Data modeling techniques used for big data in enterprise networks

International Journal of Advanced Technology and Engineering Exploration, Vol 7(65)
ISSN (Print): 2394-5443 ISSN (Online): 2394-7454
http://dx.doi.org/10.19101/IJATEE.2020.762029
79
Data modeling techniques used for big data in enterprise networks
Richard Omollo*
and Sabina Alago
Department of Computer Science and Software Engineering, Jaramogi Oginga Odinga University of Science and
Technology, Bondo, Kenya
Received: 10-March-2020; Revised: 18-April-2020; Accepted: 20-April-2020
©2020 Richard Omollo and Sabina Alago. This is an open access article distributed under the Creative Commons Attribution
(CC BY) License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is
properly cited.
1.Introduction
Data is continuously being used to make decisions
that are transforming institutions [1]. Today,
institutions are increasingly harnessing the potentials
brought about the rapid advancements in technology.
One of such potentials includes the rapid data growth
both historical and current [2]. Currently, many
institutions are apt in making efficient use of the
tremendous data in realizing strategic organizational
objectives including: improved business, improved
organizational output and productivity among many
objectives. Therefore, this paper sets forth
underpinning issues around big data modeling [3].
Institutions are sensitive in terms of institutional data,
and this is clearly exhibited by their preparedness to
turn challenges into opportunities. Data has become
an important driver for revolutionizing institutions,
thus, a corporate asset.
*Author for correspondence
In as much as this is the case, it is also important to
note that, the data are quickly becoming complex in
terms of volume, velocity, variability, veracity,
valence, value, variety, integrity, availability,
reliability, confidentiality among other characteristics
[1, 3]. These complexities have an eventual
contribution to data modeling techniques and
technologies in terms of enabled capture, storage,
distribution, management and analysis of data [4─6].
Each organization is peculiar in terms of its data
needs. These data needs vary from one institution to
the other. And as such, database needs also vary. The
database can be defined as a collection of sharable
data that originates from any kind of digital source
[7]. The value of data as an organizational asset
cannot be downplayed. The ability to understand and
manage the value of the data must be in place. Lack
of this means data becoming a liability rather than an
asset. This absurdity is in itself a vehicle that drives
the need for an optimally functioning database
management system. Database management systems
have become ubiquitous and are fundamental tools in
the management of information. Therefore, data
modeling is an integral part of database management
systems [1, 8].
Review Article
Abstract
The deployment and maintenance of enterprise networks form the bedrock of any organization, be it government,
commercial, academic, and/or non-profit making. These networks host vast amounts of information, databases, in either
temporary mode while in transit or permanent mode while stationary. The databases are managed by the information
systems with appropriate functions that meet consumers’ needs. Databases hold varying data – structured, semi-
structured, or unstructured. Data is increasingly becoming a vital organizational asset and therefore plays a crucial role
in making organizational decisions. With growth in the internet, digital data sources have become ubiquitous. In turn,
this has seen the continued growth in the volume, variety, veracity, velocity, and value of data. Big data brings with its
data complexities that have an eventual impact on the data modeling techniques. This paper presents a review of big data
modeling techniques with a concentration of enterprise networks. We started by appreciating big data before embarking
on modeling techniques for big data.
Keywords
Data, Model, Modeling techniques, Big data, Enterprise networks, Databases.

Richard Omollo and Sabina Alago
80
Databases hold varying data – structured, semi-
structured, or unstructured. Structured data highly
adheres to a pre-defined data model and such data are
easy to aggregate, store, process and access locally
and remotely. Unlike structured data, unstructured
data have no pre-defined model and are highly
characterized by lots of ambiguities, irregularities,
making them difficult to understand using traditional
structured data models, programs and databases.
Semistructured data do not conform to either
structured or unstructured data models, but rather rely
heavily on self-describing data structure - separates
semantic elements and enforces hierarchies of data
[7]. It is therefore imperative to model data before
applying analytics to any tremendous amounts of
data. Data modeling plays an important role in big
data analytics [1].
The data model can be defined as a high-level data
description that hides low-level description storage
details. Database management systems allow user-
defined data to be stored in terms of the data model.
The data model can be viewed as concepts that
describe the structure of a database – structure of how
data is held. Database management systems can be
based on a relational, non-relational, network,
hierarchical data models amongst others [1], [8].
Notably, data modeling comes in handy when there is
a need to design data structures at various levels of
abstraction [1]. Additionally, the authors suggest that
effective database designs (logical, conceptual and
physical) are grounded on effective data modelling
efforts.
a) Objectives of the study
In many application domains, the triviality of the data
modelling techniques cannot be overlooked. Data
modeling techniques allow structures that are suitable
for representing data. This paper sought to put forth
the working definitions of big data. Agreeably, large
volumes of data are generated continuously via
diverse digital data sources and in order to obtain
newer insights, knowledge of big data definitions,
technologies and analytics is imperative. With the
quickly soaring data, organizations need to be
informed on the role of big data in achieving their
predefined functions. Data can only be regarded as an
asset only if its value to the organization can be
ascertained. This partly has explained the adoption of
diverse functional definitions of big data in varied
contemporary settings.
The second concern that the paper addresses is big
data structures and big data modelling techniques.
Data need to be well structured. Inferring data
structures behind big data storage can be somewhat
misleading, knowing big data, data structures
beforehand goes a long way in facilitating our
understanding of big data management. Different
enterprises, deal with varying sets of data. Does this
always mean they are dealing with complex models?
Lastly, the paper brings forth the natural relationship
between bid data and modelling techniques. Data can
be voluminous but what is their relationship between
data models and data volumes? It is important to have
a grounded knowledge on such inferences if they are
to be made. For instance, it is possible sensor data
can be very voluminous but with simple data
structures compared to graph structures that are quite
complex.
In summary, this paper forms a modelling point of
view on big data and proposes conformity of data to
the models developed for an enterprise network.
b) Limitations of the study
In present day, the request for big data demands for
non-relational database. Notably, we have seen that
the choice of correct database is strongly linked to
storage demands by any enterprise network. These
are points to note for any enterprise network during
its operations. However, in this paper, we have
noticed the difference between different models, but
not, the difference in the models in terms of records
held within the databases manifesting the different
models. Indisputably, databases are widely used in
enterprise networks and as such, handling of the same
should be of paramount interest for any organization.
If this is so, then modelling of the data is definitely
tagged along. Some of the things to consider while
modelling include amount of data, flexibility of the
model, budget, amount of transactions, and frequency
of transactions.
However, this being the case, we acknowledge that
this study does not show the results of different
operations that have been applied by particular
databases – relational and non-relational. This is in
terms of execution of the operations- manipulations
that are essential in simultaneous user interactions in
an enterprise network.
c) Methodological approach used in the study
Our key focus was based on big data that is resident
on enterprise networks, which appropriate data
modeling techniques aids in analyzing. We
acknowledged that there are various data modeling
perspectives, methods, and techniques that can be

81
easily misconstrued by researchers and scholars in
data science. The contribution of this paper was to
review these key concepts with a view of adopting a
systematic approach to gather and critically evaluate
the concepts around data modeling on big data that is
the main component of an enterprise network. We
started by critically analyzing various dimensions of
big data used in its definitions before we embarked
on reviewing several data models and applicable data
modeling techniques. Given that our research design
integrates a review of existing concepts and follows
an interpretivism research philosophy, we
appreciated the significance of modeling big data that
resides in enterprise networks. We acknowledge that
proper understanding data modeling techniques
adopted are beneficial to decision making in an
organization.
2.Discussions on big data and its data
modeling techniques
Although there are many technical papers on big data
modelling techniques, there is still need to
continually describe big data, data modelling, and
modelling techniques used for big data. Remember,
the current global agenda largely revolves around big
data. With this in mind, despite relatively little being
written on their deployment and / or application in
enterprise networks, it does not mean we should
continue folding our hands in this regard, but rather
continuous exploratory works on the same is called
for if the gap in big data modelling techniques is to
be bridged.
That being said, during this study, we observed
several exceptions that are worth being mentioned:
[1] provides a deeper understanding of big data
modelling techniques and points out some
developments that have been made in this regard. The
author’s contributions include the role that big data
modelling has in the management of data sources.
Another study, [2], highlights the value of big data
analytics in reforming enterprise networks. For
instance, in this case, there’s a clear demonstration on
how big data are reforming higher education,
particularly the learning activities and processes. A
related study by Ale [4] contributes to this global
agenda by noting that continuous development of
such techniques results into operational big data
models. The paper emphasizes on the need of
continuous big data analytics, newer and enhanced
technological innovations if the world is to keep
abreast with contemporary societal issues. Lastly, a
related study [7], recommends that like universities,
other enterprise networks need to adopt newer data
technologies and data models, harness their potential
so as to meet the customer needs that comes along
with proper management of big data.
In summary, newer and / or enhance big data
technologies, including data models are called for.
For this reason, we sought to contribute to this
ongoing global discourse.
a) Big data and its characteristics
The use of the term “big data” was officially adopted
in the computing field in the year 2005 when Roger
Magoulus from O’Reilly tried to describe the huge
amounts of data that could not be managed and
processed by the traditional data management
techniques because it was becoming too complex and
vast in size [9, 10]. Even though Roger has been
credited for coining the term “big data”, some
sources also claim that it had been used before by
John Mashey of Silicon Graphics in the 1990s. The
originator of the term may not be the focus of this
paper, but the root description of the term “big data”
that helps in establishing a more functional
understanding is. In our review, we found more
clarity in Roger’s description since it facilitates
clearer understanding of his definition and thus it
enabled us to view “big data” in the context of its
characteristics.
Despite Roger’s viewpoint, the term “big data’ can
still be easily associated with voluminous data, that
is, quantified data that measure in zettabytes or even
yottabytes. Many do not comprehend the fact that
there are other definitions, which are beyond stored
contents in computing devices. This point of
understanding is majorly facilitated by large data sets
that stream in storage devices within computing
networks. According to current estimates by World
Economic Forum, the amount of data in the world
stands at 44 zettabytes in the year 2020.
Undoubtedly, this means that in the coming years,
there is a possibility of having huge amounts of data
that will reside in the computing devices and
networks. However, comprehensive definition of big
data can be viewed in tandem with the incorporation
of other attributes of data as summarized in the sub-
sections below.
i. Big Data in the Three Dimensions, 3 Vs: In
reference to [11, 12], the three Vs (volume,
velocity, and variety) are considered as the core
focus when defining big data and this is in
synchrony with Madden’s [13] definition of big
data that says: it is data that is too big (volume)
coming from different sources, too fast (velocity)

82
as it must be processed quickly, and too hard
(variety) to be processed by existing tools. Doug
Laney viewed big data in three dimensions and his
description has been widely accepted by data,
scientists in the key data management challenges
[12, 14].
ii. Big Data in the Four Dimensions, 4 Vs: There is
another approach in defining big data by
expanding on the 3 Vs to 4 Vs where the aspect of
veracity (too uncertain data) is incorporated from
the 3 Vs that exists in the earlier description [15].
The four Vs was proposed by IBM scientists while
trying to explain traffic patterns and downloads of
data that end up being recorded, stored and
analyzed to enable technology and services relied
upon daily. The position fronted here is that
depending on the type of industry and
organizational enterprise networks, big data
encompass information derived from both internal
and external sources. Organizations thus opt to
leverage data to adapt their products and services
to satisfy their customers’ needs, optimize their
operations and infrastructural resources, and
enhance their new sources of revenue.
iii. Big Data in the Five Dimensions, 5 Vs: It has
been argued that there is no precise definition of
big data especially when its characteristics are
considered [16]. From the Figure 2 below,
another attribute, value, is introduced to give it an
enhanced meaning, thus the five Vs of big data.
Significantly, big data are important to
organizations because it aids in gathering, storage,
management and manipulation of vast amounts of
data for making useful decisions. This qualifies the
inclusion of “value” as the fifth attributes of big
data from the added-value of collecting data
enforces intended process or predictive analysis.
The data value is related closely to other attributes
of big data like volume and variety [17].
iv. Big Data in the Six Dimensions, 6 Vs: In an
enterprise network, big data analytics
accommodate many perspectives: business,
technology, and social. These perspectives
comprise both functional and non-functional
requirements and therefore, big data impacts both
the strategy development process and the actual
strategies developed in a number of ways [18]. In
the previous definitions of big data, the attributes
of big data remain the same in an enterprise
network apart from variability that take
consideration of the network infrastructure
especially on integration, changing data and model
[19].
v. Big Data in the Seven Dimensions, 7 Vs: We
have seen from the previous definition of big data
that by introducing more attributes, a refined
understanding is realized by applying on an
environment. Remember, we are now focusing on
big data that is non-relational, and which
traditional data management techniques and tools
cannot handle. From the last definition, we have
seen the introduction of variability, which takes
care of varying data and associated models. More
understanding of non-relational data sets like in
spatial and temporal database models that deal
with data representation calls for a refined
description. This incorporates the seventh
attribute, which is a visualization [20]. This
attribute accommodates readability and
accessibility of data presentations which require
numerous spatial and temporal parameters, and
associated relationships between them [21].
vi. Big Data in the nth
Dimensions, n Vs: As it has
been demonstrated in the previous sub-sections,
the definition of big data can continue to
accommodate more attributes till the nth
, where n
is the number of attributes incorporated. This is
still a prime research field for scholars with the
goal of ending up with a more refined definition.
We may not put a caveat on this, but encourage
more research endeavours to explore better
positions on big data definition. Nowadays, there
are many V’s considered in attempts to define big
data in a more refined way [22]. For instance, the
eight Vs of big data introduces viscosity that
addresses the concern of data sticking and/or calls
for actions. In the case of the nine Vs of big data
covers volume (size of data), velocity (speed of
generating data), variety (different types of data),
veracity (accuracy of data), vocabulary (data
models and semantics), venue (distributed
heterogeneous data), variables (dynamic data),
validity (quality of data), value (useful data) and
vagueness (confusions over the meaning of data).
Other studies have discussed the ten Vs of big data
that see the incorporation of volatility that is
concerned with the life duration of data – for how
long will data be considered valid, and for how
long should it be stored [23]? This description of
data in terms of the characteristic has also attracted
divergent views and approaches. For instance, in
[24], the researchers have introduced the
components of big data intelligence that is about

83
the set of ideas, technologies, systems and tools
that can mimic human intelligence in the
management and processing of big data. This has
been considered due to its strategic role in
improving competitiveness of business
performance because of the support it offers in
decision making e.g. the cases of Google,
Facebook [25] etc. The big data analytics have
also been considered as another key characteristic
of big data since it encompasses the process of
collecting, organizing and analyzing big data to
discover and visualize patterns, knowledge and
intelligence for decision making [26]. Like in the
case of apache Hadoop ecosystem, big data
infrastructure has also been argued to be an
important characteristic that forms the ten Vs of
big data since it captures all structures,
technologies, systems, platforms, and facilities that
serves big data processes [24].
We appreciate the fact that many researchers have
different views in approaching the understanding of
big data definitions, for instance, the 42 Vs of big
data characteristics [27]. We are also aware that there
are contrary opinions on the relevance of Vs
characteristics in regard to defining big data [28] and
thus we do not want to tie one to this line of thinking.
We acknowledge overwhelming arguments that
support the integration of the Vs characteristics in
realizing functional understanding and definitions of
big data. Despite all these, variants of big data
definitions, see Figure 1. Madden [13] definition still
presents the most stable understanding of big data,
that is, at the end of the day, it is still about how big,
how fast and how hard the data is. We concur that
more research work on a comprehensive definition of
big data can still be pursued, especially in the context
of its characteristics.
Figure 1 Growing variants of big data definitions
b) Data modeling techniques
As earlier discussed, indeed, big data increasingly
becoming a corporate asset, thus the calls for
continuously harnessing of the potentials that come
with it [2]. There are multiple digital sources of big
data, including the Internet, social networks, mobile
devices, emails, blogs, videos, organizational
transactions and other consumer interactions [29].
With this in mind, organizations, therefore need to
put in place measures in terms of power, but
functional tools to store and explore the tremendous
amounts of data. Notably, users of such tremendous
amounts of data have their expectations around
availability, reliability, confidentiality, integrity,
accuracy among other data security goals [4─6]. We
still iterate data is not gathered for the sake of it. Its
value to the organization must be determined upfront
and through at usage time. This whole process is
captured in a summarized form as illustrated in
Figure 2 below. First, identify the type of data - data
collection phase; secondly, data preparation stage,
where data is cleaned from ambiguities, anomalies,
redundancies, inconsistencies and so on; thirdly, data
modelling stage, where the evaluation of different
modelling techniques is done; fourthly, and then the
best model that suits for our work is selected. After
modelling deployment is done. Thereafter, data is
stored for immediate or later usage – meaning the
value of the data to the organization has been
properly determined. This fourth phase may be
referred to as a data deployment phase. Lastly, the
next step is a data analytics phase. Here, appropriate
big data analytics tools are deployed to help
determine data trends, make a prediction. They help
to further harness the potentials that come with big
data.
Madden
volume 3 Vs
volume,
velocity,
variety
4 Vs
volume,
velocity,
variety
veracity
5 Vs
volume,
velocity,
variety
veracity
value
6Vs
volume,
velocity,
variety
veracity
value
variability
7Vs
volume,
velocity,
variety
veracity
value
variability
visualization
Nth Vs
volume,
velocity,
variety
veracity
value
variability
visualization
other nth values
viscosity,
volatility

84
Figure 2 Big data modelling and analytics process
Data modeling can loosely be translated to mean
creating a model for storage of data in a database, or
more precisely a theoretical presentation of data
objects and association among various data objects
within a particular database. The end product of data
modeling that supports business information
infrastructure (enterprise network) is called data
model, and the expertise behind this achievement is a
data modeler. The data model can deeply be viewed
as a set of conceptual tools for describing relevant
properties of the system under consideration [30]. It
consists of three distinct but closely intertwined
components, namely;
- A notation for data structure description
- A set of constraints that data values must satisfy in
order to be considered valid, and
- A set of operations for data update and retrieval.
Data model plays significant role in enterprise
network as it bridges the gap between technical and
functional areas in an organization and thus ensures
all data objects required by the database are
represented accurately. There are majorly three
perspectives [31] of viewing data model, namely;
 Conceptual data model: the highest level of
abstraction that defines what the system contains,
and its main focus is to organize, scope and define
business concepts and rules.
 Logical data model: the medium level of
abstraction that defines how the system should be
implemented devoid of any database management
system (DBMS), and its main focus is to develop
technical map for rules and data structures, and
 Physical data model: the lowest level of
abstraction that describes how a system will be
implemented using a specific database
management system (DBMS), and mainly focuses
on the actual implementation of the database.
Data modeling is important because it guarantees a
clear representation of data, making it easier to
analyze data that is properly represented in a model.
In addition, this improves data quality that in turn
enables concerned stakeholders make data-driven
decisions that are useful for the organization. There
are several ways of achieving data modeling, some of
the methods [32] include:
 Hierarchical data model: this is a tree-like data
modeling where each of the records has a single
root or parent. The hierarchical order is used as a
physical order for storing the database.
 Network data model: Built on the hierarchical
model where it allows multiple relationships
among records. This means that it has multiple
parent records with multiple child records. Again,
each record can belong to multiple sets allowing
the model to be an effective one for conveying
complex relationships.
 Relational data model: here, data segments is
combined with the help of tables and detailed
knowledge of physical data storage is adopted by
an organization on its network. It was combined
with Structured Query Language (SQL) as its de
facto database language.
 Object-Oriented (O-O) data model: this consists
of objects each having features and methods. It can
be referred to as post-relational or hybrid database
models since it is not limited to tables though it
uses tables. Examples that suit this include
multimedia, and hypertext databases.
 Entity-Relationship data model: this model
graphically represents entities and relationships. It
Step1: Data identification
Step 4: Data
storage
Step 3: Data model
Step 2: Data preparation
[Collection, noise, cleaning,
normalization]
Step 5: Data analytics
[Prediction etc.]

85
can also be referred to as Entity-Relationship
Diagram (ERD).
 Object-Relational data model: this is also called
Extended Relational Data Model (ERDM) since it
is a hybrid - simplicity of the relational data model
with some of the advanced functionalities of the
object-oriented database model. Examples of
Database Management Systems (DBMS) include
Object/Relational DBMS (O/RDBMS), and the
language is SQL3 with interface call interfaces are
ODBC, JDBC, and proprietary call interfaces.
 Semi structured data model: This is a self-
describing data model where information usually
associated with a schema is contained within the
data [32]. By its flexible design, it contains typed
scheme [33].
 Inverted file data model: this model is built with
the inverted file structure so as to facilitate fast full
text searches with data contents being indexed as a
series of keys in a look up table.
 Flat data model: this is non-scientific data model
that simply lists all the data in a single table.
Access and manipulation is achieved when
computer reads the entire flat file into its memory.
 Record base data model: This data model is used
to specify the overall structure of the database and
has many record types. Here, each record has a
fixed number of fields with fixed length.
 Multi- dimensional data model: In this data
model, structural data usually contained in the
database schema is embedded with the data itself.
It is useful for handling web-based data sources
and also for describing interactions between
databases that do not adhere to the same schema.
 Context data model: This flexible data model
since it incorporates other data models as needed -
elements of relational, object oriented, semi-
structured and network data model are cobbled
together enabling different works to be done due to
its versality. This model supports different types of
users who differ by interactions in the database.
 Associative data model: in this data model, all the
data points are divided based on whether they
describe an entity (anything that exists
independently) or an association (something that
only exists in relation to something else). It
structures data as two sets: set of items (a unique
identifier, a name and a type), and set of links
(each with unique identifier of sources, texts and
target).
 NoSQL data models: these are non-SQL data
models that emerged in contrast to the relational
model. They include:
o Graph data model: that allows any node to
connect to any other. This is more flexible
than network data model.
o Multivalued data model: this breaks
relational model by allowing attributes to
contain list of data rather than a single data.
o Document data model: This is for handling
storage and management of documents or
semi-structured data rather than atomic data.
The methods listed in the data models above are as a
result of data modeling. These are different from
modeling techniques since each method may attract a
unique technique, for instance, there are three basic
data modeling techniques that applies to relational
and non-relational models, namely;
 Entity Relationships Diagram (ERD): this
visual technique is like the default for modeling
and design relational databases. It incorporates the
use of entities, attributes, relationships,
cardinalities and participation constraints among
other components as seen case by case. It involves
adoption of different types of symbolic notations
too e.g. Crow’s Foot, Chen etc.
 Unified Modeling Language (UML) class
diagrams: this is a standardized family of
notations for modeling and designing information
systems resident in enterprise networks [32].
Different aspects of the system are represented by
different set of notations as it applies in software
engineering, and class diagrams are like ERDs in
real world and mostly applicable in designing
classes in object-oriented programming languages
e.g. Small talk, Java etc. Each Class Diagram
incorporates classes, attributes, methods,
relationships between objects and classes, among
other components per each case. This technique is
synonymous with implementation of meta-model,
the New Data Model (NDM), when migrating
conceptual model to Object Relational Database
(ORDB) [34].
 Data dictionary: this is a tabular representation of
data sets, with its core elements being data sets and
attributes [35]. They have detailed specifications
and can complement well with ERDs. Other
expanded considerations on its elements include
items descriptions, relations between tables and
additional constraints e.g. uniqueness, default
values, value constraints or calculated columns
[32]. The concept of active data dictionary is

86
when data model and application independent part
of program code combine to serve as compatible
program shell controlled by its data model. It is
useful in the case of ERDM or ORDM [30].
From the data models presented above, we
summarily grouped them according to convergence
in their characteristics and suitable modeling
techniques as in the Table 1. We appreciated that
some data models display overlapping characteristics
and may be grouped under more than one
consideration. Our clustering was premised on the
design principles exhibited in each data model.
Table 1 Clusters of data models
No. Cluster Data models Modeling techniques
I Record Based Hierarchical, Network, Inverted File, Record based, Flat, Multidimensional, ERD, Data Dictionary
II Relational Dimensional, E-R, Enhanced E-R, ERDM ERD, Data Dictionary
III Object Based E-R, O-O, ERDM, Context, UML, ERD
IV NoSQL Graph, Multi-value, Semi Structured, Associative Data Dictionary
Day by day, traditional data modeling tools are facing
their limits, especially, when it comes to handling of
big data. As a result, newer and/or enhanced data
modeling techniques are being adopted to help model
data as they grow in volume, value, veracity and
variety [29]. In big data analytics, there are two main
categories of tools [36]:
 Univariate big data analysis tools: that engage
single variable and useful in forecasting, growth
rate and instability indices
 Bivariate big data analysis tools: that establish
the relationships between two variables as in
conjoint analysis, predictive modeling, and
 Multivariate big data analysis tools: that involve
more than two variables and more pronounced in
artificial neural network, cluster analysis,
propensity score modeling among others.
Data modeling techniques and tools come in handy in
capturing and translating complex systems designs
into easily understood representations of the data
flows and processes. They also go a long way in
creation of database blue print for actual construction
or re-engineering. It is therefore possible to deploy
multiple modeling techniques in viewing similar sets
of data to help ascertain all processes, entities,
relationships and data flows have been identified and
captured [29].
In the previous sections, we have observed that big
data goes beyond structured data as exhibited by its
characteristics. That is why we approached big data
modeling with a map onto non-relational data
modeling, where we discussed each family. Non-
relational data modeling referred to also as NoSQL
data modeling has not been well studied compared to
relational data modeling. This has been exhibited by
lack of systematic theories on the same. Databases,
NoSQL Databases, derived from this modeling
approach get identified by various number of non-
functional criteria like their scalability, performance
and consistency.
We considered approaching this non-relational data
modeling from a systematic viewpoint of NoSQL
data models thus showing the trends and
interconnections. An imaginary “evolution” of the
major non-SQL (No SQL or Not Only SQL) systems
families captured by [37] consists of, namely; key-
value stores, big table-style databases, document
databases, full text search engines, and graph
databases.
We note that the acronym NoSQL got its use in 1998
by Carlos Strozzi while naming his Strozzi NoSQL
Open Source Relational Database, a light open source
“relational” database that did not use SQL [38]. Later
in 2009, Eric Evans and Johan Oskarsson came to use
it for describing non-relational databases. The main
motivation behind the development of NoSQL was to
address web data that necessitated faster processing
of unstructured data. The NoSQL data model uses a
distributed database management system (DDBMS)
thus is quicker, uses ad-hoc approach for organizing
data, and process large amount of differing kinds of
data. Again, NoSQL databases can handle large and
unstructured data sets better than relational databases
and this is attributed to their speed and flexibility.
Therefore, NoSQL systems not only handle both
structured and unstructured data but they also process
unstructured big data quickly [39].
The non-relational databases, premised on NoSQL
data model, are built on the Consistency,
Availability, Partition (CAP) Theorem [40], which
guides the simultaneous establishment of the three
guarantees i.e. consistency, availability, and partition

87
tolerance, and thus quantifies tradeoffs between
Atomicity, Consistency, Isolation, Durability (ACID)
and Basic Availability, Soft state, Eventual
consistency (BASE) models. Consistency, for
instance, is adopted by both ACID and BASE models
[41] even though BASE model provides less
assurance in providing safe environment for
processing data compared to the ACID model. Most
NoSQL databases use the ACID constraints for
ensuring safe and consistent storage of data. On the
other hand, availability of scaling purposes, which is
an important feature of BASE data stores, enables it
to be used by aggregate stores [39, 42]. Therefore,
non-relational data storage, which is often open-
source, non-relational, schema-less and horizontally
scalable, uses BASE model for consistency. This
elasticity feature allows for rapid changes and
replication, something that has been accomplished by
designing NoSQL data storage from bottom up and
optimized for horizontal scaling [39]. It is evident
that modeling with non-relational data systems is
completely different from modeling used in relational
data systems because of its reliance on different
design philosophies. In conclusion, NoSQL systems
uses data stores optimized for specific purposes and
can be achieved in one of the four categories; key-
value, document, wide-column, and graph database.
We started by briefly describing each NoSQL System
family before exploring the characteristics, merits
and demerits for each case. This enabled us to
appreciate their suitability in adoption in different
practical scenarios of big data cases, as summarized
in Table 3.
i) key-value stores-based/database model
Description
This adopts data storage system designed for storage,
retrieval, and managing associated arrays. Unlike
relational databases, it prioritizes a variety of optimal
options when clarifying data types.
Characteristics
 It is a non-relational model.
 It consists of a pair, a unique key that can be used
to identify data and a value. This means that, a row
contains several columns each with a pair of
unique column value and column key which are
used to identify unique data [1, 29]
 It is based on a big table that stores contents and
comments from whence data can be queried
Benefits
 Largely used with social networks
 It is suitable for fast retrieval of information [43].
 It improves aggregation capabilities [29].
 It limits the requirement of formatted data for
storage. Meaning data can be primitive, string,
integer, an array or an object [3].
 There is no need for fixed data model [3].
 Can be partitioned across distributed servers
Challenges
 Focus is on the ability to store and retrieve data
rather than the structure of that data [29].
 Poor applicability to cases that requires processing
of key ranges [37].
ii) Document stores/database model
Description
This is a system designed for storage, retrieval, and
managing document-oriented information. It has
similarities to key-value store apart from the way the
data gets processed i.e. it uses internal structure of
document for identification and storage [39]. It also
advances the BigTable model by improving on values
with schemes of arbitrarily complexities, and
database managed indexes in some implementations.
Characteristics
 They are schema-less and querying is based on
multiple attribute value constraints [3].
 It extends the key-value model; key values are
stored in a structured format that the database can
understand.
 Originally intended to store traditional documents.
 Horizontal scalability and sharding across the
cluster nodes [44]
Merits
 Supports more complex data compared to
key/value based- supports secondary indexes and
multiple types of objects
 Good for content-oriented applications (with a
single query entire content can be fetched)
 Can store huge collections of textual documents
 Good also for semi structured data or de-
normalized data [43].
 Enhanced to support any domain object [43].
 Easily find and reuse information through
metadata [45].
 Quickly disseminate contents to many recipients.

88
Demerits
 Not suitable for ACID transactions [44]
iii) BigTable/column-oriented stores model
Description
Also referred to as column family databases, it
focuses on groups of columns for storing, processing
and accessing data.
Characteristics
 Data values are modeled as map-of-maps-of-maps,
namely, column families, columns, and time-
stamped versions [29].
 Column keys in Big Table get grouped together as
column families and data within a family shares a
similar data type [29].
 Within a specified column family, data is stored
row-by-row, with columns for a specified row
being stored together instead of each column being
stored individually [37].
 Horizontal Scalability [44]
Merits
 Column oriented layout is very effective to store
very sparse data as well as multi-value cell [29].
 Data within a column family can share the same
data type [29].
 Makes reads and writes more efficient in a
distributed environment because big tables are
distributed in nature [29].
 Suitable for batch- oriented data processing [43].
 Suitable for large scale data processing [43].
 Suitable for exploratory and predictive analytics
[43].
 Unlike in row-oriented database models, it has
better capabilities to manage data and storage
space [44].
Demerits
 Manage data as sections of columns rather than
rows like in relational database management
systems [43].
iv) Graph-oriented stores model
Description
It evolved from graph theory [46] which is designed
to represent entities and their relationships as nodes
and edges respectively [47]. Unlike in traditional
databases or other NoSQL stores, graph databases
embrace connected and semantically strong
relationships between large amount of data [44]. It is
considered as a side branch of evolution that origins
from the ordered-key value models [37].
Characteristics
 Some graph database exploit native graph storage
and native processing [44].
 Are schema-less non-relational databases [48]
 Data are represented as a network of nodes
(domain entities) and are connected by edges
(connections/relationships between the entities)
[10].
 Properties are represented as key-pair values - a
node is used as a unique identifier to and from
each node there are edges and these form a pair of
key values – together they are a pointer to a
relationship [29].
Merits
 Designed to handle complex relationship data –
relationships that are unpredictable and slow in
nature.
 Handy when analyzing interconnections.
 Can be used in data mining
 Can be used in mapping a document
 Adopts visual paradigm -user friendly in nature
because of their visual representation [43].
 Graph data can be queried more efficiently
because intensive joins are not necessarily required
in graph query languages [44].
 Allow one model business entity transparently
related to document database because many
implementations allow one model a value as a map
or document. [37]
Demerits
 Are slow
 Are based on identified entities
v) Natural language processing (NLP) and
ontology ecology or full text searches engines
Characteristics
 Ontology model is a classification of entities and
models the relationship among those entities
 NLP is to identify the entities and understanding of
relationship among those entities. [29].
 Models the unstructured data - Unstructured data
is stores specific format – meets grammatical rules
 Based on grammar to express thoughts and extract
information.
 Uses natural language processing text.
Merits
 Enables computers makes sense of human
language
 Making sense of natural language enhances
inferencing [29]

89
c) A comparative analysis of big data modeling
techniques
We have successfully presented, in the previous
sections, an understanding of big data and its
characteristics. We acknowledged that the definitions
of big data by various authors follow particular
patterns of integrating dimensions to enrich the
understanding of this fundamental concept. We noted
that the definitions may vary but the core of its roots
is premised on Madden [13]. This presentation
enabled us to view modeling of big data in a simpler
manner, and thus we effectively outlined the various
modeling techniques that applies for unstructured
data. It is clear for structured data, which calls for
modeling approaches synonymous with relational
databases, but becomes complex with data that are
catered for by non-relational (NoSQL) databases. In
the Table 2 below, we managed to classify various
data modeling techniques that are used in big data
cases.
After studying various data modeling techniques use
in big data for enterprise networks, we concluded
based some parameters. There have been favoured
considerations while modeling data and these apply
to big data case, namely;
 It is advisable to consider traditional techniques to
see if they can handle the big data in question
 The modeler is also required to design system or
schema that apply for the situation
 Consideration of big data modeling tools thus
choosing an appropriate one to use
 The data modeler should also focus on data that is
core to organization’s business and its philosophy
 Not to deviate from the real business of data
modeling where the main goal is quality of data
 The data modeling techniques applied should
appreciate performance, cost, efficiency and
quality
In our summary analysis of data modeling
techniques, we came up with a guideline that can be
adopted when viewing appropriate data modeling
techniques in big data for an enterprise network. We
integrated the parameters (performance, cost,
efficiency, quality) in good modeling techniques
applied in the various big data cases by ranking their
suitability under High (H), Medium (M) and Low (L).
The summary is in Table 3 below.
Our analysis in Table 3 above interpreted the
performance of the data modeling, the cost (resource
wise) of implementing the techniques, how efficient
is the technique and does it provide quality modeling
results. This was our proposal based on the
understanding of the considered parameters of data
modeling techniques. We agree that there is need for
more research work on this with empirical approach
in implementation of various data modeling on
different data sets in different computing
environment. We know that different big data sets
require special considerations and some even calls for
combination of approaches for better results.
Table 2 A comparative analysis of NoSQL data modeling techniques
Modeling techniques Specific modeling Descriptions NoSQL Model
Conceptual
(Basic Principles)
Denormalization Query processing simplification
/ optimization
Key-value, Document, BigTable
Aggregates Soft schema, complex internal
structures
Application Side Joins Design time joins vs query time
joins
Key-value, Document, BigTable,
Graph
General
(General techniques for
NoSQL
implementation)
Atomic Aggregates Guaranteed atomicity, locks,
test-&-set instructions
Enumerable Keys Generated sequential IDs, daily
buckets
Key-value stores
Dimensionality
Reduction
Multi-Dimensional data
mapping
Index Table Special tables with keys BigTable Style
Composite Key Index Multi-dimensional indexing BigTable Style
Aggregation with
Composite Keys
Records aggregation indexing Ordered key-value, BigTable
Inverted Search – Direct
Aggregation
Data aggregation, criteria
indexing
Hierarchy (index based
and mapping)
Tree Aggregation Modeling arbitrary graphs Key-value, Document
Adjacency Lists Independent Node Modeling Key-value, Document
Materialized Paths Denormalization, node Key-value, Document,Search Engines

90
Modeling techniques Specific modeling Descriptions NoSQL Model
attribution to identifiers
Nested Sets Leafs Storage, Leafs Mapping Key-value, Document
Nested Documents
Flattening: Number
Field Names
Business entities to plain
documents mapping
Search engines
Nested Documents
Flattening: Proximity
Queries
Proximity queries for
acceptable distance limitations
Search Engines
Batch Graph Processing MapReduce, Message Passing
Pattern
Table 3 A comparative ranking of NoSQL data modeling techniques
Modeling
Techniques
Specific Modeling Performance Cost Efficiency Quality
Conceptual
(Basic Principles)
Denormalization M L H H
Aggregates M H M H
Application Side Joins M M M H
General
(General techniques
for NoSQL
implementation)
Atomic Aggregates M L H M
Enumerable Keys M M M H
Dimensionality
Reduction
M H M H
Index Table H M H H
Composite Key Index H H H H
Aggregation with
Composite Keys
H M M H
Inverted Search –
Direct Aggregation
H M M H
Hierarchy (index
based and mapping)
Tree Aggregation H L H H
Adjacency Lists M M H H
Materialized Paths M L H H
Nested Sets H M H H
Nested Documents
Flattening: Number
Field Names
M M M M
Nested Documents
Flattening: Proximity
Queries
M M M M
Batch Graph Processing M L M M
3.Conclusion and future work
Indeed, new and/or enhanced data management
techniques are imperative if big data potentials are to
be optimally harnessed. New knowledge is needed
for working around big data and enterprise network
systems. Where enterprise network systems are,
interdisciplinary teams are inevitable. And as such,
 Understanding big data on enterprise network
systems perspective is key - data structure and data
models.
 Besides, any new or enhanced horizon of big data
knowledge is needed, especially with an
inclination towards enterprise networks and
computational modeling when tackling
voluminous data - user interactions with the
database.
In this paper, we have set apart the big data modeling
techniques in enterprise network systems. Data
characteristics do not necessarily infer complexities
in the data model. We also introduced another angle
of interpreting various data modeling techniques as
applicable for big data by ranking them against the
parameters and attributes.
We also have observed that it is important to identify
the logical dependencies between data entities, that is
how data is classified, organized and stored.
Appropriate data modeling techniques offer an
appropriate storage environment, thus allowing
optimal performance while meeting data security

91
goals. It would be of utmost importance if more
studies on Intelligence on data models are done.
Suppose they will be helpful in the management of
data, including those in enterprise network systems.
Acknowledgment
None.
Conflicts of interest
The authors have no conflicts of interest to declare.
References
[1] Jyothi BS, Jyothi S. A study on big data modelling
techniques. International Journal of Computer
Networking, Wireless and Mobile Communications.
2015; 5(6):19-26.
[2] Tulasi B. Significance of big data and analytics in
higher education. International Journal of Computer
Applications. 2013; 68(14):21-3.
[3] M. M. Huda, M. L. Hayun and Z. Martun. Data
modelling for big data. ULTIMA Infosys. 2015;
6(1):1-11.
[4] Ale B. Risk analysis and big data. In Safety and
Reliability 2016 (pp. 153-65). Taylor & Francis.
[5] Baumann SU, Erber IR, Gattringer MA. Selection of
risk identification instruments. ACRN Oxford Journal
of Finance and Risk Perspectives. 2016; 5(2):27-41.
[6] Chen J, Tao Y, Wang H, Chen T. Big data based fraud
risk management at Alibaba. The Journal of Finance
and Data Science. 2015; 1(1):1-10.
[7] Logica B, Magdalena R. Using big data in the
academic environment. Procedia Economics and
Finance. 2015; 33:277-86.
[8] Patel A, Patel J. Data modeling techniques for data
warehouse. International Journal of Multidisciplinary
Research. 2012; 2(2):240-6.
[9] Chaorasiya V, Shrivastava A. A survey on big data:
techniques and technologies. International Journal of
Research and Development in Applied Science and
Engineering. 2015; 8(1):1-4.
[10] Ularu EG, Puican FC, Apostu A, Velicanu M.
Perspectives on big data and big data analytics.
Database Systems Journal. 2012; 3(4):3-14.
[11] Zhu J, Wang A. Data modeling for big data. CA,
Beijing. 2012.
[12] Austin C, Kusumoto F. The application of Big Data in
medicine: current implications and future directions.
Journal of Interventional Cardiac Electrophysiology.
2016; 47(1):51-9.
[13] Madden S. From databases to big data. IEEE Internet
Computing. 2012; 16(3):4-6.
[14] Laney D. 3D data management: controlling data
volume, velocity and variety. META Group Research
Note. 2001.
[15] Saabith AS, Sundararajan E, Bakar AA. Parallel
implementation of Apriori algorithms on the hadoop-
mapreduce platform-an evaluation of literature.
Journal of Theoretical and Applied Information
Technology. 2016; 85(3):321.
[16] Hadi HJ, Shnain AH, Hadishaheed S, Ahmad AH. Big
data and five v’s characteristics. In IRF international
conference 2014.
[17] Anuradha J. A brief introduction on Big Data 5Vs
characteristics and Hadoop technology. Procedia
Computer Science. 2015; 48:319-24.
[18] Demchenko Y, De Laat C, Membrey P. Defining
architecture components of the Big Data Ecosystem.
In international conference on collaboration
technologies and systems 2014 (pp. 104-12). IEEE.
[19] Lněnička M, Máchová R, Komárková J, Čermáková I.
Components of big data analytics for strategic
management of enterprise architecture. In SMSIS:
proceedings of the 12th international conference on
strategic management and its support by information
systems 2017. Vysoká škola báňská-Technická
univerzita Ostrava.
[20] Alexandru A, Alexandru C, Coardos D, Tudora E.
Healthcare, big data and cloud computing.
Management. 2016; 1:2.
[21] Malik BH, Cheema SN, Iqbal I, Mahmood Y, Ali M,
Mudasser A. From cloud computing to fog computing
(C2F): the key technology provides services in health
care big data. In international conference on material
engineering and advanced manufacturing technology
2018 (p. 03010). EDP Sciences.
[22] Patgiri R, Ahmed A. Big data: the v's of the game
changer paradigm. In international conference on high
performance computing and communications; IEEE
international conference on smart city; IEEE
international conference on data science and systems
2016 (pp. 17-24). IEEE.
[23] Khan N, Alsaqer M, Shah H, Badsha G, Abbasi AA,
Salehian S. The 10 Vs, issues and challenges of big
data. In proceedings of the international conference on
big data and education 2018 (pp. 52-6).
[24] Sun Z, Strang K, Li R. Big data with ten big
characteristics. In proceedings of the international
conference on big data research 2018 (pp. 56-61).
[25] Sun Z, Wang P, Strang K. A mathematical theory of
big data. IEEE Transactions on Knowledge and Data
Engineering. 2017; 13(2):83-99.
[26] Sun Z, Sun L, Strang K. Big data analytics services for
enhancing business intelligence. Journal of Computer
Information Systems. 2018; 58(2):162-9.
[27] Farooqi MM, Shah MA, Wahid A, Akhunzada A,
Khan F, ul Amin N, Ali I. Big data in healthcare: a
survey. In applications of intelligent technologies in
healthcare 2019 (pp. 143-52). Springer, Cham.
[28] https://tombreur.wordpress.com/2018/12/16/the-three-
vs-of-big-data-or-four-five-seven-10-or-42/. Accessed
15 December 2019.
[29] Hashem H, Ranc D. An integrative modeling of
bigdata processing. International Journal of Computer
Science and Applications. 2015; 12(1):1-15.
[30] Mišić V, Velašević D, Lazarević B. Formal
specification of a data dictionary for an extended ER
data model. The Computer Journal. 1992; 35(6):611-
22.

92
[31] https://www.guru99.com/data-modelling-conceptual-
logical.html. Accessed 15 December 2019.
[32] Liu L. Encyclopedia of database systems. Özsu MT.
New York, NY, USA: Springer; 2009.
[33] Chakraborty S, Chaki N. A survey on the semi-
structured data models. In computer information
systems–analysis and technologies 2011 (pp. 257-66).
Springer, Berlin, Heidelberg.
[34] El Alami A, Bahaj M. The migration of a conceptual
object model com (conceptual data model CDM,
unified modeling language UML class diagram...) to
the object relational database ORDB. MAGNT
Research Report. 2(4):318-2.
[35] Ermolayev VA, Keberle NG. Active data dictionary: a
method and a tool for data model driven information
system design.2000.
[36] Ramadas, Sendhil. (2017). Big data analytics: tools
and approaches. ICAR-Indian Institute of Wheat and
Barley Research. 2017; 1-4.
[37] https://highlyscalable.wordpress.com/2012/03/01/nosq
l-data-modeling-techniques/. Accessed 20 December
2019.
[38] https://www.datastax.com/blog/2018/08/evolution-
nosql. Accessed 20 December 2019.
[39] https://www.dataversity.net/a-brief-history-of-non-
relational-databases/. Accessed 20 December 2019.
[40] https://dzone.com/articles/understanding-the-cap-
theorem. Accessed 20 December 2019.
[41] https://www.dataversity.net/choose-right-nosql-
database-application/. Accessed 20 December 2019.
[42] Pritchett D. Base: an acid alternative. Queue. 2008;
6(3):48-55.
[43] Ribeiro A, Silva A, Da Silva AR. Data modeling and
data analytics: a survey from a big data perspective.
Journal of Software Engineering and Applications.
2015; 8(12):617.
[44] Farooq H, Mahmood A, Ferzund J. Do NoSQL
databases cope with current data challenges.
International Journal of Computer Science and
Information Security. 2017; 15(4):139-46.
[45] https://www.dataversity.net/graph-database-vs-
document-database-different-levels-of-abstraction/.
Accessed 20 December 2019.
[46] Besta M, Peter E, Gerstenberger R, Fischer M,
Podstawski M, Barthels C, Alonso G, Hoefler T.
Demystifying graph databases: analysis and taxonomy
of data organization, system designs, and graph
queries. arXiv preprint arXiv:1910.09017. 2019.
[47] Ramachandran S. Graph Database Theory Comparing
Graph and Relational Data Models.
https://www.lambdazen.com/assets/pdf/GraphDatabas
eTheory.pdf. Accessed 20 December 2019.
[48] Srinivasa S. Data, storage and index models for graph
databases. In graph data management: techniques and
applications 2012 (pp. 47-70). IGI Global.
Dr. Richard Omollo, PhD, is a faculty
member and researcher in the
Department of Computer Science and
Software Engineering, Jaramogi
Oginga Odinga University of Science
and Technology, Kenya. His research
interests include Programming
Systems, Modeling, Information
Security, IP Networks and Artificial Intelligence.
Email: comolor@hotmail.com
Sabina Alago, MA, is an adjunct
faculty and researcher in the
Department of Computer Science and
Software Engineering, Jaramogi
Oginga Odinga University of Science
and Technology, Kenya. Currently, she
is working on a research study in Big
Data where she endeavors to utilize
Bayesian Methods for Big Data Security in Enterprise
Networks.

Data modeling techniques used for big data in enterprise networks

More Related Content

Similar to Data modeling techniques used for big data in enterprise networks (20)

More from Dr. Richard Otieno (6)

Recently uploaded (20)

Data modeling techniques used for big data in enterprise networks