NoSQL Databases: Critical Analysis and Comparison

Adity Gupta, Swati Tyagi, Nupur Panwar, Shelly Sachdeva Upaang Saxena
Jaypee Institute of Information Technology, India Minjar Cloud Services Private Limited, India
[email protected], [email protected], [email protected]
[email protected], [email protected]

Abstract— The current research explores and differentiates dirty data so as to perform effectively. The most common,
between various forms in which NoSQL databases exist. It well-known conception to store this data is through relational
examines the need of NoSQL and how they have become an model. Structured Query Language (SQL) extracts relevant
important alternative to relational databases. NoSQL databases data from the pool of database.
can be categorized into four major classifications which are: key
Relational databases are the most common type of database
value stores, graph databases, wide column stores, and document
stores. These categories are compared on the basis of functional because of its simplicity. In an RDBMS [26] data is broken
features and non-functional features. The non-functional features into several tables which can be accessed as per the
include performance, scalability, flexibility, structure and requirements without actually making changes in the table.
complexity. The functional features include de-normalization, Operations like join, aggregation, addition, creation, retrieval
joins, atomicity, aggregation and keys. Then for further analysis, and deletion are easily performed in relational databases and it
one database is selected from each of these categories that is, is also very easy to extend or modify existing tables.
MongoDB (document stores), Cassandra (wide column stores), Examples: SQL Server [23], Oracle Database [24] and MySQL
Redis (key value stores), and Neo4j (graph databases). Selected [25].
databases are compared on their data model, CAP theorem,
distributive properties and other factors. By performing the B. Why NoSQL Databases are used?
comparison on non-functional features, it has been found that a
The major challenge with the growing data is its non-
document store can be used if high performance, flexibility and
scalability are required and if we have represented the data in uniformity. Due to this problem, in recent years, a non-
JSON format. Column store can be used for semi structured data relational database is needed to scale the growing need of
which requires high performance and scalability. Redis is anin- industry and at the same time, must be highly efficient. This
memory store and performs exceptionally fast in the case of gave rise to NoSQL databases which are highly scalable,
single shard operation. Graph databases can be used when it efficient and can store large amount of data.
comes to highly interconnected data and continuously evolving Though RDBMS are able to manage all three kinds of data
data models. The comparison between MongoDB, Cassandra, i.e., structured, semi-structured and unstructured, but labor and
Redis and Neo4j concluded that all of them follow horizontal compromises are required to achieve efficient storage of
scaling and are schema free. Except Neo4j, others don’t have
unstructured and semi-structured data. RDBMS stores
complete ACID properties. Write and delete operations are fast
for databases MongoDB, Redis and Cassandra, whereas read structured data as it is because they are already in required
operation is comparatively slow in Cassandra. In case of Neo4j, form. But, storing the semi-structured data involves a few
REST performance is similar to MongoDB, whereas embedded is complexities. The semi-structured data needs to be converted
comparatively slow. We also discuss how these databases work in in relational data before storage. Also, in case of unstructured
a distributed environment. data the data is saved as a blob object and is not stored
directly.
Keywords—database; NoSQL; comparison; database Hence, to satisfy this non-uniformity of data a fresh thought
systems; was given to the storage of data, leading to the creation of
NoSQL (Not only SQL) Databases.
I. INTRODUCTION
NoSQL databases have emerged as an important substitute to
The recent advancements in distributed web applications and relational databases and we choose them according to features
cloud computing have generated large volumes of data which like scalability, availability, and fault tolerance. They do not
cannot be managed by single nodes systems. Thus, distributed follow the general table/row/column approach which is
storage offers the solutions that provide high availability and practiced by all RDBMSs. NoSQLs are primarily called
scalability are needed. Examples of distributed (non-relational distributed or non-relational database. They support horizontal
storage) are Dynamo by Amazon and Google’s Big Table. scalability, so to scale number of servers are increased rather
A. Relational Database than upgrading hardware of the system which happens in
RDBMS where vertical scalability is performed.
Initially, every record was maintained manually, but the
advent of technology has led to drastic changes over the years. C. Importance of NoSQL
To make maintaining data easier databases were created. A NoSQL [16][39]databases are geared towards management of
database varies from a simple text document to much more large, varied and continuously changing data sets. They are
complex databases. These databases have to be refined often used in distributed systems or cloud databases. In
periodically to remove any kind of redundant, inconsistent or

978-1-5386-0627-8/17/$31.00 2017
c IEEE 293
Authorized licensed use limited to: UNIVERSIDAD VERACRUZANA. Downloaded on October 15,2024 at 13:45:10 UTC from IEEE Xplore. Restrictions apply.
NoSQL databases rigid schemes and many other limitations processing [9]. NoSQL also supports data which cannot be
are avoided. They were initially introduced as databases to easily expressed in terms of SQL [17].
provide an alternative to the long existing relational databases.
For these NoSQL databases scalability, fault tolerance and Figure 1: Classification of relevant papers
availability are the most important deciding factors. They do
not follow the strict schema approach of RDBMSs [26].
NoSQLs have a certain edge over relational databases as they
are able to tackle big data efficiently, provide high velocity,
and can handle variety of data with varied complexities. As
they are horizontally scalable, managing them is also simpler,
which can be done by the addition of a new node to the cluster
which will handle load efficiently. To avoid failure as the data
is distributed amongst several servers, so even if one fails
others are still in working condition, and hence can easily
continue the work of the faulty node. This guarantees that
single point of failure doesn’t exist in the database and also
depicts true fault tolerance of a NoSQL database. In case of
data and function, it also enjoys the ability of built-in
redundancy.
There are four general types of NoSQL databases where
every database has its own properties: Hence NoSQL databases have been adopted as a
widespread substitute to conventional SQL databases,
x Graph database: The basis of this type of databases is especially in the scenario where we are managing extremely
graph theory. Examples: Neo4j [27] and Titan [28]. large scale of data [13]. NoSQL was developed to overcome
x Key-Value store: In this database, we store the data in the disadvantages of relational databases. Therefore, many
two parts, namely key and value. Examples: Redis companies invested into researching the field of these
[29], DyanmoDB [30], Riak [31]. databases [9]. Nowadays ACID properties can be achieved by
NoSQL databases also with the help of middleware [3].
x Column store: Here, data is stored in the form of NoSQL databases rely on the services and capabilities of the
sections of columns of data. Examples: HBase[32], underlying storage systems [8].
BigTable[18][20] and Cassandra [33].
x RDBMS/SQL
x Document database: This database is higher version of
key-value stores. Here values are saved as documents Relational databases are the most common type of
which are data in the form of complex structures (like database because of its simplicity. In an RDBMS [26] data is
JSON). Examples: MongoDB [34] and CouchDB divided into multiple tables which are usually in their
[35]. normalized form for more efficiency. While accessing data it
can be reassembled as per the requirements of the user.
CAP [19] theorem explains the limitation posed on all Structured Query Language (SQL) consists of four types of
databases. It states that anyone can pick only any two out of the queries that are data definition language (DDL), data control
three features abbreviated as CAP in which C stands for language (DCL) and data manipulation language (DML). Each
Consistency, A for Availability, and P stands for Partition one has its own set of queries which are executed to define
tolerance. The main statement of Brewer's theorem says that data i.e. create table, alter table etc. , to manipulate existing
for any shared-data system, a maximum of two properties can data as per the requirements using update, insert etc. and
be exist from these properties [36]. define the control of transaction using queries like roll back,
commit etc. respectively. A detailed comparison of NoSQL
II. LITERATURE SURVEY versus RDBMS on features such as data validity, query
The detailed summary of related papers has been presented in language, data type, data storage, schema, flexibility,
appendix (Table 5). scalability and ACID compliancy is presented [40] Generally
Figure 1 depicts that we surveyed papers from different in NOSQL, only single record transactions and an eventual
sources such as, VLDB, IEEE, ACM and SIGMOD etc. These consistency replica system are supported, where it is assumed
papers have been categorized into six groups (NoSQL, that transactions are commutative. Thus, ACID transactions
SQL/RDBMS, Redis, MongoDB, Cassandra and Neo4j) and are compromised for performance [41].
critical analysis of each group has been performed. x Document store (MongoDB):
x NoSQL:
MongoDB resides on the CP side of CAP theorem.
MongoDB supports format BSON [37] which is JSON [38]
NoSQL, Not Only SQL, is distributive data model that does not like document with dynamic schemas which make data
follow relational database guidelines. It supports huge data integration easier and faster. Some of the common features of
storage, horizontal scaling and massive- parallel data MongoDB are that it has a document-oriented storage layer and

294 2017 International Conference on Computing and Communication Technologies for Smart Nation (IC3TSN)
Authorized licensed use limited to: UNIVERSIDAD VERACRUZANA. Downloaded on October 15,2024 at 13:45:10 UTC from IEEE Xplore. Restrictions apply.
for replication of data between servers it uses asynchronous and is semi structured, then column store is the appropriate
replication [15]. In MongoDB and other NoSQL databases database as it will provide high performance and scalability. If
additional implementation decisions are made which were not data can be represented in JSON format, then document store
required in SQL databases. These decisions have an effect on should be preferred as it has high performance, flexibility and
the performance of databases[12]. Other advantages of using usually high scalability. If we need to store data which can be
MongoDB include easy replication, MapReduce, and represented using graph theory or if the data is strongly inter-
clustering [11]. related, then we use graph store model which provides high
stability, but performance and scalability is variable.
x Column store (Cassandra):
Table 2 compares the four categories of NoSQL databases
Cassandra resides on the AP side of CAP theorem. It on the basis of functional features, such as, De-normalization,
provides its users scalability as it’s linearly scalable and Single aggregate (adding multiple composite keys to a single
availability without compromising with its performance. key), Atomicity, Unordered Keys, Derived Table (a table can
Cassandra is easily capable of managing heaps of data across be created on the basis of master class this helps in sorting
number of commodity servers while maintaining high according to multi-dimensional indices), Composite Key,
availability without any single point of failure. [6][10] Composite Aggregation, Aggregation, Aggregation and Group
x Key value store (Redis): by, Adjacency Lists (each node is designed as an
individualistic record that accommodates arrays of immediate
Redis resides on the CP side of CAP theorem. Redis key- ancestors or descendants), Nested Sets and Joins.
value data store with a choice for data durability. It is an in-
memory NoSQL database, Redis supports various data Key value store should be avoided if we want to use
structure servers like strings, lists, sets, hashes and sorted sets. composite key, joins or derived table operations on the
It can be replicated using relax master slave architecture. database.

x Graph database (Neo4j): Document Store should be avoided if we want to use de-
normalization, unordered key, composite key, composite
Neo4j utilizes labeled property graph model. In Neo4j aggregation, joins or derived table operations on the database.
nodes and edges can have properties associated with them.
Nodes can be further associated with labels which categorize Wide Column store should be avoided if we want to use
different them according to their roles. Neo4j is a full ACID unordered keys, aggregation and group by, adjacency lists,
transaction compliant graph database. It can be used as both nested sets or joins operations on the database.
standalone server (REST interface) or in embedded form [22]. Graph Store should be used if we want to perform just de-
normalization.
III. COMPARITIVE ANALYSIS OF NOSQL DATABASES
Table 2: Different NoSQL databases on basis of functional features
A. Comparison of NoSQL databases on the basis of
functional and non functional requirements. S. Features Key Value Document Wide Graph
No Store Store Column Store
Table1: Different NoSQL databases on basis of Non-functional features . store
Data Perform Scalability Flexibility Structure Complexity 1. Denormalization Applicable Not Applicable Applicable
model ance of of data of schema of of values Applicable
queries database 2. Single Aggregate Applicable Applicable Applicable Not
Key- High High High Primary None Applicable
value key with
store some value 3. Atomicity Applicable Applicable Applicable Not
Applicable
Column High High Moderate row Low 4. Unordered Keys Applicable Not Not Not
Store consisting Applicable Applicable Applicable
multiple
5. Derived Table Not Not Applicable Not
columns
Applicable Applicable Applicable
Document High Variable High JSON in Low
Store (High) form of 6. Composite Key Not Not Applicable Not
tree Applicable Applicable Applicable
Graph Variable Variable High Graph – High 7. Composite Applicable( Not Applicable Not
Database entities and Aggregation ordered) Applicable Applicable
relation 8. Applicable Applicable Applicable Not
Aggregation
Applicable
Table 1 compares key-value store, column store, document 9. Aggregation and Applicable Applicable Not Not
Group by Applicable Applicable
store and graph database based on their non-functional
10. Adjacency Lists Applicable Applicable Not Not
features such as, Performance of queries, Scalability of data,
Applicable Applicable
Flexibility of schema, Structure of database and Complexity of
11. Nested Sets Applicable Applicable Not Not
values.Table1 depicts that for a simple data that can be Applicable Applicable
represented as a key-value pair form easily; key value store
12. Joins Not Not Not Not
may be chosen as it will provide high performance, scalability Applicable Applicable Applicable Applicable
and flexibility. If the value can be represented in column from

2017 International Conference on Computing and Communication Technologies for Smart Nation (IC3TSN) 295
Authorized licensed use limited to: UNIVERSIDAD VERACRUZANA. Downloaded on October 15,2024 at 13:45:10 UTC from IEEE Xplore. Restrictions apply.
 Table 3: Differentiation of Cassandra, MongoDB, Redis, Neo4J and MySQL.
S. Feature Wide Column Store Document Store Key Value Store SQL (MySQL) Graph
No. (Cassandra) (MongoDB) (Redis) Database
(Neo4j)
1 Database Wide column store Document Store Key-Value Store Relational DBMS Graph database
Model
2 Description It is one of the most It is one of the well- It is an in-memory Widely used open Open source
popular wide column known document store data structure store source RDBMS graph database
store database. It is database and an important key
based on the concept value store
of BigTable
3 DB Key space Database Database Database Graphs
4 Table Column Family Collection Hash set, List, Set, Relation Label
Sorted set and String
5 Value Rows Documents Key value pair Rows Node and edges
6 Read Slow[4] Fast[4] Fast[5] Slow (Join dependent) Data dependent
Operations
7 Write Fast[4] Fast[4] Fast[5] Slow Data dependent
Operations
8 Delete Fast [4] Fast[4] Fast[5] Slow Data dependent
Operations
9 Language Java C++ C[14] C and C++ Java, Scala
10 License Open Source Open Source Open Source Open Source Open Source
11 Data scheme Schema-free No particular schema Schema-free Yes Schema-free
is followed but usually
contents of same
documents as a
convention have
similar structures
though it is not
mandatory
12 Predefined Yes; ASCII, int, Yes; Boolean, date, Partial; data types Yes; int, float, double, Yes; Boolean,
types blob, counter, object_id, String, supported for value date, time, bit, char, byte, short, int
decimal, double, list, Integer, double. are strings, Bit enum, binary, blob, long, float,
map, set, text, arrays, hyper logs, Boolean double, char,
timestamp, varchar hashes, lists, sets, string
sorted sets, and
geospatial indexes
13 Server side No JavaScript Lua Yes Yes
scripts
14 Triggers Yes No No Yes Yes
15 Partitioning Sharding (In this Sharding with no Sharding Horizontal Partitioning
methods very large databases individual point of partitioning, sharding should be
are divided or failure with MySQL Cluster avoided in Neo4j
partitioned into or MySQL Fabric
much smaller and
manageable units
called shards)
16 Foreign Keys No Usually, not used, No Yes Yes
however equivalent
operation with DBRef
can be done
17 Transaction Atomicity and Atomic operations can Optimistic locking, ACID ACID
Concepts Isolation are be performed within atomic execution of
supported for single single document command blocks and
operations scripts
18 User Access rights for Access rights for users Simple password – Users with fine Users, roles and
Concepts users can be defined and roles based access control. grained authorization permissions
per object concepts; no user
groups or roles
19 Website cassandra.apache.org www.mongodb.org redis.io www.mysql.com www.neo4j.com

20 Developer Apache Software MongoDB, Inc. Salvatore Sanfilippo Oracle Neo Technology
Foundation
21 Initial 2008 2009 2009 1995 2007
Release

296 2017 International Conference on Computing and Communication Technologies for Smart Nation (IC3TSN)
Authorized licensed use limited to: UNIVERSIDAD VERACRUZANA. Downloaded on October 15,2024 at 13:45:10 UTC from IEEE Xplore. Restrictions apply.
B. Comparison on the basis of categories of NoSQL Feature Wide Document Key Graph
databases. Column Store Value Database
Store (MongoDB) pair (Neo4j)
In Table 3, comparison is made using features such as, (Cassandra) Store
Database Model, Description, Database, Table, Value, Read (Redis)
Operations, Write Operations, Delete Operations, Language Sharding Auto sharding Built in and Auto Supports
License, Data scheme, Predefined types, Server side scripts, and and order order sharding sharding
Partitioning preserving preserving and no but should
Triggers, Partitioning methods, Replication Methods, Scaling, order be avoided
Foreign Keys, Transaction Concepts, User Concepts, Website, Scaling Horizontal Horizontal Horizontal Horizontal
Developer, Initial Release and Current Release. Replication Selectable Master slave Relaxed Causal
The current research has taken databases from each Replication Master Clustering
category of NoSQL databases that is Cassandra, MongoDB, Factor slave using Raft
protocol
Redis) and Neo4j. Table 3 presents the differentiation of (master
various NOSQL databases with an example from each slave)
category. SQL is a Relational DBMS; Cassandra falls under In Cassandra vast quantity of data is divided across many
the category of Wide Column stores which are based on the nodes which imparts user with very high availability and
ideas of BigTable and DynamoDB. MongoDB is a Document without failure. It also supports horizontal scaling, selectable
Store, whereas Redis follows the concepts of a Key-Value replication factor and cross data center replication.
Store.
Redis is designed for in-memory data using master-slave
Cassandra has a keyspace analogous to a database in SQL architecture. Categorically, Redis supports less strict practice
and a column family instead of a table. MongoDB makes a of master-slave replication, wherein information from any
collection, while Redis has options of hashes, lists, sets and master is easily replicated to whatever number of slaves,
sorted sets instead of a table. whereas a slave itself can act as a master to other slaves. This
Read operations are slower in Cassandra and SQL database doesn’t partition data across nodes in an order
compared to the other two. For both write and delete preserving manner.
operations, SQL falls short in comparison to all NoSQL Subject to Neo4J scalability package is noted as high
databases. In case of Neo4j, even though the embedded availability. It does not support partitioning and complete
version is slow REST’s performance is roughly similar to dataset is replicated across whole cluster.
MongoDB[21]. For partitioning methods, SQL is the only one
to use Horizontal Partitioning, while the rest use sharding. IV. CONCLUSIONS
Also SQL is the only one which uses the concept of foreign
SQL databases are scale vertically (hardware) while the
keys. Coming to the transaction concepts, SQL and Neo4j
NoSQL databases are horizontally scalable (server). This paper
follow the ACID properties. For single operations, atomicity
has the aim of giving a thorough overview and introduction of
and isolation are supported in Cassandra. Atomic operations
NoSQLs, which have recently emerged in the market as an
are possible inside a long document in MongoDB, while Redis
alternative to predominant relational database management
supports optimistic locking and atomic execution of command
systems. The first half discusses the motives and rationales
blocks and scripts.
behind the development and usage of non-relational
MongoDB supports access rights for different types of management systems, while the next half categorizes NoSQLs
users. For Cassandra, access rights can be established per into types, namely, Document stores, Key-value stores and
object. Redis supports uncomplicated password based access Column based stores, and then elucidate on their models and
control [44]. In MongoDB, authorization and authentication workings. Each database performs and behaves in a different
are disabled by default. Here, the authorization is provided by manner and all of them are constantly evolving. The current
following a role-based approach on a per-database level. research has taken databases from each category that is
Provision for authentication on a per-database level has been Cassandra (wide column store), Neo4j (Graph database) Redis
made available in basic MongoDB where the users subsist (Key value pair store) and MongoDB (Document store) and
particularly for a single logical database [42]. Authorization compared them on the basis of data models, distributive
and authentication is enabled by default in Neo4j [45]. properties and other features. The research compares them on
their non-functional features. It has been found that for a
C. Comparison on the basis of distributive properties simple data that can be represented in the form of key value
Table 4 explains how the four databases work when easily, a key value store should be chosen as it will provide
database is spread on multiple computers which may or may high performance, scalability and flexibility. If the value can
not be in same physical location. In case of MongoDB auto be represented in column form, and is semi structured, then
sharding is used to partition data amongst multiple nodes in column store is the appropriate database as it will provide high
order preserving manner. MongoDB supports horizontal performance and scalability. If data can be represented in
scaling which enables it to scale data across multiple nodes. JSON format, document store should be preferred as it has
The load is distributed equally across nodes an if balance is high performance, flexibility and usually high scalability. If
disrupted it automatically redistributes the load equally. the graph theory represents the data then we use graph store
Table 4: Analysis of NOSQL databases based on distributive properties. model which provides us high stability, but performance and
scalability is variable. Following this, the comparison is made

2017 International Conference on Computing and Communication Technologies for Smart Nation (IC3TSN) 297
Authorized licensed use limited to: UNIVERSIDAD VERACRUZANA. Downloaded on October 15,2024 at 13:45:10 UTC from IEEE Xplore. Restrictions apply.
on the basis of functional features. It has been concluded that [15] Lanjun Wang, Shuo Zhang, Juwei Shi, Limei Jiao,Oktie Hassanzadeh,
Key value store ought to be avoided if one needs to use Jia Zou, Chen Wangz, “Schema Management for Document Stores”,
Proceedings of the VLDB Endowment, 2015
composite key, joins or derived table operations on the
[16] http://www.planetcassandra.org/what-is-nosql/
database. Document Store ought to be avoided if one needs to
[17] Concurrent Programming for Scalable Web Architectures
use de-normalization, unordered key, composite key, (2012),Benjamin Erb Research Assistant, University of Ulm Distributed
composite aggregation, joins or derived table operations on the Systems (Cited by 14):http://berb.github.io/diploma-
database. Wide column store should be avoided if we want to thesis/original/061_challenge.html
use unordered keys, aggregation and group by, adjacency lists, [18] Fay Chang, Jeffrey Dean, Sanjay Ghemawat, Wilson C. Hsieh, Deborah
nested sets or joins operations on the database. Graph Store A. Wallach, Mike Burrows, Tushar Chandra, Andrew Fikes, Robert E.
should be used if we want to perform just de-normalization. Gruber, "Bigtable: A Distributed Storage System for Structured Data",
ACM Transactions on Computer Systems,Volume 26, issue 2,Article 4,
Redis is not optimized for maximum security [43] but for June 2008
maximum performance and simplicity. Stonebraker [41]
[19] Seth Gilbert, Nancy Lynch, “Brewer's conjecture and the feasibility of
considered various performance arguments in support of consistent, available, partition-tolerant web services”, ACM SIGACT
NOSQL databases and observed them insufficient. Thus, these NewsVolume 33 Issue 2, June 2002 ,Pages 51-59
systems have various limitations also. http://dl.acm.org/citation.cfm?id=564601
[20] "Google Bigtable, Compression, Zippy and BMDiff". 2008-10-12.
V. REFERENCES Archived from the original on 1 May 2013. Retrieved 14 April 2015..
[1] AvriliaFloratou, Nikhil Teletia , David J. DeWitt, Jignesh M. Patel, [21] Santosh S. Ravi, Kalyanaraman santhanam , “Performance of Neo4j
Donghui Zhang, “Can the Elephants Handle the NoSQL Onslaught?”, versus MongoDB for social actions”; 2014
Proceedings of the VLDB Endowment, 2012 [22] Chad Vicknair, Michael Macias, Zhendong Zhao, Xiaofei Nan, Yixin
[2] Tilmann Rabl, Mohammad Sadoghi, Hans-Arno Jacobsen, Sergio Chen, Dawn Wilkins, “A Comparison of a Graph Database and a
G´omez Villamor, Victor Munt´es Mulero and Serge Mankovskii, Relational”, ACM SE '10 Proceedings of the 48th Annual Southeast
“Solving Big Data Challenges for Enterprise Application Performance Regional Conference; 2010
Management”, The 38th International Conference on Very Large Data [23] DatabaseSQL Server by Microsoft: https://www.microsoft.com/en-
Bases, August 27th - 31st 2012, Istanbul, Turkey. Proceedings of the us/sql-server/sql-server-2016
VLDB Endowment,2012 [24] Oracle Official website: https://www.oracle.com/database/index.html
[3] Bogdan George Tudorica, Cristian Bucur, “A comparison between [25] MySQL Documentation: https://dev.mysql.com/doc/refman/5.7/en/what-
several NoSQL databases with comments and notes”, RoEduNet is-mysql.html
International Conference 10th Edition: Networking in Education and
Research, 2011 [26] Definition of RDBMS:
http://searchsqlserver.techtarget.com/definition/relational-database-
[4] Yishan Li and Sathiamoorthy, “A performance comparison of SQL and management-system
NoSQL databases.”, IEEE Pacific Rim Conference on Communications,
Computers and Signal Processing, 2013 [27] Neo4j Official Website: https://neo4j.com/developer/graph-database/
[5] Laurie Butgereit, “Four NoSQLs in Four Fun Fortnights: Exploring [28] TITAN’s (a distributed graph database) Official website:
NoSQLs in a Corporate in a Corporate IT Environment”, SAICSIT http://titan.thinkaurelius.com/
'16 Proceedings of the Annual Conference of the South African Institute [29] Redis Documentation: https://redis.io/documentation
of Computer Scientists and Information Technologists, 2016 [30] Amazon DynamoDB Official Page:
[6] Avinash Lakshman, Prashant Malik, “Cassandra - A Decentralized http://docs.aws.amazon.com/amazondynamodb/latest/developerguide/Int
Structured Storage System”, Operating Systems Review, 2010 roduction.html
[7] John Klein, Ian Gorton, Neil Ernst, Patrick Donohoe, Kim Pham, [31] Riak Official Website: http://basho.com/products/
Chriisjan Master, “A comparison between several NoSQL databases [32] Apache HBASE official website: https://hbase.apache.org/
with comments and notes”, PABS '15 Proceedings of the 1st Workshop
on Performance Analysis of Big Data Systems, 2015
[33] Cassandra Documentation by Datastax Academy:
[8] Jiri Schindler, “I/O performance of NoSQL Databases(VLDB)”, https://academy.datastax.com/resources/brief-introduction-apache-
SIGMETRICS '13 Proceedings of the ACM; cassandra
SIGMETRICS/international conference on Measurement and modelling
of computer systems, 2013 [34] MongoDB Official Website: https://www.mongodb.com/what-is-
mongodb
[9] Amal W. Yassien, Amr F. Desouky, “RDBMS, NoSQL, Hadoop: A
Performance-Based Empirical Analysis”, AMECSE ‘16 Proceedings of [35] Apache CouchDB Official Website: http://couchdb.apache.org/
the 2nd Africa and Middle East Conference on Software Engineering, [36] Brewer’s CAP Theorem:
2016 http://www.royans.net/wp/2010/02/14/brewers-cap-theorem-on-
[10] Hua Fan, Aditya Ramaraju, Marlon McKenzie, Wojciech Golab, distributed-systems/
Bernard Wong, “Understanding the Causes of Consistency Anomalies in [37] BSON Official Website and Documentation: http://bsonspec.org/
Apache Cassandra”, Proceedings of the VLDB Endowment, 2015 [38] JSON Official Website and Documentation: http://www.json.org/
[11] Sumitkumar Kanoje, Varsha Powar, Debajyoti Mukhopadhyay,”Using [39] NoSQL Explanation by Datastax
MongoDB for Social Networking Website: Deciphering the Pros and Academy:http://www.planetcassandra.org/what-is-nosql/
Cons”, IEEE Sponsored 2nd International Conference on Innovations in [40] Antonios Makrisa, Konstantinos Tserpesa, Vassiliki Andronikoub,
Information Embedded and Communication Systems, 2015 Dimosthenis Anagnostopoulosa, “A classification of NoSQL data stores
[12] Zachary Parker, Scott Poe, Susan V. Vrbsky, “Comparing based on key design characteristics”, Cloud Futures: From Distributed to
NoSQLMongoDB to an SQL DB”, ACMSE '13 Proceedings of the 51st Complete Computing, CF2016, 18-20 October 2016.
ACM Southeast Conference, 2013 [41] Michael Stonebraker. 2010. SQL databases v. NoSQL
[13] Gansen Zhao, Weichai Huang, Shunlin Liang, Yong Tang,”Modelling databases. Commun. ACM 53, 4 (April 2010), 10-11. DOI:
MongoDB with Relational Model”, Fourth International Conference on https://doi.org/10.1145/1721654.1721659.
Emerging Intelligent Data and Web Technologies, 2013 [42] “MongoDB Security Guide Release3.2.1”, MongoDB,Inc. February09,
[14] Rick Cattell, “Scalable SQL and NoSQL Data Stores”, SIGMOD 2016,©MongoDB,Inc.2008–2015.
Record, December 2010 (Vol. 39, No. 4), 2010

298 2017 International Conference on Computing and Communication Technologies for Smart Nation (IC3TSN)
Authorized licensed use limited to: UNIVERSIDAD VERACRUZANA. Downloaded on October 15,2024 at 13:45:10 UTC from IEEE Xplore. Restrictions apply.
[43] IBM systems power solution for redis: https://www- Hadoop:
Performance-
A Yassien, Amr F.
Desouky
‘16 Proceedings
of the 2nd
insight into choosingthe suitable
data model by conducting a
304.ibm.com/partnerworld/wps/servlet/download/DownloadServlet?id= Based Empirical Africa and benchmark usingYahoo! Cloud
Analysis Middle East Serving Benchmark (YCSB) on
6sXLmWUMhJJiPCA$cnt&attachmentName=ibm_power_systems_sol Conference on three types of database systems
ution_for_redis.pdf&token=MTQ1NTc4NTMxMTgzMA==&locale=en Software i.e., MySQL for RDBMS,
Engineering MongoDB for NoSQL data
_ALL_ZZ model, and HBase for Hadoop
framework.
[44] Security: http://redis.io/topics/security/ 10 Understanding Hua Fan, Aditya 2015;Proceeding This paper studies the staleness
the Causes of Ramaraju, s of the VLDB of values returned by read
[45] Neo4j Security, https://neo4j.com/docs/operations- Consistency Marlon Endowment operations applied to Cassandra
manual/current/security/checklist/ Anomalies in McKenzie, which supports eventual
Apache WojciechGolab, consistency using quorum-based
Cassandra Bernard Wong replication.
VI. APPENDIX 11 Using MongoDB
for Social
Sumit kumar
Kanoje, Varsha
2015; IEEE
Sponsored 2nd
This purpose focuses on the
drawbacks and advantages
Networking Powar, International offered by MongoDB, so that
Table 5: Summary of Related Work Website: Debajyoti Conference on the developer can make a wise
S.No Paper Title Authors Publication Description Deciphering the Mukhopadhyay Innovations in decision while choosing a
Details Pros and Cons Information database for a social networking
1 Can the Avrilia Floratou, 2012;Proceeding Comparative study between Embedded and website.
Elephants Handle Nikhil Teletia , s of the VLDB SQL and NoSQL is done on Communication
the NoSQL David J. DeWitt, Endowment interactive data serving and Systems
Onslaught? Jignesh M. decision support analysis. With 12 Comparing Zachary Parker, 2013; ACMSE A comparison has been made
Patel, Donghui the help of this evaluation NoSQLMongoDB Scott Poe, Susan '13 Proceedings between one of the NoSQL
Zhang conclusions are drawn and to an SQL DB V. Vrbsky of the 51st ACM solutions, MongoDB, to the
future trends are speculated. Southeast standard SQL relational
2 Solving Big Data Tilmann Rabl, 2012; The 38th Various benchmark programs Conference database, SQL Server. The
Challenges for Mohammad International are run on DBMSs i.e. performance, in terms of
Enterprise Sadoghi, Hans- Conference on Cassandra, HBase, Voldemort, runtime, of these two databases
Application Arno Jacobsen, Very Large Data MySQL and Redis and the for a modest-sized structured
Performance Sergio G´omez Bases, August throughput and the latencies of database has been compared
Management Villamor, Victor 27th - 31st 2012, operations like read write etc 13 Modelling Gansen Zhao, 2013; Fourth This paper explores the
Munt´es Mulero Istanbul, and create graphs for these MongoDB with Weichai Huang, International problems by modelling of
and Serge Turkey. comparisons Relational Model Shunlin Liang, Conference on MongoDB, with relational
Mankovskii Proceedings of Yong Tang Emerging algebra. The dissimilarities of
the VLDB Intelligent Data semantic expression powers
Endowment and Web have been highlighted between
3 A comparison Bogdan George 2011; RoEduNet The paper explains the Technologies RDBMS and MongoDB.
between several Tudorica, International relevance of SQL and NoSQL 14 Scalable SQL Rick Cattell 2010; SIGMOD This paper compares many SQL
NoSQL databases Cristian Bucur Conference 10th databases in different and NoSQL Data Record, data stores against the various
with comments Edition: environments. NoSQL database Stores December 2010 NoSQL data stores which are
and notes Networking in models are those databases in (Vol. 39, No. 4) meant to scale simple DLTO-
Education and which fixed schemas are not style application loads. The new
Research. required, scaling in done data models sacrifice some
horizontally, joins are avoided dimensions like consistency,
and SQL interface isn’t exposed. availability, high scalability etc.
4 A performance Yishan Li and 2013; IEEE This paper compares the 15 Schema Lanjun Wang , 2015; A schema management
comparison of Sathiamoorthy Pacific Rim performance on the basis of key- Management of Oktie Proceedings of framework has been presented
SQL and NoSQL Conference on value store implementations on Document Stores Hassanzadeh, the VLDB in this paper for document
databases. Communications NoSQL and SQL databases. Shou Zhang, Endowment stores. The simplicity offered by
, Computers and Various (CRUD) operations are Juwei Shi, Limei JSON document stores can
Signal performed on various NoSQL Jiao, Jia Zou, cause snags in certain database
Processing and SQL databases and drastic Chen Wang management tasks.
variations are recorded as a 16 Brewer's Seth Gilbert, June 2002, CAP theorem has been
result, even among the NoSQL conjecture and Nancy Lynch ACM SIGACT explained in this paper.
databases the feasibility of News Volume
5 Four NoSQLs in Laurie Butgereit 2016; SAICSIT NoSQL databases are been consistent, 33 Issue 2
Four Fun '16 Proceedings divided into groups: key-value available,
Fortnights: of the Annual stores, columnar databases, partition-tolerant
Exploring Conference of graph databases, and document web services
NoSQLs in a the South databases. One database has 17 MongoDB vs Alexandru 2012, Emerging This paper describes the
Corporate in a African Institute been picked from each category: Oracle - database Boicea, Florin Intelligent Data distinctions we come across
Corporate IT of Computer Redis, Cassandra, Nep4j and comparison Radulescu, and Web while comparing an SQL
Environment Scientists and MongoDB Laura Ioana Technologies database i.e. Oracle database
Information Agapin (EIDWT), Third and NoSQL form of database
Technologists. International i.e. document store database.
6 Cassandra - A Avinash 2010; Operating This paper explains how Conference
Decentralized Lakshman, Systems Review Cassandra works. Cassandra is a 18 Comparative Supriya S. Pore, 5, May 2015, This research paper aims to
Structured Prashant Malik distributed storage system which Study of SQL & Swalaya B. International evaluate and compare these two
Storage System is used to manage a large NoSQL Pawar Journal of SQL & NOSQL databases and
amount of structured data Databases Advanced examines which of these is
7 A comparison John Klein, Ian 2015 ;PABS '15 This paper describes methods Research in better when it comes to
between several Gorton, Neil Proceedings of and the corresponding results of Computer performance and scalability.
NoSQL databases Ernst, Patrick the 1st a study that is conducted for the Engineering & Index Terms— RDB
with comments Donohoe, Kim Workshop on three selected NoSQL databases Technology
and notes Pham, Chriisjan Performance when we consider the scenario 19 NoSQL Veronika 2013, C3S2E '13 This research paper revolves
Master Analysis of Big of the data of a large healthcare Databases: Abramova, Proceedings of around NoSQL databases, their
Data Systems. organization. MongoDB vs Jorge the International features and operational
Cassandra Bernardino C* Conference principles. It compares and
8 I/O performance Jiri Schindler 2013; The paper aims to evaluate the on Computer evaluates two NoSQL
of NoSQL SIGMETRICS architecture of a few selected Science and databases i.e. MongoDB and
Databases '13 Proceedings NoSQL databases to lay the Software Cassandra.
of the ACM; foundations for understanding Engineering
SIGMETRICS/i the functioning of the new 20 Have Your Data Dipti Borkar, 2016, This paper explains the
nternational emerging database systems. and Query It Ravi Mayuram, SIGMOD’16 architectural alterations that are
conference on They base their work on an Too: Gerald Sangudi, Proceedings of vital to be made to tackle the
Measurement electronic healthcare record From Key-Value Michael Carey the 2016 requirements of future
and modelling of (HER) System where they were Caching to Big International generation applications which
computer given use cases by a customer Data Conference on employ databases.
systems. that he primarily wanted them to Management Management of
work on. Data
9 RDBMS, NoSQL, Amal W. 2016; AMECSE This paper aims to provide an

2017 International Conference on Computing and Communication Technologies for Smart Nation (IC3TSN) 299
Authorized licensed use limited to: UNIVERSIDAD VERACRUZANA. Downloaded on October 15,2024 at 13:45:10 UTC from IEEE Xplore. Restrictions apply.