Journal of Theoretical and Applied Information Technology
31
st
March 2017. Vol.95. No 6
© 2005 – ongoing JATIT & LLS
ISSN: 1992-8645 www.jatit.org E-ISSN: 1817-3195
1372
A NOVEL METHOD FOR EVALUATION OF NOSQL
DATABASES: A CASE STUDY OF CASSANDRA AND REDIS
1
HADI HASHEMI SHAHRAKI,
2*
TAGHI JAVDANI GANDOMANI,
2
MINA ZIAEI NAFCHI
1
Department of Computer Engineering
Shahrekord Branch, Islamic Azad University, Shahrekord, Iran
2
Department of Computer Engineering
Boroujen Branch, Islamic Azad University, Boroujen, Iran
E-mail:
1
hadihashemi_sh@yahoo.com,
2
t_javdani@azad.ac.ir, ziaei.mina@yahoo.com
ABSTRACT
In today's complex world, due to the increasing use of web-based programs, smart phones and social
networks, production rate is increasing constantly as well as volume of data. Also, companies try to provide
their services with more features in order to stay ahead of their competitors. That increases the production
of information by the users too. In this case, it seems that moving from rational databases to non-relational
databases in a necessity. With regard to the various of use cases and their requirements, also the primary
features of NoSQL databases, choosing the most appropriate NoSQL database can be big concern of
developers. The right selection of NoSQL databases will avoid to wasting time, money, and energy. On this
paper, we propose a general and structured method to help developers, customers, and managers to evaluate
NoSQL databases in the right way and facilitate the process of decision making to select one of them.
Keywords: Non-Relational Databases, Nosql Databases, Evaluation, Performance, Scalability
1. INTRODUCTION
Nowadays, the volume of data is
increasing constantly; hence the cost of relational
databases scalability will be very expensive. In
contrast, NoSQL databases are designed for
appropriate horizontal scalability and
implementation of the community hardware.
On the one hand, concept of "one size fit's it all" is
not appropriate for current application scenarios and
mainly is suitable for systems with high workload
applications. [1]. Also, according to report of
Digital Universe, the volume of data is expected to
be double in every two years over the world. So,
often systems that use RDBMS will be restricted
against of the rapid data growth. In addition, since
the emergence of RDBMS, most of the information
systems have been built based on them [1].
Furthermore, some of information systems need to
higher performance and distributed databases rather
than higher reliability [2]. Existing cloud
environments should to be supported by particular
feature including flexible schema, fault-tolerance,
simple invocations, optimum use of distributed
indexes and RAM, availability, and replication of
data over the multiple servers [3-4]. NoSQL
databases always have many goals and advantages.
One of their main goals is cost and risk reduction by
using of community servers. This is the idea used
by Google in BigTable.
The purpose of performance can be many factors
such as throughput, run time, average operation
latency, diffusion index points, being suitable for
real time applications, etc. [5].
MapReduce is one of the effective techniques in
enhancement of the performance [2]. However, it
had been used for parallel processing by Google.
Discussion, analysis, benchmark and evaluation of
NoSQL databases together with comparison of
them will help developers and customers to select
the best choice to meet their requirements [3].
Careless in choosing databases and low attention to
nature of data and environment results in to
increasing re-work for revising structure and
selection of database. Therefore, evaluation and
analysis of databases should be done carefully by
using of the best tools to save money, cost, and
resources [6]. Indeed, paying enough attention
during the evaluation process is crucial. On the
other hand, regarding to data growth on the web,
concerns about scalability, maintenance,
management, and inefficient performance of