li S (2021) Multi-access edge computing architecture, data security
and privacy: a review. IEEE Access 9: 18706–18721.
Cao K, Liu Y, Meng G, Sun Q (n.d.) An overview on edge computing research. Available
at: https://doi.org/10.1109/ACCESS.2020.2991734.
Chen B, Wan J, Celesti A, Li D, Abbas H, Zhang Q (2018) Edge computing in IoT-based
manufacturing. IEEE Communications Magazine 56(9): 103–109.
Cheng B, Solmaz G, Cirillo F, Kovacs E, Terasawa K, Kitazawa A (2018) FogFlow: easy
programming of IoT services over cloud and edges for smart cities. IEEE Internet of
Things Journal 5(2): 696–707.
De Souza PRR, Matteussi KJ, Veith ADS, Zanchetta BF, Leithardt VRQ, Murciego AL,

224
Athens Journal of Technology & Engineering December 2023

et al. (2020) Boosting big data streaming applications in clouds with burstflow. IEEE
Access 8: 219124–219136.
Fridelin Panduman YY, Ulil Albaab MR, Anom Besari AR, Sukaridhoto S, Tjahjono A,
Nourma Budiarti RP (2019) Implementation of data abstraction layer using kafka on
SEMAR platform for air quality monitoring. International Journal on Advanced
Science, Engineering and Information Technology 9(5): 1520–1527.
Froiz-Míguez I, Fernández-Caramés TM, Fraga-Lamas P, Castedo L (2018) Design,
Implementation and Practical Evaluation of an IoT Home Automation System for
Fog Computing Applications Based on MQTT and ZigBee-WiFi Sensor Nodes.
Sensors 18(8): 2660.
Hassan N, Yau KLA, Wu C (2019) Edge computing in 5G: A review. IEEE Access 7:
127276–127289.
Heinze T, Aniello L, Querzoni L, Jerzak Z (2014) Cloud-based data stream processing. In
DEBS 2014 - Proceedings of the 8th ACM International Conference on Distributed
Event-Based Systems, 238–245.
Hugo A, Morin B, Svantorp K (2020) Bridging MQTT and Kafka to support C-ITS: a
feasibility study. In Proceedings - IEEE International Conference on Mobile Data
Management, 2020-January, 371–376.
Ibryam B (n.d.) Fine-tune Kafka performance with the Kafka optimization theorem. Red
Hat Developer. Available at: https://developers.redhat.com/ articles/2022/05/03/fine-
tune-kafka-performance-kafka-optimization-
theorem#the_kafka_optimization_theorem.
Javed A, Heljanko K, Buda A, Framling K (2018) CEFIoT: a fault-tolerant IoT
architecture for edge and cloud. In IEEE World Forum on Internet of Things, WF-
IoT 2018 - Proceedings, 2018-January, 813–818.
Jezdović I, Popović S, Radenković M, Labus A, Bogdanović Z (2021) A crowdsensing
platform for real-time monitoring and analysis of noise pollution in smart cities. In
Sustainable Computing: Informatics and Systems, 31.
Khriji S, Benbelgacem Y, Chéour R, Houssaini DE, Kanoun O (2022) Design and
implementation of a cloud-based event-driven architecture for real-time data
processing in wireless sensor networks. Journal of Supercomputing 78(3): 3374–
3401.
Korab J (n.d.) How to survive a Kafka outage. Available at: https://www.confluent.io/
blog/how-to-survive-a-kafka-outage/.
Koziolek H, Grüner S, Rückert J (n.d.) A comparison of MQTT brokers for distributed IoT
edge computing.
Kraft R, Birk F, Reichert M, Deshpande A, Schlee W, Langguth B, et al. (2020) Efficient
processing of geospatial mhealth data using a scalable crowdsensing platform.
Sensors (Switzerland) 20(12): 1–21.
Markham T, Payne C (2001) Security at the network edge: a distributed firewall
architecture. In Proceedings - DARPA Information Survivability Conference and
Exposition II, DISCEX 2001, 1, 279–286.
Martin Fernandez C, Diaz Rodriguez M, Rubio Munoz B (2018) An edge computing
architecture in the internet of things. In Proceedings - 2018 IEEE 21st International
Symposium on Real-Time Computing, ISORC 2018, 99–102.
Mitrović N, Đorđević M, Veljković S, Danković D (2023) View of IoT enabled software
platform for air quality measurements. Available at: https://www.ebt.rs/journals/in
dex.php/conf-proc/article/view/188/135.
MQTT and Kafka. How to combine two complementary… | by Techletters | Python Point |
Medium (n.d.) Available at: https://medium.com/python-point/ mqtt-and-kafka-
8e470eff606b.

225
Vol. 10, No. 4 Miletic et al.: A Data Streaming Architecture for Air Quality Monitoring…

Rausch T, Nastic S, Dustdar S (2018) EMMA: Distributed QoS-aware MQTT middleware

for edge computing applications. In Proceedings - 2018 IEEE International
Conference on Cloud Engineering, IC2E 2018, 191–197.
Raza S, Wang S, Ahmed M, Anwar MR (2019) A survey on vehicular edge computing:
architecture, applications, technical issues, and future directions. Wireless
Communications and Mobile Computing 2019: 3159762.
Runsewe O, Samaan N (2021) Cloud resource scaling for time-bounded and unbounded
big data streaming applications. IEEE Transactions on Cloud Computing 9(2): 504–
517.
Sahni J, Vidyarthi DP (2021) Heterogeneity-aware elastic scaling of streaming
applications on cloud platforms. Journal of Supercomputing 77(9): 10512–10539.
Samizadeh Nikoui T, Rahmani AM, Balador A, Haj Seyyed Javadi H (2021) Internet of
Things architecture challenges: a systematic review. International Journal of
Communication Systems 34(4): e4678.
Shi W, Cao J, Zhang Q, Li Y, Xu L (2016) Edge computing: vision and challenges. IEEE
Internet of Things Journal 3(5): 637–646.
Sotskov YN, Tchernykh A, Werner F, Vitorino JP, Simão J, Datia N, et al. (2023)
IRONEDGE: stream processing architecture for edge applications. Algorithms 2023
16(2): 123.
Tudorica BG, Bucur C (2011) A comparison between several NoSQL databases with
comments and notes. In Proceedings - RoEduNet IEEE International Conference.
Varghese B, Wang N, Barbhuiya S, Kilpatrick P, Nikolopoulos DS (2016) Challenges and
opportunities in edge computing. In Proceedings - 2016 IEEE International
Conference on Smart Cloud, SmartCloud 2016, 20–26.
What is Mosquitto MQTT? (n.d.) Available at: https://www.eginnovations.com/documen
tation/Mosquitto-MQTT/What-is-Mosquitto-MQTT.htm.
Xiao Y, Jia Y, Liu C, Cheng X, Yu J, Lv W (2019) Edge computing security: state of the
art and challenges. In Proceedings of the IEEE.
Агенција за заштиту животне средине - Министарство заштите животне
средине (n.d.) Available at: rom http://www.sepa.gov.rs/.

226