Content uploaded by Naglaa Saeed
Author content
All content in this area was uploaded by Naglaa Saeed on Jun 21, 2020
Content may be subject to copyright.
Int. J. Advanced Networking and Applications
Volume: 11 Issue: 05 Pages: 4423-4428(2020) ISSN: 0975-0290
4423
Big Data with Column Oriented NOSQL
Database to Overcome the Drawbacks of
Relational Databases
NaglaaSaeedShehata
Faculty of Computers &artificial intelligence,Helwan University, Cairo,egypt
Email:nagla_sd@yahoo.com
Amira Hassan Abed
Egyptian Organization for Standardisation & Quality, Cairo,egypt
Email:Mirahassan61286@gmail.com
---------------------------------------------------------------------- ABSTRACT-----------------------------------------------------------
Due to the Era of Big Data with the large amount of distributed databases in the web and the rapid growth in the
smart systems a rapid growth happening in database models and the relational database fails to dealing with such
a big amount of data and have many limitations the need to new technologies comes up, which makes DBMS
developers move towards column oriented NOSQL database. The main goal of this paper is to provide a survey on
NOSQL Model especiallya column oriented NOSQL database, providing the user with the benefit of using
NOSQL database, Instead of using the (row database) relational to overcome the drawbacks of the relational
database Model.
Keywords - Relational Databases, NoSQL, Columnar Database, BASE properties.
----------------------------------------------------------------------------------------------------------------------------- ---------------------
Date of Submission: Feb 16, 2020 Date of Acceptance: May 08, 2020
----------------------------------------------------------------------------------------------------------------------------- ----------------------
1. INTRODUCTION
The rapid growth of the web technologies and
cloud applications that changes the nature of stored
data which included social media information's,
transactions, online purchases and because of the
relational database model scalability issues, the need
of new and easier approach arises to overcome those
problems researchers provides NOSQL Model [1],
that provides a new data stores techniques rather than
the relational database tabular data store.
1.1 Relational Database
The relational Database model presented by coded
in the 1960s and "it's the model that deals with data
and organizes it into one or more tables, or we
could call it a relations, which consists of columns
and rows, defining that relation with a unique
key"[2][3] identifying each row "Primary Key " we
also called the Rows in this kind of databases as
records or tuples and Columns called "attributes",
each relation represents one entity type and rows
represent instances of that entity and the columns
representing values attributed to the instance
thatconnect between two or more tables is called a
relation and it has some characteristic as
follows[4] :
- Optional attributes (NULLs),
- Depends on defined schema.
- Use joins to aggregate related data
- Dealing with large data VOLUME and high
rate of READ (scalability(
And it has a number of advantagesas [5]:
Based on ACID
Depends on Strong consistency,
concurrency,and recovery.
Mathematical background
Using The Standard Query language (SQL)
Vertical scaling (up scaling)
But in the other hand it has some drawbacksas it can't deal
easily with a huge amount of data and the distributed
databases that contains a variety of data type (semi
structured – unstructured).
1.2 Big Data
The Big data is a new technology that deal with a huge
number of data (terabytes – zettabytes) and it can be
defined in different ways, but the 3 V’s namely volume,
variety, and Velocity are sufficient to represent the most
general characteristics of big data [6]:fig.1 showing that
the three characteristics of big data.
- Volume: which is refers to the magnitude of
data.
- Variety: that refers to data come from a
number of variety sources.
- Velocity: and it's refers to the streamed data
collected on real time [7].
There is great interest in deploying big data technology
inthe healthcare industry to manage massive sets of
diversehealth datasets such as electronic health records
and sensor data, which are increasing in magnitude and
variety due to the commoditization of electronic devices
Int. J. Advanced Networking and Applications
Volume: 11 Issue: 05 Pages: 4423-4428(2020) ISSN: 0975-0290
4424
such as mobilephones and wireless sensors. The
newfangled medical andhealthcare systems have to be
augmented with new “bigdata” computing and analysis
capabilities.[8]
Fig .1 Big Data 3v's
1.3 NOSQL Database Model
NoSQL is a non-relational databasemanagement system
that neither uses SQL query languagefor operation data
nor is based on tabular relations that areextremely good in
dealing with the large amount of datainvolved in big data.
The main concept upon whichNoSQL is based on is the
notion of distributed storage ofdata alongside to the
handling of parallel processing [27].
NoSQL is based mainly on horizontal scalability and there
are a lot of different implementations, different systems
and techniques in building a NoSQL database system.
NoSQL databases mainly differ in the way data is stored
and accessed they can be classified into many different
types for example, wide-column store, document store,
and value store each of which has its own characters and
these three categories cover most of the techniques
involved.
NoSQLis “Class of database management systems
(DBMS)” [1] which Stands for "Not Only SQL“
Which Characteristics according to[4] [20] as follows:
No fixed schema (schema less)
No joins (typical in databases operated with
SQL)
Does not use SQL as querying language
Distributed, Partition-tolerant architecture
Characteristics Of NOSQL Databases [4][22]:
NOSQL have a number of Characteristics like it's an open
free source,didn't depended on a schema which make it
easy to use and it could be considered as The most cluster
friendly[2]
Also the researchers say that it provides the users [9] with
the ability to add frequent changes to DB, some good
Solution to Impedance mismatch, also NOSQL depends
on BASE usability (Basically available, soft state,
Eventual consistency) and CAP (consistency, Availability,
Partition tolerance) .
NoSQL databases types are classified into four major
data models are [25]: which are showed in Fig.2
Key-value model
Document model
Column family model (our point of scope)
Graph model
(Each DB Model in NOSQL has its own query
language)
Fig.2 NOSQL Models
1.4 Column-Oriented Database Systems
It's called (Column - stores) [21] that referring to it stores
data in column rather than rows as in the relational
database because the column is the smallest and lowest
instance of data.It contains the data name – value, also in
NOSQL, database stores each table with different
columns, with many values that are belongs to the same
column stored contiguously, compressed, and densely
packed, unlike the traditional database systems that store
main records (rows) after each other in a continually
manner.
In this category of NoSQL database, the columns are
realized and determined in relevant to each row in state of
predefined by the table organization owned uniform sized
columns for each tuple. Such these stores introduce a two
dimension gross/aggregate organization, a key and a row
gross that is defined as a set of columns. This allows any
column to be added to any particular row, and in this case
the rows can own a lot of various columns. In other words,
each row possesses a number of different columns that
were maintained and stored. It also is able to maintain
datain tables like segments of data columns.
2. State of The Art
In (Row Database) relational the data stored as tables
containing entities which relate to each other relationally
depends on a structured data type to fit them In the
relational tables but due to the huge evolving in the web
based application [27] nowadays which contains a
different types of data "semi structured and un-structured
data" (social media – on line purchasing process and any
kind of online activity and according to that there are
significant researches to solve the problem of scale that
Big Daya 3V's
Velocity
Variety volume
NOSQL
Key-value
model
DOCUMEN
T MODEL
COLUMN
FAMILY
MODEL
Graph
model
Int. J. Advanced Networking and Applications
Volume: 11 Issue: 05 Pages: 4423-4428(2020) ISSN: 0975-0290
4425
occurred in the relational databases and connect with the
big data using NOSQL databases to overcome the
drawbacks in the (Row Database) relational which focused
especially on who use the "Column-oriented database
systems"and the result of the surveyin this filed listed in
the next few lines[11]:
2.1 NOSQL Vs. Relational Database
Researchers in [22] discuss the NOSQL database
concepts and explained how it differ from relational
database, also they mention why NOSQL database is
needed through the era of the big data also provides the
features of NOSQL data base and the no SQL model types
then focus on the consistency methods for NOSQL
database, but the paper did not mention the advantages and
disadvantage of NOSQL database or the relational
database throw their work.
[19] Presented a comparative studywhich compares
between relational and NOSQL databases they focusesin
the presented work on the processes and (features and
characteristics) constraints for each two types of
databases.This paper could be considered as a qualitative
research paperthat based on deeply analysis and detailed,
they produced their work on the latest researches that
published during the last few years, but the researcher
didn't answer the important questionthrough their work,
which is the best database solution?,also the
researcherdidn't take in consideration a number of key
points such as "flexibility, scalability, performance, query
language, security, and availability which considered as
strong and important issues".
Researchers provide in [16] strength comparison between
NOSQL and SQL databases. Focuses on the overhead, the
conclusion of their work is that the "overhead is not
related to SQL, but to other components defined main four
components such as buffer management,locking, logging,
and latching". The researcherexaminedtheir work through
the previous four components. Gettingtheir comparative
with that avoidance of the overhead, specialized with one
or more of these components, can provide a speedup of
two of them.
[23] Presents a Survey with more focuses on the
difference between Relational and Non-Relational
Database, shown the main differences between each
databases, take into consideration the advantages and
disadvantages of each database and the tools used through
each types of them but the work didn’t give a clear results
that could help the user which database type to choose to
apply their needs.
2.2 NOSQL Models
In [29] provides comparative study for NOSQL and
They explained the main concepts and the analysis of the
NOSQL database's architecture like"Mongo DB,
Cassandra, and HBase", the researcher focuses on
Cassandra as a case study and explain briefly the
performance evaluation of the other databases in the
aspects of read and write performance. But the researcher
ignores or they didn’t mention during his workto the result
of their analytical study, which should be mentioned
together as results of the study they made.
Researcher in [14]compared between relational and
NOSQL data base considering well the performance of
each databases and some technologies that is used during
his research, the researcher found that NOSQL databases
perform better than relational databases duo to the good
facility in non-relational databases, the researcher also
compared thedifferent types of NOSQL databases, testing
or checking some operations which are read, write, delete
and instantiate that are considered the main operations.
2.3 Column-Oriented Database Systems
The study in [28] provides extensive solutions to the
problem of relational DBmigration to HBase. The
researcher uses MySQL which is considered relational
DBas input to the model output with column oriented
database HBase. They extract the features of objects using
semantic enrich, encompassinginheritance, aggregation
and composition which"are represented in a New
Optimized Data Model (NODM)" .the goal of the
proposed model to store data in a column oriented
database through novelty methodfocusing onMap structure.
The model ignoresthe details of the relational database
limitations; they focus only on the goals only and how to
achieve it .
[15] Presents a Tutorial for Column-oriented where
some open research problems were discussed as by
authors column-store systems includes physical database
design, indexing techniques, parallel query execution,
replication, and load balancing, as a conclusion of this
work the authors compared between the column- stored
systems as a commercial products which exists in the real
market.
Authors in [17] provide fine comparison between
Column-Oriented Database Systems and XML their work
explained the relationships between XML compressors
and column-stores. They illustrated that a permuting XML
compressor, called XSAQCT with the DBMS back-end
has essentially the same functionality as a column-store
(while ignoring things such as SQL Joins), including in
their work a specific kind of compression, Also they test
the compression ratio achieved with the compressor they
made, experiments were performed on an XML corpus
and the test showed a very good results that make their
work strong an applicable to use instead of the XML they
also describes the existing XML compressor showing the
similar inherent between its compression technique and
column-stores.
3. Column-Oriented Database Systems
The main purpose of using the columnar database is
toreach to high performance in the operation ofread and
write data from and to stored data and get speed upin the
Int. J. Advanced Networking and Applications
Volume: 11 Issue: 05 Pages: 4423-4428(2020) ISSN: 0975-0290
4426
process returns of a query results, the column – oriented
has so many uses as in customer relationship management
(CRM) systems, data warehouses, and in any other inquiry
systems.
3.1 Advantages Of Column-Oriented DBMS[12]
[21]
Uses less disk space through usingthe self-
indexing.
Highly Data compressed.
Use lower disk space duo to the compressed
schemes
Don't read unnecessarycolumns.
Faster to perform operations avoiding
decompression costs.
3.2 NOSQL database (column oriented) vs.
relational database (row oriented)[13][24]
Table (1) (Column Oriented) Vs. (Row Oriented)
Category
Column oriented
Row oriented
Description
Is a direction system
(DBMS) that stores
information tables by
column instead of by
row
Relational DBMS that
Stores the data in
two-dimensional
table, of columns and
rows
Stored
systems
A column-oriented info
serializes all of the
values of a column
along, then the values of
the following column,
and so on
A common technique
of storing a table is to
arrange every row of
information
Benefits
suited for OLAP
like workloads (e.g.,
Data warehouses)
(INSERTs) should be
separated into columns
and compressed as
they're keep, creating it
less suited to OLTP
workloads
suited for OLTP
like workloads that
are a lot of heavily
loaded with
interactive
transactions
Compression
Allow compression
through using uniform
data type
Duo to the lake of its
uniform data type not
available in row-
oriented data.
Table 2 differentiate between relational database and
column data base [19].
Table (2): shows how the data stored in relational
(row) differ from to be stored in column data base.
The example assumes that we have three variables
sales, product and country; we need to store values
and store them in the database. In the row store each
row contains values about product and sales and
country but in column store seriously all country then
all products then sales.
4. Survey Results
After presenting and surveying a number of related papers
and researches that related to our research scoop we could
say that now:
Column-arranged associations are progressively
productive when a total should be registered over
numerous lines yet just for a remarkably littler
subset of all segments of information, since
perusing that littler subset of information can be
quicker than perusing all information.
Column-oriented organizations are a lot of
economical once new values of a column
areequipped for all rows without delay, as a result
of that column information will be written
expeditiously and replace previous column
information while not touching the other columns
for the rows.
Row-oriented organizations are additional
economical once several columns of one row are
needed at the identical time, and once row-size is
comparatively little, because the entire row will
be retrieved with one disk request.[18]
Row-oriented organizations are additional
economical once writing a new row if all of the
column information is provided at the identical
time, because the entire row will be written with
one disk.
Table.2 Difference between Column Data Store and Row Store [10]
Int. J. Advanced Networking and Applications
Volume: 11 Issue: 05 Pages: 4423-4428(2020) ISSN: 0975-0290
4427
And as result of the survey done in the filed we will
compare between NOSQL database (column oriented) and
the relational database (row oriented) some other
perspective as follows in table (3):
Table .3The Difference Between Column Oriented and Row
Oriented
Category
Column oriented
Row oriented
1-Transaction
reliability
Transaction can
occurred when
Column range from
BASE to ACID.
high transaction
reliability
2- Data Model
Many models
Depend on the
mathematics
3- Scalability
Horizontal
scalability.
Vertical scalability
4- Cloud
suitable with cloud
Not suitable
5- Big data
handling
Used mainly for big
data
Face difficult to
dealing with big
data
6-Data
warehouse
Can serve
datawarehouse
Difficult to
manage increased
data
7-Complexity
Support
differentData types
High complexity
only table/row
formula
8- Crash
Recovery
Recovery achieved
through replication
achieved through
log files and ARIS
algorithm
9- Security
Need many solution
to be secure
Very secure also
can use different
tools to be more
secured
Security in NOSQL database (column oriented) and the
relational database (row oriented): in table 4
Table 4Column Oriented and Row Oriented
Category
column oriented
row oriented
1.Authentication
Could be achieved
using some methods
and techniques
Authentication
mechanisms applied
to All relational
databases
2- Data Integrity
data integrity is not
always achieved
achieved using
ACID properties
3.Confidentiality
Difficult to be
achieved
using encryption
techniques
4. Auditing
don’t provide
auditing
mechanisms to audit
using advanced
language are
Provided
5. Client
communication
Client
communication
Security is missed.
provide secure client
communication
mechanism via
using SSL protocols
and encryption
Table (4) illustrated a brief comparison in the security
issues between the column and row databases as a result of
the survey taking in consideration the main 5 categories
Authentication, data integrity, confidentiality, auditing, and
the client communication.
5. Conclusion
Finally concluding of the work done here that column
oriented systems are used when a new data are bringing in
the data set while those data are un or semi –structured
data, and the consistency could be relaxed for a while in
the situation that the performance will come first, in the
case we choose column a huge number of user requests
can be answered with eventually consistent unlike the row
DB which is focuses on having a strong consistency but at
the cost of scale and performance speed which makes the
column oriented systems a good choice in so many fields
but we still have a great problem, that we still need
distributed DBMS that having the four main properties:
- High availability
- Consistency
- Scalability
- Fault tolerance
Which founds here is no way to achieve them together
according to the CAP theorem which makes the column
also not the best solution but the future will be using a
combination of (SQL and NOSQL)the researchers named
it a NEWSQL and it could considered as future point of
research .
REFERENCES
[1] Raj, P., &Deka, G. “A Deep Dive into NoSQL
Databases”. San Diego: Elsevier Science &Technology.,
2018.
[2] M. State, “Relational Database Management Systems
Semester - III,” Management, no. 9038, pp. 1–8.
[3] M. State, “Relational Database Management Systems
Semester - III,” Management, no. 9038, pp. 1–8.
[4] Radoev, M. (2017). “A Comparison between
Characteristics of NoSQL Databases and Traditional
Databases”. Computer Science And Information
Technology, 5(5), 149-153. doi:
10.13189/csit.2017.050501
[5] M. A. Mohamed and O. G. Altrafi, “Relational
vs .NoSQL Databases : A Survey,” vol. 03, no. 03, pp.
598–601, 2014.
[6] A. Corbellini, C. Mateos, A. Zunino, D. Godoy and S.
Schiaffino. “Persisting Big Data: The NoSQL landscape”.
Information Systems. Vol. 63, pp. 1-23. Elsevier Science,
2017.
[7] A. E. YOUNESS KHOURDIFI, MOHAMED
BAHAJ, “A new approach for migration of a relational
Database into column-oriented A NEW APPROACH FOR
Int. J. Advanced Networking and Applications
Volume: 11 Issue: 05 Pages: 4423-4428(2020) ISSN: 0975-0290
4428
MIGRATION OF A RELATIONAL DATABASE INTO
COLUMN-ORIENTED NOSQL DATABASE ON
HADOOP,” no. November, 2018.
[8] Amira Hassan. Mona Nasr, Walaa Saber “The Future
of Internet of Things for Anomalies Detection using
Thermography”, International Journal of Advanced
Networking and Applications (IJANA), Volume 11 Issue
1, pp. Pages: 4142-4149 (2019).
[9] S. Gajendran, “A Survey on NoSQL Databases,” IBM
J. Res. Dev., vol. 5, no. 8, 2014.
[10] HANA .s “row store vs column store Internet”:
http://www.saphanacentral.com/p/rowstore-vs-column-
store.html,2016 accessed on [Jan. 19, 2019].
[11] H.R.V.yawahere,Drp.pkarde&DrV.M.Thakare,
"Nosql database". International journal of evoloutionary
scientific research and technology ,Conference Paper ·
October 2017
[12] N. D. Karande, “A Survey Paper on NoSQL
Databases : Key-Value Data Stores and Document Stores,”
vol. 6, no. 2, pp. 153–157, 2018.
[13] K.Dwivedi, A., S. Lamba, C., &Shukla, S.
“Performance Analysis of Column Oriented Database Vs
Row Oriented Database”. International Journal Of
Computer Applications, 50(14), 31-34, 2012.
[14] Y. Li and S. Manoharan, “A performance comparison
of SQL and NoSQL databases” in IEEE Pacific Rim
Conference on Communications, Computers and Signal
Processing, Canada, Aug. 2013, pp. 15- 19.
[15] D. J. Abadi and S. Harizopoulos, “Column-Oriented
Database Systems ( Tutorial ) Column-oriented Database
Systems,” no. August 2009.
[16] M. Stonebraker, “SQL databases v. NoSQL
databases,” Commun.ACM Journal, vol. 53, no. 10–12,
2010.
[17] T. Corbin, T. Müldner, and J. K. Miziołek, “Column-
Oriented Database Systems and XML Compression
Column-Oriented Database Systems and XML
Compression,” no. November, 2014.
[18] M. Abdellatif, M. Salah, and N. Saeed,
“ScienceDirect Overcoming business process
reengineering obstacles using ontology-based knowledge
map methodology,” Futur.Comput.Informatics J., vol. 3,
no. 1, pp. 7–28, 2018.
[19] K. Sahatqija, J. Ajdari, X. Zenuni, B. Raufi, and F.
Ismaili, “Comparison between relational and NOSQL
databases,” no. May, 2018.
[20] M. Radoev, “A Comparison between Characteristics
of NoSQL Databases and Traditional Databases,” vol. 5,
no. 5, pp. 149–153, 2017.
[21] N. C. Ug and J. Steemann, “Column-oriented
databases,” pp. 1–34, 2012.
[22] H. Vyawahare, “NoSql Database,” no. June, 2017.
[23] C. Győrödi, R. Győrödi, and R. Sotoc, “A
Comparative Study of Relational and Non- Relational
Database Models in a Web- Based Application,” vol. 6, no.
11, pp. 78–83, 2015.
[24] N. Jatana, “A Survey and Comparison of Relational
and Non-Relational Database,” vol. 1, no. 6, pp. 1–6, 2012.
[25] A. Nayak, “Type of NOSQL Databases and its
Comparison with Relational Databases,” vol. 5, no. 4, pp.
16–19, 2013.
[26] Amira Hassan. Mona Nasr, “Diabetes Disease
Detection through Data Mining Techniques”, International
Journal of Advanced Networking and Applications
(IJANA), Volume 11 Issue 1, pp. Pages: 4142-4149
(2019).
[27]Amira H. A., “Recovery and Concurrency
Challenging in Big Data and NoSQL Database Systems”,
International Journal of Advanced Networking and
Applications (IJANA), Volume 11 Issue 04, pp. Pages:
4321-4329 (2020).
[28] Oliveira, F., Oliveira, A., & Alturas, B..Migration of
Relational Databases to NoSQL - Methods of
Analysis. Mediterranean Journal Of Social Sciences, 9(2),
227-235, 2018.
[29] T. Of and V. Of, “COMPARATIVE STUDY OF
NOSQL DOCUMENT , COLUMN S TORE D
ATABASES A ND EVALUATION OF CASSANDRA,”
vol. 6, no. 4, pp. 11–26, 2014.
