Cluster Computing

https://doi.org/10.1007/s10586-019-02986-2 (0123456789().,-volV)(0123456789().
,- volV)

The role of structured and unstructured data managing mechanisms

in the Internet of things
Poopak Azad1 • Nima Jafari Navimipour2 • Amir Masoud Rahmani1 • Arash Sharifi1

Received: 23 January 2019 / Revised: 15 May 2019 / Accepted: 16 September 2019

 Springer Science+Business Media, LLC, part of Springer Nature 2019

Abstract
Internet of Things (IoT) has the idea of receiving data and sending them for each object within the communication network.
One of the main issues in this type of networks is handling a growing volume of data with various data sources and data
types to satisfy the performance requirements of applications. In this regard, data management in the IoT plays an
important role in its efficient operations and has become a major research topic. However, the data management has a
crucial role in the IoT, there is not any comprehensive and systematic work to analyze its approaches. Thus, main propose
of this paper is to systematically review and study the existing data management approaches in the IoT. The data
management approaches are classified into three main classes, including SQL database, NoSQL database, and graph
database. In addition, the detailed comparison of the important mechanisms in each category brings a recommendation for
further works.

Keywords Data management
Data storage
Internet of things
Systematic literature review

1 Introduction frameworks and mechanisms have been proposed to col-

lect, manage and store the data produced in the IoT pro-
Internet of things (IoT) presents services for daily use and cesses in order to deal with this challenge [1, 26]. The rise
it is built through purposeful connectivity within objects of the IoT will change traditional database systems to adapt
and network platforms [25, 35]. It works by receiving and to huge volumes of unstructured and semi-structured data
integrating information between objects connected together from different sources. Data management refers to the
to offer information-based services [40]. In the IoT plat- practices, procedures, and architectures for properly data
form, many of the devices that are around us will be on the lifecycle management. In the field of IoT, data manage-
network [24, 59]. Sensor network and Radio Frequency ment is considered as a layer among the devices and
IDentification (RFID) technologies are integrated to offer objects producing the applications and data [2].
these services where communication and information sys- On the other hand, despite the importance of data
tems are invisibly placed in the network [19]. This leads to management issue in the IoT, as far as we know, there is
massive volumes of data that produced by devices and has not any systematic and comprehensive research about its
to be presented, processed, and then stored in an efficient, important mechanisms that realize the requirement for
interpretable and seamless form [22]. researchers to do more study on this issue. Thus, the aim of
One of the main research issues for the following years this paper is to study the existing data management tech-
is storing the data generated by devices [62]. Some niques in the IoT systematically and comprehensively. The
contributions of this paper are as follow:
& Nima Jafari Navimipour • Providing an overview of the present issues in the
jafari@iaut.ac.ir domains integrated with data management in the IoT.
1
Department of Computer Engineering, Science and Research
• Providing an overview and systematic study of the
Branch, Islamic Azad University, Tehran, Iran present techniques for data management in the IoT.
2 • Providing anatomy of different important mechanisms
Department of Computer Engineering, Tabriz Branch,
Islamic Azad University, Tabriz, Iran on the aspect of the IoT data management.

123
 Cluster Computing

• Outlining key areas to improve the IoT data manage- Furthermore, the utilization of IoT technology in
ment methods in future studies. applications has resulted in the rise of real-time data, which
makes the data access and storage more challenging and
The structure of this article is as follows. After the
difficult. Shona and Arathi [44] have discussed the issues
introduction, the related work is reviewed and analyzed in
for data management systems in the IoT to manage the high
Sect. 2. Then, backgrounds of IoT data management are
amount of operational data generated by devices and
provided in Sect. 3. The paper selection approaches are
physical objects. Moreover, they have reviewed IoT
provided in Sect. 4. The selected data management tech-
requirements that consists of operational real-time queries,
niques in IoT are studied in Sect. 5. The results and com-
indexing, and storage analytics of IoT data and continuous
parison of the selected techniques are provided in Sect. 6.
machine scale ingestion. This paper reviews two groups of
Also, Sect. 7 illustrates open issues. Finally, in Sect. 8 the
databases including relational and non-relational databases
paper concludes.
and discusses their advantages and disadvantages. In this
paper, the author has not studied related works in a sys-
tematic way.
2 Related work
Finally, Cai et al. [9] have presented a functional
framework that identifies the obtaining, management,
In this section, some survey papers that reviewed the
mining and processing aspects of big data in IoT and some
database issues in the IoT environment will be mentioned
technical modules are reviewed in the aspect of their major
and their main advantageous and disadvantageous will be
properties and characteristics. Then research in IoT appli-
outlined.
cation has been studied, in addition, the opportunities and
Cooper and James [12] have discussed database man-
challenges related to the high amount of IoT data have
agement challenges in the IoT. They have provided some
been described. They also provide a review of important
scenarios to represent the new world that will be con-
IoT application publications and research areas according
structed by the IoT, where devices are fully integrated into
to associated industry and academic publications. Also,
information. In this paper, the various types of data that
some typical samples and open issues have been provided.
will be part of the IoT were discussed. These include
However, recently published articles have not been
identification, environmental, positional, descriptive and
reviewed and open issues have not been well described.
historical data. Also, they have taken into account how to
manage high data quantities across heterogeneous systems.
In particular, they have considered the areas of querying,
process modeling, indexing, transaction handling and
3 Background
integration of heterogeneous systems. The paper refers to
IoT systems have a huge and growing number of data
the works that might provide solutions for these problems.
sources including sensors, mobile devices, RFIDs and
Finally, a road map for the IoT and technical priorities
embedded systems [39, 50]. Data are produced by millions
were discussed. However, the papers selection mechanism
of these data sources connected over a network and hence
is not clear and old papers are reviewed.
the volume of generated data are high and heterogeneous
Gurav and Kudale [23] have focused on different kinds
[23, 62]. Database system should be able to handle all
of data in IoT, data management and its problems in the
kinds of unknown data. As time passes, IoT will find a
IoT environment. Database for IoT (DoT) demands is
more prominent role in telemetry. Telemetry involves the
different than the existing database demands. DoT should
wireless transmission of small amounts of data such as
be reliable, fast and scalable. Wireless Sensor Network
environmental conditions. Telemetry data are often
(WSN) data management solutions are applied to IoT but
unstructured and small. The same is true not just about
WSN is just a part of IoT and it requires other solutions for
telemetry, but about other forms of data as well [18]. As it
data management. They have described the data lifecycle
is shown in Fig. 1, most of the data that is transmitted over
within an IoT environment. Some data management chal-
the internet is now unstructured.
lenges such as high volume, velocity, diversity, integration
It is necessary that a system that is created to support
and real-time data processing have been reviewed. Also,
IoT data storage meets some requirements. First, it should
this paper considers requirements of IoT database and
be able to identify each object uniquely. Second, it should
focuses on DoT selection by considering device type and
allow the data to be transferred using a variety of protocols.
its abilities. They only have discussed IoT data type and
IoT devices can communicate using a number of protocols
data management issue and have not reviewed related
like hypertext transfer protocol (HTTP) or constrained
papers.
application protocol (CoAP). Third, it should be a RESTful
web service. The RESTful architecture allows for a

123
Cluster Computing

60000 ability to process the data and respond in real-time puts

50000 an extra load on the data infrastructure.
Exabytes (billions of GB)

Data integration IoT data have multiple use cases.

40000
Device data may be used operationally and also it may
30000 be leveraged to glean analytic insight regarding trends
20000
and for predictive goals. Typically, these requirements
need various tools and introduce a critical data integra-
10000
tion issue.
0
200620072008200920102011201220132014201520162017201820192020
Year
Structured Data Unstructured Data 4 Research method
Fig. 1 The growth of structured versus unstructured data over the past
This section provides a systematic literature review (SLR)
decade [41]
to advance the understanding of the data management
uniform way to access and modify a resource. Fourth, it mechanisms in the IoT. The SLR is a significant evaluation
should allow for real-time triggers. As IoT is about real- and analysis of all investigation that deals with a specific
time data, there should be a way to trigger actions when challenge [40]. In this section, it is used to achieve a
certain conditions are met. For example, it sends an email comprehensive review of the data management mecha-
when the temperature drops to freezing. Fifth, it should nisms in IoT. Then, the validity of the study collection
provide a security mechanism like a password to protect method is evaluated as given below. The following sub-
the data. Finally, it should handle a big number of con- sections include the search process like paper selection
current, I/O-bound requests [23]. Four important factors of process and classification.
databases have been described as follows:
4.1 Question formalization
• Storage cost: it refers to the cost of storing data in
storage systems. The propose of this study is identifying, analyzing, and
• Query speed: it indicates the response time of queries. reviewing all related works that have evaluated in data
• Scalability: it is the capability of a network, or system management in the IoT. Particularly, this research will
to control an increasing amount of work or its potential consider the key characteristics and mechanisms and their
to be enlarged to compatible that development. features. Another aim of this research is understanding the
• Flexibility: it is the quality of being able to change major challenges and problems in term of data manage-
easily based on the condition. ment. The following research questions are considered to
By knowing the purpose of IoT and the type of devices provide a precise examination of the selected mechanisms.
used in the IoT, the following challenges are considered: Q1: What is the importance of data management in the
Data diversity IoT includes multiple connected objects IoT?
and the data requirements are described by weaving This question deals with the importance, effectiveness,
together disparate data sources. It needs the integration and role of the data management mechanisms in the IoT.
of customer data, device data, information, web services Q2: What is the data management? And what are the
data and more. open issues?
Data volume and velocity The two key properties of the This question was addressed in Sect. 3 and the open
data are velocity and volume. Devices produce a massive issue will be provided in Sect. 7.
amount of data that must be processed, assessed for trend Q3: How is the paper searching and selecting to evaluate
data and anomalies and used to trigger different oper- the data management in the IoT?
ations. Data velocity is the frequency of incoming data The question will be answered in Sect. 4.
that needs to be processed and puts a high load on data Q4: What are the groups of research mechanisms?
management mechanisms. This question will be discussed in Sect. 4.
Real-time data processing Real-time data processing is Q5: How have researchers conducted the research?
the execution of data in a short time period, providing Section 8 will discuss the answer.
near-instantaneous output. IoT system has the ability to
respond to a user’s actions, more or less in real-time. The

123
 Cluster Computing

4.2 Article selection process IEEE Springer Science Direct Sage Wiley Scienfic ACM

The paper selection process consists of three major steps as 8%

follows: 3%
2%
2%
4.2.1 Stage 1: automated search based on keywords

Since Google Scholar holds the most related databases, it

17%
was selected as the search engine. Some keywords were
used to find related papers including (IoT database, data- 57%
base Internet of Things, data management IoT, data man-
agement Internet of Things, IoT storage) and the papers
resulting from the initial search were refined in three steps 11%
to identify the papers for this review. So, 412 papers are
found from the books, conference and journal papers using
an automatic search.
Fig. 3 Publication and relevant analysis (after Stage 2)
4.2.2 Stage 2: selection based on the title, abstract
and quality of the publisher 50
45

Stage 2 starts by setting important criteria to ensure that in Number of published arcles 40
35
the review, only qualified publishers are selected. The
30
reports, working papers, commentaries, editorial notes and
25
survey papers and papers which are written in other lan- 20
guages than English are removed in this study. By means of 15
this strategy, 198 articles are found for analysis. The dis- 10
tribution of the remained papers by year of publication is 5
represented in Fig. 2. It represents the relevant papers from 0
2010 2011 2012 2013 2014 2015 2016 2017 2018
2010 to 2018. In 2018, the published papers have the lar- IEEE Springer Science Direct Sage Wiley ACM Scienfic
gest number. Figure 3 represents the classification of the
papers by selected publishers. Eventually, Fig. 4 represents Fig. 4 Distribution of articles by the publisher (after Stage 2)
the distribution of the papers over years in each class,
including Springer, IEEE, Elsevier, Scientific, ACM, Sage, article and published before 2012 were excluded. In the
and Wiley. end, the proposed method in each paper has been evaluated
and 42 papers remained that their method is directly related
4.2.3 Stage 3: publisher and relevant analysis to the data storage. Table 1 represents details of the
selected papers such as publisher, year and authors.
In the previous stage, 198 articles at searching are founded.
In Stage 3, 156 articles which are conference or review
5 Review of the selected mechanisms
60

In this section, ten selected articles based on the considered

50
Number of published arcles

criteria will be analyzed. Moreover, their approaches and

40 differences, as well as their basic properties, benefits and
weaknesses, will be described and discussed. The reviewed
30
methods can be categorized into three classes, including
20 SQL database, NoSQL database, and Graph database as
shown in Fig. 5. This category is chosen because these are
10
the three most common types of databases and all data-
0 bases can be placed in one of these categories.
2010 2011 2012 2013 2014 2015 2016 2017 2018
Year

Fig. 2 Number of published articles from 2010 to 2018

123
Cluster Computing

Table 1 Details of the selected papers (after Stage 3)

SQL Database
Publisher Year Paper (Fazio, Celesti, Villari, & Puliafito, 2014)
(Ding, Gao, Xu, & Wu, 2013)
IEEE 2012 Li et al. [34]

Database categories
2013 Ding et al. [15]
2014 Jiang et al. [29] NoSQL Database
2014 Fazio et al. [16] (Abbasi et al., 2017)
(Chen, Chen, Wu, & Fu, 2014)
2014 Chen et al. [11]
(Li, Liu, Tian, Shen, & Mao, 2012)
2015 Ning et al. [37]
(Paethong, Sato, & Namiki, 2016)
2016 Paethong et al. [38]
2016 D’silva et al. [13]
2016 Ueta et al. [52] Graph Database
2017 Ali-Ali et al. [3] (D'silva, Thakare, & Bharadi, 2016)
2017 Tripathi et al. [51] (Ueta, Xue, Nakamoto, & Murakami, 2016)
(Shafagh, Burkhalter, Hithnawi, & Duquennoy, 2017)
2018 Babar et al. [6]
(Tripathi, Sharma, & Rajvanshi, 2017)
2018 Wang et al. [57]
2018 Fu et al. [17]
Fig. 5 IoT database categories
2018 Zhao et al. [63]
Elsevier 2015 Jiang et al. [28]
SQL is readily available in the developer community and
2015 Gonizzi et al. [20]
there are many resources available to the help developers
2016 Kim and Jeong [31]
learn and use the technology. They support high integrity
2016 Bokefode et al. [8]
and even possibilities of data normalization. In SQL
2016 Urbanczyk and Peter [53]
databases, subsets of data can be accessed immediately and
2017 Gu et al. [21]
fast through indexes. On the other hand, IoT data coming
2017 Tao et al. [47]
from a variety of sensors and devices might imply a broad
2017 Terroso-Saenz et al. [49] spectrum of different schemas, which depending on the use
2017 Sood et al. [46] case rapidly changes. SQL databases are hard to scale. Its
2017 Ku [33] storage cost is high because SQL databases need very
2017 Yang et al. [61] reliable storage systems.
2017 Witkowski [58]
2017 Teing et al. [48] 5.1.1 Review of the SQL databases
2018 Yang et al. [60]
2018 Din and Paul [14] Fazio et al. [16] have presented an architecture with two
2018 Liono et al. [35] layers using a hybrid storage system which is able to
Springer 2017 Karkouch et al. [30] support a Platform as a Service (PaaS) federated cloud
2017 Alelaiwi [4] approach. The proposed method has the advantages of both
Scientific 2012 Song [45] storage systems. Specifically, on one hand, it enables
2012 Wang et al. [55] extending SQL-like legacy systems, and on the other hand,
2014 Wang et al. [56] managing big data through an XML-like and non-SQL
Sage 2015 Mishra et al. [36] distributed storage system based on a Cloud federation
2015 Ko et al. [32] scenario. They have considered that cloud uses a SQL-like
Wiley 2016 Bagci et al. [7] database for services and data management. This consid-
ACM 2017 Shafagh et al. [43] eration is very practical, assuming the wide diffusion of
IJACSA 2017 Abbasi et al. [1] SQL-like technologies. The proposed architecture enables
providers to take advantages of SQL-like benefits on the
aspect of flexibility and to develop effective services
without divesting their legacy systems. Also, it ensures
5.1 SQL databases high efficiency and stability.
Furthermore, Ding et al. [15] have proposed a general
The most common type of data storage system is a SQL statistical database cluster mechanism for big data analysis
database and for IoT data, it might be a valid storage in the IoT (IoT-StatisticDB). In IoT-StatisticDB, statistical
option. SQL databases are well-known and understood. functions are executed through statistical operators in the

123
 Cluster Computing

data base management system (DBMS) kernel, so that proposed method can integrate and extend multiple data-
complicated statistical queries are able to be shown in the bases and Hadoop to store and handle various data types
standard SQL format. So, a general statistical database generated by RFID readers and sensors. Moreover, in order
cluster mechanism is provided with operators and data to extend the Hadoop to realize a distributed file repository
types for statistical analyzing. The proposed method is a some components have developed, which can efficiently
general model that is able to perform complicated statis- process huge unstructured data. Based on the proposed
tical queries through standard SQL statements. Besides, framework, a prototype system has also developed to rep-
statistical analysis is performed in a parallel and distributed resent the effectiveness of the framework. In the proposed
way over different servers so that the performance may be data storage method, only limited kinds of adapters for
highly affected, which is confirmed by the experiments. databases are constructed. However, the complexity of the
They have not concerned with how to deal with anomaly proposed framework is high.
presence. Furthermore, Chen et al. [11] have proposed a data
storage solution based on a NoSQL database for Internet
5.1.2 Summary and comparison of the reviewed applications and mobile devices for monitoring appliance
mechanisms status in a home in real time. Also, home energy man-
agement (HEM) mobile application is developed to moni-
In this section, the selected SQL database papers are tor this information. In order to evaluate the proposed
reviewed. Table 2 represents the comparison of the most method, an HBase cluster is implemented as a testbed. The
significant advantages and disadvantages of each paper in assessment was carried out based on the developed appli-
this category. cation accessing stored data by the proposed method and it
represents proper efficiency. This application with the
5.2 NoSQL databases proposed method is able to support users be aware of the
home energy usage details and even can support other
NoSQL databases do not force the data to comply with any remote software applications/agents to deliver further
pre-defined schema and generally do not follow the rela- useful applications, for example, large scale monitoring on
tional model. Also, they do not provide tables with flat demand response. They have not considered the results
fixed-column records. They allow heterogeneous structures from those clients whose network latency are high. Also,
of data in the same domain. Therefore, each entry can have when the number of clients raises, clients will spend a long
its own schema. They are linearly scalable. But, this flex- time for a query.
ibility comes at a price. So, the programs interacting with Li et al. [34] have presented a storage management
them have to take care of these properties if needed. strategy named IOTMDB by using NoSQL to solve the
Generally, accessing data is slower than SQL databases. storage and management issues of huge IoT data as exist-
The main reason is the databases must read and parse each ing storage strategies are not well support storing hetero-
document in order to respond to a search query. geneous and huge data in IoT. In addition, the storage
solutions about organizing and expressing IoT data in a
5.2.1 Overview of the NoSQL databases uniform manner was presented and some assessments were
performed. Moreover, they not only just concerned about
IoT data volume can be huge and rapidly grows and the describing the data itself but also cared for data sharing,
data types can be diverse. In order to address these issues, therefore an ontology-based data sharing mechanism was
Abbasi et al. [1] have proposed a data storage method designed. Eventually, the index was studied and a set of
which not only can effective storing of huge IoT data but in query syntaxes to satisfy the requirements of different IoT
addition, integrates structured and unstructured files. The query types based on NoSQL was provided.

Table 2 A comparison of the SQL databases in IoT

Paper Technique Advantages Disadvantages

Fazio et al. A two-layer architecture based on a hybrid storage High stability Not suitable for big IoT data
[16] system High efficiency
Ding et al. Statistical database cluster mechanism for big data High support of complicated Not obvious how to treat with
[15] analysis in IoT statistical queries anomaly presence
High stability

123
Cluster Computing

Finally, Paethong et al. [38] have explained how to build and so on. In these type of applications, the data and
database server for IoT middleware that has low power connection between them are usually at the same level [5].
consumption and data distribution by utilizing credit card
size device, for example, Raspberry Pi with affordable 5.3.1 Overview of graph databases
price and acceptable performance. It is a platform for
interfacing with many ubiquitous sensor nodes of the IoT application has lots of different forms of things, peo-
environments like homes, farms or offices. The major ple, locations and sensors which they all have relations that
purpose is to describe the benefits and weaknesses of uti- leads to generating a complex connection and big data.
lizing credit card size device as a database server for When to model this interconnected data a relational data-
middleware and review the abilities of its usage in the base or another type of NoSQL database is utilized, their
development of the next generation IoT. The proposed performance is not suitable. Since those relational and
concept is able to present a flexible and simple database NoSQL type models are not optimized for data connection,
with the use of a NoSQL database with low power and cost they are optimized for different connection types. To solve
among cloud computing and sensing devices. this issues, D’silva et al. [13] have proposed a highly
scalable, faster, robust and pluggable architecture which
5.2.2 Summary and comparison of the reviewed attempts to overcome the problems of connected data in
mechanisms IoT domain with the unification of three open source
technologies including Neo4j, Apache spark and Ejabberd
In this section, the selected NoSQL database papers are database and to get more efficient and secure performance
reviewed. Table 3 represents the comparison of the most this three open source technologies are developed on a
significant advantages and disadvantages of each paper in Microsoft cloud Azure. However, the graph databases are
this category. poor at aggregating data.
Also, Ueta et al. [52] have considered a distributed
5.3 Graph databases graph database model for data IoT management because
graph database models are able to determine relationships
A graph database is a kind of unstructured database models in a straightforward way between information, devices, and
that use graphs with nodes and edges for storage in which users that forms an IoT system. Although, applying a graph
nodes represent entities and edges represent their rela- database to the data management of IoT systems increases
tionships and has the ability to define connections among challenges in term of distribution operations integrated
data items [42]. The graph databases can be described as with REST APIs. They have extended a graph edge prop-
those in which data structures for the schema and instances erty by adding access protocol permissions and checking
are modeled as graphs or generalizations of them and data permissions using the APIs with authentication. They have
manipulation is described by graph oriented operations and presented the demands for a use case scenario besides the
type constructors. The graph database models are applied distributed graph database properties for IoT data man-
in areas where information about topology or data inter- agement to overcome data management problems and they
connectivity is critical. Some typical use cases are location- have implemented a graph database prototype. But, the
based services, recommendation services, social networks proposed method is not efficient at processing high vol-
umes of transactions.

Table 3 A comparison of the NoSQL-based data management in the IoT

Paper Technique Advantages Disadvantages

Abbasi et al. [1] A data management framework to securely manage the huge amount of data High security High complexity
Chen et al. [11] A data storage method of NoSQL database for mobile devices and Internet Low query response High network latency
applications time High query response
time
Li et al. [34] MongoDB-based data storage High availability Low scalability
High efficiency
Paethong et al. Low-power distributed NoSQL database Low cost Low scalability
[38] Low power
consumption

123
 Cluster Computing

Shafagh et al. [43] have presented a design based on articles that are in a language other than English are
blockchain for IoT data that brings a distributed and excluded. The papers are chosen by their title and the
auditable data management and access control. They depart publishers are Elsevier, IEEE, Springer, Sage, ACM, Sci-
from the current trust model that delegates data access entific, and Wiley. According to this process, 156 papers
control to a centralized trusted authority and instead are removed. Then, the proposed method in each paper has
empower the users with data ownership. The proposed been evaluated and 42 papers remained that their method is
system enables cryptographically securing data sharing directly related to the data storage. Eventually, 10 papers
which frequently updates keys and tailored for IoT data were selected to be thoroughly reviewed. The selected
streams. The proposed method enables a resilient and papers are categorized into three groups that 2 of them are
secure access control management by using the blockchain SQL-based database, 4 of them are NoSQL-based database
as a distributed and auditable access control layer to the and 4 of them are a graph-based database.
storage layer. Moreover, it helps the storage of time-series SQL databases are the often right choice to save and
IoT data at the edge of the network using a locality-aware query high volume data. The reason for selecting this kind
decentralized storage system that is controlled with the of database is that the developers know them well and due
blockchain technology. The proposed system is agnostic of to the stability of SQL databases. Although, SQL databases
the physical storage nodes and supports as well utilization distribution to large scales is challenging. Since these type
of cloud storage resources as storage nodes. However, it of databases are proposed for availability and consistency
has low scalability. support, there is less tolerance for network partitions, so
Finally, Tripathi et al. [51] have focused on the attri- horizontally scale SQL databases is hard. [54].
butes of the tactical military attributes monitoring systems Some NoSQL DBMS stores and handles unstructured
(TMAMS) according to combining the IoT technique and data in the key-value model. The NoSQL databases have a
graph database computing methodology and use it in the free schema structure which can have some features like
battlefield terrain. They have proposed a mechanism to dynamical schema, distributed storage, horizontal scala-
query the graph database for situational awareness pattern bility and etc. Also, NoSQL database cannot work suit-
about every parameter of the soldiers and to have a better able at keeping consistency, atomicity, durability, and
decision support system for tactical advantage over the isolation of data. Also, it cannot support distributed quer-
opponents. In other mathematical issues, they have utilized ies, appropriately. In addition, in terms of data access,
graphs to store and represent data in a graphical manner. some approaches have been proposed on the potential
Traversing and searching the graph is simple in the graph integration of an ontology-based data access method in
database and therefore information retrieval becomes rel- NoSQL DBMS [10]. Thus, some related studies have
atively simpler than a relational database. It is a conceptual attempted to merge the IoT data stores with a distributed
method where the standard method of using the soldier’s file system [9].
data from the gateway to the cloud is used. They have Graph DBMS is employed graph structure for semantic
provided the case of multiple queries which is more helpful queries with nodes, edges to store and represents data
and intuitive in situational awareness on the battlefield. records. The connectivity between data sensors can be
handled effectively with graph DBMS [9]. A graph data-
5.3.2 Comparison and summary of the reviewed graph base model is designed to handle data that can be shown
database mechanisms like a graph data structure [27]. A graph database is an
appropriate way for the data management of IoT systems
In this section, selected graph database papers are studied. because the graph database is able to determine commu-
Table 4 represents the most significant advantages and nication in a straightforward way between objects like
disadvantages of each paper in this category. users, information, and devices that build an IoT system
[52].
The researchers evaluate and compare the properties that
6 Results and comparison have effects on the data management to find which prop-
erty is more important. The researchers also identify the
Three steps are performed in order to do the paper selection least and most significant properties that have effects on the
process. In the first step, Google Scholar is utilized as the data management. Table 5 gives a summary of the selected
search engine in order to find related papers for considered data management mechanisms and their major factors such
keywords. The search result is 412 papers from books, as scalability, power consumption, security, query response
conference papers, journals and any papers that includes time, stability, suitability for big data, cost and flexibility
one of the mentioned keywords. The working papers, that the effect of these factors may be hurtful or useful.
commentaries, book reviews, editorial notes, reports, and

123
Cluster Computing

Table 4 A comparison of the graph databases in IoT

Paper Technique Advantages Disadvantages

D’silva et al. An architecture to resolve the issues of High scalability Poor in aggregating data
[13] connected data High robustness
Ueta et al. [52] A distributed graph database High security Low efficiency at processing high volumes of
transactions
Shafagh et al. A blockchain-based storage Low overhead Low scalability
[43] High security High processing power and time
Tripathi et al. A conceptual approach based on the graph Low dependency on the Only appropriate for less volume of data
[51] database data type High processing power
High scalability
High flexibility

Table 5 An overview of the selected data management techniques and their major factors
Category Paper Scalability Security Query Availability Stability Suitability for Power Cost Flexibility
response big data consumption
time

SQL Fazio et al. 9 9 9 9 4 9 9 9 9

database [16]
Ding et al. 9 9 4 9 4 9 9 9 9
[15]
NoSQL Abbasi 9 4 9 9 9 9 9 9 9
database et al. [1]
Chen et al. 9 9 4 9 9 9 9 9 9
[11]
Li et al. 9 9 9 4 9 9 9 9 9
[34]
Paethong 9 9 9 9 9 9 4 4 9
et al. [38]
Graph based D’silva 4 9 9 9 9 9 9 9 9
database et al. [13]
Ueta et al. 9 4 9 9 9 9 9 9 9
[52]
Shafagh 9 4 9 9 9 9 9 9 9
et al. [43]
Tripathi 4 9 9 9 9 9 9 9 4
et al. [51]

In Table 6 performance of SQL (MySql, Oracle, and SQL databases offer users a powerful and flexible data
SQLServer), NoSQL (MongoDB, CouchDB, HBase, and retrieval mechanism which is the SELECT statement. In
Cassandra), Graph database (Neo4j, Azure Cosmos DB, this type of databases, indexes are used to accelerate data
and JanusGraph) are compared from the IoT application retrieval operations. In SQL and NoSQL, database indexes
perspective. Also, Table 7 provides information about serve the same purpose and that is for more optimized and
selecting the most efficient database for different quicker retrieval of data. But, they do this in different
applications. ways, because of different database architectures and dif-
After studying database categories, we also compared ferences how data is stored in database engine. While in
SQL, NoSQL, and Graph databases using three points: data SQL indexes are in form of B-Trees which show hierar-
retrieval, scalability, and robustness. chical structure of relational data, in NoSQL databases they
Faster data retrieval feature will be required when user point to documents or parts of documents which, in gen-
has to fetch data from database for further processing. The eral, have no relations among them. Meanwhile, graph

123
 Cluster Computing

Table 6 Performance comparison of SQL, NoSQL and Graph databases from the IoT perspective
Category Advantages Disadvantages

SQL They perform better in update, query with non-key attributes Whenever data becomes more and more connected and large in
databases and for aggregate queries size, SQL databases show worse performance than NoSQL and
Graph database
NoSQL They perform faster in insert, update and simple query when For NoSQL deployment, Security is a limiting factor
databases data amount is large Not all NoSQL databases are good at automating the process of
They have dynamic schema which are valuable in spreading a database across multiple nodes. Therefore, it cannot
developing a big data application scale up or down automatically
Graph They support complex data queries which are also performed For an IoT application, if just a simple sensor is connecting to the
databases in a relatively smaller period of time compared to other cloud and logging the data, a graph database is not well suited
databases

Table 7 Selecting the most efficient database for different adding resources such as memory or processors to the same
applications node. In horizontal scalability, number of nodes (servers) is
Category Application increased so as to share the system load. For new IoT
applications which are in developing stage, using a data-
SQL database • Data analysis base that has future expansion scope will be a practical
• Quality assurance choice. This will allow expansion of resources as and when
• Banking systems required by the IoT application and no need for high initial
• Systems administration investment. On the other hand, graph databases can scale
• Software engineering more naturally to large data sets as they do not typically
• Computerized medical records need costly join operations.
• Online shopping Robustness is concerned with the ability of the database
NoSQL database • Big data to deal with errors during execution. SQL databases is
• Real-time web applications known for their robustness, but many questions still arise
• Mobile applications when such a topic is discussed in the context of NoSQL.
• Internet of things From one point of view, if one might consider that NoSQL
• Digital communications databases are more robust due to their replication, from
Graph database • Fraud detection another, lack of code maturity and extensive testing might
• Real-time recommendation engines make NoSQL less robust in general. As cluster scale and
• Master data management workloads increase, robustness becomes a more evident
• Network and IT operations problem. Lack of robustness is a question in big data
• Identity and access management scheduling platforms and, in particular, in the NoSQL
databases. Meanwhile, Neo4j is the world’s leading graph
based database and it is a robust and proven database
technology with native graph support, which is used in
databases are built to handle highly interconnected datasets mission critical systems worldwide.
and therefore it is the most efficient when it comes to the
storage and data retrieval. Moreover, it allows for data
retrieval at deeper levels of abstraction. These type of 7 Open issues
databases simply provides speedy data retrieval for con-
nected data. In IoT applications, the massive data are generated from
Database scalability is the ability to scale out or scale up heterogeneous devices like traffic sensors, vehicles, pas-
a database to allow it to hold increasing amounts of data sengers, drivers, geological sensors and medical devices. It
without sacrificing performance. One of the main differ- results in heterogeneous sensing data forms such as text,
ences between NoSQL and SQL is that NoSQL databases voice, and video. Heterogeneous data processing gets sig-
are considered to be more scalable than SQL databases. nificant issues and moreover provides new opportunities
SQL database supports vertical scalability while NoSQL and advantages for system improvement. IoT systems may
supports horizontal scalability. Vertical scalability refers to have a lot of sensor types whose data are heterogeneous in
the ability to increase the performance of single node with data structures. Each category can be classified into various

123
Cluster Computing

sensor forms. Such as a traffic sensor can include an RFID employ scale of integrity meaningfully with some standard
reader, GPS sensor, traffic loop sensor, a road condition rules and procedures. Also, data aggregation is a rapidly
sensor, video-based traffic-flow analysis sensor and so on. emerging research area. It represents one of the processing
The sampling data from many devices have dissimilar data challenges of IoT and can be used to mitigate big data
structures and semantics which increases the problems in challenges within IoT.
data processing critically.
As interconnected objects are growing, there will be lots
of data as well. Enormous semi-structured and unstructured 8 Conclusion
data are gathered from the objects in the IoT network.
Thus, disposing of these large amounts of data is a very In this study, the state-of-art data management techniques
challenging issue. High volume data from heterogeneous in the IoT are reviewed systematically. Based on the
devices with complex semantic is a significant problem for accomplished SLR in the data management techniques
IoT processing and the lack of semantic support in the from 2012 to 2018, most of the articles are published in
processing makes it more challenging for some applica- 2018. Also, the large number of them are published in well-
tions. High volume data refers to the data that its volume known conferences and journals. IEEE with 57% has the
may be terabytes or more that named big data. The data- maximum number of papers. Sage with 2%, Wiley with 2%
center requires to store the latest forms of the sampling and Scientific with 3% have the lowest published articles
data. It also requires to store past forms of the data for a within the selected publishers. The selected papers divided
period in order to data analyzing, state monitoring, and into three classes including SQL database, NoSQL data-
query processing. The data size can be big and its pro- base, and graph-based database.
cessing is a great problem. By observing the interconnected The evaluation and comparison are carried out for all
objects that send various data types and the collected data selected techniques and advantages, disadvantages and
has a huge amount, the storage datacenters which save differences of them are reviewed for future works. More-
generated data will require more power sources, energy, over, all techniques are reviewed based on important fac-
and spaces. This collected data needs to be processed and tors like scalability, power consumption, security, query
organized. response time, stability and suitability for big data, cost and
Data storage will be an issue as objects in the whole flexibility. In general, each category has some disadvan-
world will be interconnected in IoT. Controlling the gen- tages that in futures research must be optimized. The
erated data from the IoT is a practice in big data that has NoSQL database and graph-based database in comparison
three steps: data ingestion, data storage, and data analytics. to SQL database have been able to optimize scalability and
In this way, enterprises must make use of some technolo- flexibility to deal with different kind of data at the same
gies such as MapReduce and Hadoop. It should be able to time. The SQL database has been able to increase stability
give adequate compute capacity, disk, and network to and pay more attention to security issues. Also, the SQL
continue with new data. The data volume cannot be han- databases are efficient in static environments that can
dled by traditional databases and goes beyond the pro- reduce the overhead of topology maintenance and stabilize
cessing capacity of the databases. Although, if there is a the network topology. Traversing and searching is simple
high volume of data, it needs various methods such as in the graph databases and therefore information retrieval
architectures, tools, and techniques with the propose to becomes comparatively simpler than SQL databases. In
tackle old challenges or new problems in a better way. order to have extra control over the queries and data, the
The IoT poses unprecedented challenges for database graph as a modeling tool in the IoT world can have benefits
administrators in terms of scalability, flexibility and con- for data representation because the IoT environment has
nectivity. In order to cater to these challenges in IoT many connections among heterogeneous and homogeneous
applications, we need to employ mechanisms and frame- devices. Graph databases become more beneficial when the
works to gather, store and manage data generated in IoT stored data on the nodes become associative.
processes. In addition to this, we need analysis tools which
may help analyze the produced data for better industrial
decisions and enhancing the performance and production
of different applications. References
Moreover, data integrity and data aggregation in data
1. Abbasi, M.A., Memon, Z.A., Memon, J., Syed, T.Q., Alshboul,
centric environments are challenging. Sensing objects must R.: Addressing the future data management challenges in iot: a
collect and share only data essential to perform a required proposed framework. Int. J. Adv. Comput. Sci. Appl. 8(5),
operation and assure that data is not kept or shared indef- 197–207 (2017)
initely. Data collection and sharing mechanisms must

123
 Cluster Computing

2. Abu-Elkheir, M., Hayajneh, M., Ali, N.A.: Data management for conference on computing for sustainable global development
the internet of things: design primitives and solution. Sensors (INDIACom), 2016
13(11), 15582–15612 (2013) 19. Golsorkhtabaramiri, M., Issazadehkojidi, N., Pouresfehani, N.,
3. Al-Ali, A., Zualkernan, I.A., Rashid, M., Gupta, R., Alikarar, M.: Mohammadialamoti, M., Hosseinzadehsadati, S.M.: Comparison
A smart home energy management system using IoT and big data of energy consumption for reader anti-collision protocols in dense
analytics approach. IEEE Trans. Consum. Electron. 63(4), RFID networks. Wirel. Netw. (2018). https://doi.org/10.1007/
426–434 (2017) s11276-018-1670-y
4. Alelaiwi, A.: A collaborative resource management for big IoT 20. Gonizzi, P., Ferrari, G., Gay, V., Leguay, J.: Data dissemination
data processing in Cloud. Clust. Comput. 20(2), 1791–1799 scheme for distributed storage for IoT observation systems at
(2017) large scale. Inf. Fus. 22, 16–25 (2015)
5. Angles, R., Gutierrez, C.: Survey of graph database models. 21. Gu, F., Ma, B., Guo, J., Summers, P.A., Hall, P.: Internet of
ACM Computi. Surv. (CSUR) 40(1), 1 (2008) things and big data as potential solutions to the problems in waste
6. Babar, M., Khan, F., Iqbal, W., Yahya, A., Arif, F., Tan, Z., electrical and electronic equipment management: an exploratory
Chuma, J.M.: A secured data management scheme for smart study. Waste Manage. 68, 434–448 (2017)
societies in industrial internet of things environment. IEEE 22. Gubbi, J., Buyya, R., Marusic, S., Palaniswami, M.: Internet of
Access 6, 43088–43099 (2018) things (IoT): a vision, architectural elements, and future direc-
7. Bagci, I.E., Raza, S., Roedig, U., Voigt, T.: Fusion: coalesced tions. Future Gener. Comput. Syst. 29(7), 1645–1660 (2013).
confidential storage and communication framework for the IoT. https://doi.org/10.1016/j.future.2013.01.010
Secur. Commun. Netw. 9(15), 2656–2673 (2016) 23. Gurav, T., Kudale, R.: DoT (Database for IoT): requirements and
8. Bokefode, J.D., Bhise, A.S., Satarkar, P.A., Modani, D.G.: selection criteria. Int. J. Comput. Appl. 159(8), 975–8887 (2017)
Developing a secure cloud storage system for storing IoT data by 24. Hamzei, M., Navimipour, N.J.: Toward efficient service compo-
applying role based encryption. Procedia Comput. Sci. 89, 43–50 sition techniques in the internet of things. IEEE Internet Things J.
(2016) 5(5), 3774–3787 (2018)
9. Cai, H., Xu, B., Jiang, L., Vasilakos, A.V.: IoT-based big data 25. Ibrahim, H., Bao, W., Nguyen, U.T.: Data rate utility analysis for
storage systems in cloud computing: perspectives and challenges. uplink two-hop internet-of-things networks. IEEE Internet Things
IEEE Internet Things J. 4(1), 75–87 (2017) J. 6(2), 3601–3619 (2018)
10. Charalampidis, P., Tragos, E., Fragkiadakis, A.: A fog-enabled 26. Jayashree, K., Abirami, R., Babu, R.: A collaborative approach of
IoT platform for efficient management and data collection. Paper IoT, big data, and smart city. In: Dey, N. (ed.) Big Data Analytics
presented at the IEEE 22nd international workshop on computer for Smart and Connected Cities, pp. 25–37. IGI Global, Hershey
aided modeling and design of communication links and networks (2019)
(CAMAD), 2017 27. Jeong, S., Zhang, Y., O’Connor, S., Lynch, J.P., Sohn, H., Law,
11. Chen, W.-C., Chen, Y.-H., Wu, C.-L., Fu, L.-C.: An efficient data K.H.: A NoSQL data management infrastructure for bridge
storage method of NoSQL database for HEM mobile applications monitoring. Smart Struct. Syst. 17(4), 669–690 (2016)
in IoT. Paper presented at the Internet of Things (iThings) 2014 28. Jiang, H., Shen, F., Chen, S., Li, K.-C., Jeong, Y.-S.: A secure
IEEE international conference on green computing and commu- and scalable storage system for aggregate data in IoT. Future
nications (GreenCom), IEEE and cyber, physical and social Gener. Comput. Syst. 49, 133–141 (2015)
computing (CPSCom), IEEE, 2014 29. Jiang, L., Da Xu, L., Cai, H., Jiang, Z., Bu, F., Xu, B.: An IoT-
12. Cooper, J., James, A.: Challenges for database management in the oriented data storage framework in cloud computing platform.
internet of things. IETE Tech. Rev. 26(5), 320–329 (2009) IEEE Trans. Industr. Inf. 10(2), 1443–1451 (2014)
13. D’silva, G.M., Thakare, S., Bharadi, V.A.: Real-time processing 30. Karkouch, A., Mousannif, H., Al Moatassime, H., Noel, T.: A
of IoT events using a Software as a Service (SaaS) architecture model-driven framework for data quality management in the
with graph database. Paper presented at the international con- Internet of Things. J. Ambient Intell. Humaniz. Comput. 9(4),
ference 2016 on the computing communication control and 977–998 (2017)
automation (ICCUBEA), 2016 31. Kim, H.-W., Jeong, Y.-S.: Efficient auto-scaling scheme for rapid
14. Din, S., Paul, A.: Smart health monitoring and management storage service using many-core of desktop storage virtualization
system: toward autonomous wearable sensing for internet of based on IoT. Neurocomputing 209, 67–74 (2016)
things using big data analytics. Future Gener. Comput. Syst. 91, 32. Ko, J., Lee, B.-B., Lee, K., Hong, S.G., Kim, N., Paek, J.: Sensor
611–619 (2018) virtualization module: virtualizing iot devices on mobile smart-
15. Ding, Z., Gao, X., Xu, J., Wu, H.: IOT-StatisticDB: a general phones for effective sensor data management. Int. J. Distrib.
statistical database cluster mechanism for big data analysis in the Sens. Netw. 11(10), 730762 (2015)
internet of things. Paper presented at the IEEE international 33. Ku, A.Y.: Anticipating critical materials implications from the
conference on green computing and communications (Green- Internet of Things (IOT): potential stress on future supply chains
Com), 2013 IEEE and internet of things (iThings/CPSCom) and from emerging data storage technologies. Sustain. Mater. Tech-
IEEE cyber, physical and social computing, 2013 nol. 15, 27–32 (2017)
16. Fazio, M., Celesti, A., Villari, M., Puliafito, A.: The need of a 34. Li, T., Liu, Y., Tian, Y., Shen, S., Mao, W.: A storage solution for
hybrid storage approach for iot in paas cloud federation. Paper massive IoT data based on NoSQL. Paper presented at the IEEE
presented at the 28th international conference on advanced international conference on green computing and communica-
information networking and applications workshops (WAINA), tions (GreenCom), 2012
2014 35. Liono, J., Jayaraman, P.P., Qin, A., Nguyen, T., Salim, F.D.:
17. Fu, J., Liu, Y., Chao, H.-C., Bhargava, B., Zhang, Z.: Secure data QDaS: quality driven data summarisation for effective storage
storage and searching for industrial IoT by integrating fog com- management in Internet of Things. J. Parallel Distrib. Comput.
puting and cloud computing. IEEE Trans. Industr. Inf. 14(10), 127, 196–208 (2018)
4519–4528 (2018) 36. Mishra, N., Lin, C.-C., Chang, H.-T.: A cognitive adopted
18. Gogawale, A., Khatib, F., Sontakke, P., Saigaonkar, S.: Database- framework for IoT big-data management and knowledge dis-
as-a-Service for IoT. Paper presented at the 3rd international covery prospective. Int. J. Distrib. Sens. Netw. 11(10), 718390
(2015)

123
Cluster Computing

37. Ning, H., Belanger, D.G., Xia, Y., Piuri, V., Zomaya, A.Y.: Guest 56. Wang, H.Z., Lin, G.W., Wang, J.Q., Gao, W.L., Chen, Y.F.,
editorial special issue on big data analytics and management in Duan, Q.L.: Management of big data in the internet of things in
Internet of things. IEEE Internet Things J. 2(4), 265–267 (2015) agriculture based on cloud computing. Paper presented at the
38. Paethong, P., Sato, M., Namiki, M.: Low-power distributed applied mechanics and materials, 2014
NoSQL database for IoT middleware. Paper presented at the Fifth 57. Wang, W., Xu, P., Yang, L.T.: Secure data collection, storage and
ICT international student project conference (ICT-ISPC), 2016 access in cloud-assisted IoT. IEEE Cloud Comput. 5(4), 77–88
39. Pashazadeh, A., Navimipour, N.J.: Big data handling mechanisms (2018)
in the healthcare applications: a comprehensive and systematic 58. Witkowski, K.: Internet of things, big data, industry 4.0–inno-
literature review. J. Biomed. Inf. 82, 47–62 (2018) vative solutions in logistics and supply chains management.
40. Pourghebleh, B., Navimipour, N.J.: Data aggregation mecha- Proced. Eng. 182, 763–769 (2017)
nisms in the Internet of things: a systematic review of the liter- 59. Wu, D., Shi, H., Wang, H., Wang, R., Fang, H.: A feature-based
ature and recommendations for future research. J. Netw. Comput. learning system for internet of things applications. IEEE Internet
Appl. 97, 23–34 (2017) Things J. 6(2), 1928–1937 (2018)
41. Rizzatti, L.: Digital data storage is undergoing mind-boggling 60. Yang, Y., Zheng, X., Guo, W., Liu, X., Chang, V.: Privacy-
growth. 14th September 2016 preserving smart IoT-based healthcare big data storage and self-
42. Robinson, I., Webber, J., Eifrem, E.: Graph Databases: New adaptive access control system. Inf. Sci. 479, 567–592 (2018)
Opportunities for Connected Data. O’Reilly Media, Inc., Newton 61. Yang, Y., Zheng, X., Tang, C.: Lightweight distributed secure
(2015) data management system for health Internet of Things. J. Netw.
43. Shafagh, H., Burkhalter, L., Hithnawi, A., Duquennoy, S.: Comput. Appl. 89, 26–37 (2017)
Towards blockchain-based auditable storage and sharing of IoT 62. Yassine, A., Singh, S., Hossain, M.S., Muhammad, G.: IoT big
data. Paper presented at the proceedings of the 2017 on cloud data analytics for smart homes with fog and cloud computing.
computing security workshop, 2017 Future Gener. Comput. Syst. 91, 563–573 (2019)
44. Shona, M., Arathi, B.: A survey on the data management in IoT. 63. Zhao, X., Lucani, D.E., Shen, X., Wang, H.: Reliable IoT storage:
Int. J. Sci. Tech. Adv. 2(1), 261–264 (2016) minimizing bandwidth use in storage without newcomer nodes.
45. Song, X.H.: The study on cloud storage data management method IEEE Commun. Lett. 22(7), 1462–1465 (2018)
based on minimum access cost for internet of things. Paper pre-
sented at the applied mechanics and materials, 2012 Publisher’s Note Springer Nature remains neutral with regard to
46. Sood, S.K., Sandhu, R., Singla, K., Chang, V.: IoT, big data and jurisdictional claims in published maps and institutional affiliations.
HPC based smart flood management framework. Sustain. Com-
put. Inf. Syst. 20, 102–117 (2017)
47. Tao, M., Ota, K., Dong, M.: Ontology-based data semantic Poopak Azad received her B.S.
management and application in IoT-and cloud-enabled smart in computer engineering, soft-
homes. Future Gener. Comput. Syst. 76, 528–539 (2017) ware engineering, from Univer-
48. Teing, Y.-Y., Dehghantanha, A., Choo, K.-K.R., Yang, L.T.: sity College of Daneshvaran,
Forensic investigation of P2P cloud storage services and back- Tabriz, Iran, in 2014 and the
bone for IoT networks: bitTorrent Sync as a case study. Comput. M.Sc. in computer engineering,
Electr. Eng. 58, 350–363 (2017) software engineering, from
49. Terroso-Saenz, F., González-Vidal, A., Ramallo-González, A.P., Tabriz Branch, Islamic Azad
Skarmeta, A.F.: An open IoT platform for the management and University, Tabriz, Iran, in
analysis of energy data. Future Gener. Comput. Syst. 92, 2017. Her research interest
1066–1079 (2017) includes Cloud Computing,
50. Terroso-Saenz, F., González-Vidal, A., Ramallo-González, A.P., Internet of Thing and Data
Skarmeta, A.F.: An open IoT platform for the management and Management.
analysis of energy data. Future Gener. Comput. Syst. 92,
1066–1079 (2019)
51. Tripathi, G., Sharma, B., Rajvanshi, S.: A combination of internet
of things (IoT) and graph database for future battlefield systems.
Paper presented at the international conference on computing, Nima Jafari Navimipour received
communication and automation (ICCCA), 2017 his B.S. in software computer
52. Ueta, K., Xue, X., Nakamoto, Y., Murakami, S.: A distributed engineering, from Tabriz
graph database for the data management of IoT systems. Paper Branch, Islamic Azad Univer-
presented at the 2016 IEEE international conference on internet sity, Tabriz, Iran, in 2007; the
of things (iThings) and IEEE green computing and communica- M.Sc. in computer engineering,
tions (GreenCom) and IEEE cyber, physical and social comput- computer architecture, from
ing (CPSCom) and IEEE smart data (SmartData), 2016 Tabriz Branch, Islamic Azad
53. Urbanczyk, T., Peter, L.: Database development for the urgent University, Tabriz, Iran, in
department of hospital based on tagged entity storage following 2009; the Ph.D. in computer
the IoT concept. IFAC-PapersOnLine 49(25), 278–283 (2016) engineering, computer architec-
54. Van der Veen, J.S., Van Der Waaij, B., Meijer, R.J.: Sensor data ture, from Science and Research
storage performance: SQL or NoSQL, physical or virtual. Paper Branch, Islamic Azad Univer-
presented at the 2012 IEEE 5th international conference on cloud sity, Tehran, Iran in 2014. He is
computing (CLOUD), 2012 an assistance professor in the
55. Wang, F.Y., Li, G., Du, L. M., Han, Z.Y.: Elementary discussion Department of Computer Engineering at Tabriz Branch, Islamic Azad
on data management of the internet of things. Paper presented at University, Tabriz, Iran. His research interests include Cloud Com-
the advanced materials research, 2012 puting, Social Networks, Fault-Tolerance Software, Computational
Intelligence, Evolutionary Computing, and Network on Chip.

123
 Cluster Computing

Amir Masoud Rahmani received Arash Sharifi received B.S.

his B.S. in computer engineer- degree in computer hardware
ing from Amir Kabir University, engineering, from Islamic Azad
Tehran, in 1996, the M.S. in University, South Tehran
computer engineering from branch, Iran in 2005. He also
Sharif University of technology, received M.Sc. and the Ph.D.
Tehran, in 1998 and the PhD degree in Artificial intelligence
degree in computer engineering from the Science and Research
from Islamic Azad University Branch, Islamic Azad Univer-
(IAU), science and research sity, Tehran, Iran in 2008 and
branch, Tehran, in 2005. He is a 2012, respectively. He is cur-
full professor in the department rently an Associate professor in
of computer engineering at the the Department of Computer
IAU, science and research Engineering at Islamic Azad
branch. He is the author/co-au- University, Science and
thor of more than 200 publications in technical journals and confer- Research Branch. His research interests include Machine Learning,
ences. His research interests include Internet of Things, Fog and Fuzzy Systems, Artificial Neural Networks, and Swarm Optimization.
Cloud computing, Wireless networks and Evolutionary computing.

123