This document discusses NoSQL databases and compares MongoDB and Cassandra. It begins with an introduction to NoSQL databases and why they were created. It then describes the key features and data models of NoSQL databases including key-value, column-oriented, document, and graph databases. Specific details are provided about MongoDB and Cassandra, including their data structure, query operations, examples of usage, and enhancements. The document provides an in-depth overview of NoSQL databases and a side-by-side comparison of MongoDB and Cassandra.

1. XXX-X-XXXX-XXXX-X/XX/$XX.00©20XX IEEE
In-depth understanding of NoSQL and comparison of two different
NoSQL Databases
Abstract— Growing firms confront a variety of obstacles and
possibilities that needlong-term planning. What worked a year
ago may no longer be the greatest strategy today. Similarly,
while developing a new enterprise.NET application, you must
consider carefully and select the best technology for the job. If
you make the wrong decision, it will have a long-term impact on
your entire firm. Choosing the correct database technology is
one of the factors that will influence how your application is
designed (1). With the growth of the Internet and cloud
computing, databases must be able to store and process large
amounts of data efficiently, as well as provide high performance
while reading and writing, therefore the classic relational
database is facing several newissues. Using a relationaldatabase
to store and query dynamic user data has proven insufficient,
particularly in large-scale and high-concurrency applications
such as search engines and social networking sites. In this
instance, a NoSQL database was constructed. This paper
discusses the history, fundamental features, and data model of
NoSQLand finally some renownedNoSQLdatabases are being
discussed to help corporate firms decide which database to use
(2).
Keywords—NoSQL; Key-value; Column Oriented; document;
database; Big Data;
1. Introduction
The fundamental question of why do we need NoSQL
database when we can use SQL database can be answered
using the analogy of different types of screwdrivers. Almost
all screwdrivers do the same job i.e., tightening and losing the
screws. So why don’t we use the best one eliminated the rest?
Standard/slotted screws are less expensive to produce.
They're difficult to use with electric screwdrivers, though,
since the driver keeps falling out.Phillips-head screwdrivers
are preferable for this, although they can peel out if the metal
isn't strong enough. And so forth. Each type of screw has a
purpose and is beneficial for a variety of applications. It's the
same with a lot of other forms of technology. There are
tradeoffs between complexity and ease,cost to create vs.cost
to use, and so on (3).
Likewise different types of databases are used for different
purposes:
 A relational database is intended to meet the needs
of the data. You create a schema that ensures well-
formed data. When you know that the data is
properly structured, it is easy to create queries (4).
 NoSQL databases are more flexible, allowing you to
begin storing data in a variety of formats. These are
more versatile in terms of data format, but there is
no assurance of data consistency therefore writing
queries is more difficult. Furthermore, the database
architecture is centered on your query needs rather
than the data requirements. So, you must first assess
the queries that must be executed (5).
With the evolution of database technology over the ages
many different database systems have emerged. With the
advancements in internet technologies and hardware
components,the databases are now high in demand than ever
before for the following reasons:
1) Connectivity and centralization:
Most DBMS only support one formofdatabase,such as SQL,
NoSQL, or Key-value. This means that you'd require a
graphical user interface (GUI) for each of these databases in
order to handle them as an administrator. A good DBMS
should help consolidate all of these databases and operate as
a data warehouse.
2) Processing in memory:
An in-memory database stores all data in a computer's main
memory, or RAM. A typical database pulls information from
hard discs. When it comes to data processing, in-memory
databases have a significant performance advantage.They are
significantly quicker than traditional databases because they
utilize fewer CPU instructions and eliminate the time
required to read data from a drive.
3) Performance and distributed computing:
When it comes to RDBMS, huge parallel computing has
numerous major advantages. A. It can help to calculate a
greater quantity of data simultaneously, making it
considerably quicker. B. Users are not required to access the
database in the same network environment.
4) Secure and IAM-enabled
Any DBMS must have Identity Access Management.Despite
the fact that most systems are not inherently collaborative.
This implies a significant knowledge gap between the
administrator and the end users. The finest solutions should
facilitate cooperation while maintaining a high level of
security.
5) Excellent UI/UX
The majority of DBMS pay no attention to UI/UX. The best
should appeal to customers from all backgrounds,since data
consumers nowadays might range from company managers
to data analysts. A great user interface should assist achieve
significantly higher efficiency than an IT-dependent solution
(6).
Although relational databases have a prominent position in
the data storage sector, when confronted with the following
criteria, they have certain inherent limitations:
6) Poor Interface:
SQL databases have very complex interface which makes it
difficult for the user to interact.

2. 7) Cost Inefficient:
SQL databases cost huge amount which make them very
expensive.
8) Partial Control:
Due to some best kept secret business policies SQL databases
grants limited control to its users.
9) Security:
Due to huge amount of sensitive data being stored on SQL
databases it is at huge risk concerning security (7).
A variety of different types of databases have emerged to
address the aforementioned demands. Because these novel
databases differ significantly from typical relational
databases, they are referred to as "NoSQL" databases (8) .
II. CHARACTERISTICS, DATA MODEL, AND
CATEGORIZATION
A. Features of NoSQL Database
1) The main advantages of NoSQL database are
following: 1) Support for Multiple Data Models; 2) Easily
Scalable via Peer-to-Peer Architecture; 3) Flexibility :
Versatile Data Handling; 4) Distribution Capabilities; 5)
Zero Downtime (9).
One of the key characteristics of NoSQL database and
probably the reason of its development is how to handle
diverse data structures before writing it Thus, NoSQL refers
to a wide range of databases that can readily store and
manage enormous amounts of semi-structured and
unstructured data. They can manage a large volume of reads
and writes while also expanding data horizontally. Even if
there is no logical grouping, NoSQL allows us to arrange any
type of stored data. It provides several benefits through the
use of various types of databases, such as hybrid cache
stores, graph databases, and so on (10).
I've (for the sake of research) recently experimented with
NoSQL databases such as MongoDB, Cassandra, and
CouchDB. NoSQL is not designed for server-side querying,
as far as I am aware. There is no join procedure, thus
embedding records into one collection is ineffective. It is not
suited for use in commercial transactions. Aggregation
framework for these NoSQL databases currently lacks
performance adjustment. And so forth. It is suitable for
applications that do not require numerous transactions but
require a large amount of data storage. Also, there is no
common, standardized query language which is the industry
standard as there is in SQL for relational databases.
NoSQL databases have been existed since the 1960s, but it
wasn't until the early twenty-first century that businesses
began to embrace them, particularly to manage huge data and
real-time online and cloud applications. Since then, the
NoSQL database has grown in popularity and utility, albeit
relational databases remain useful (11).
B. Data Model
Data models describe how a database's logical structure is
represented. Data Models are basic elements in a DBMS for
introducing abstraction. Data models determine how data is
linked to one anotherand how it is handled and stored inside
the system. The very first data model might be flat data-
models, in which all data is held on the same plane. Earlier
data models were not as scientific; therefore, they were
prone to duplication and update abnormalities (12). The
following are the NoSQL database fields and the standard
data model:
a) Key-Valued:
The key-value database may be viewed as an associative
array, which is a generalization of a standard array, i.e. A
["anything"]= "something," where "anything" is the key and
"something" is the value. This is a strong notion; forexample,
an in-memory key-value database such as Redis may act as a
distributed hash table,expanding the RAM concept.You may
use this simple yet effective program to pool (cluster) the
RAM of numerous commodity PCs to build a massive main-
memory to offer yourapplication a significant speed increase.
Of course, there are additional factors to consider, but let's
avoid the gory specifics (13).
b) Column Family:
Column Family databases may be viewed as C structure
objects with an arbitrary number of fields. This class includes
Cassandra and HBase. These can alternatively be viewed as
"relaxed" variations of typical RDBMS databases with the
primary purpose of scalability. This work is suitable for
processing column-oriented data.
c) Document:
The structure of a document database and a key-value
database is quite similar; however, the value of a document
database is semantic and is saved in JSON or XML format.
Furthermore, document databases may typically have a
Secondary Index to value to help with the higher
application, although key-value databases cannot (14).
d) Graph:
Graph databases model the most complicated data by
making data relationships first-class residents of the
system. These databases can perform fast graph focused
queries such as path traversals by optimally modelling the
data as graphs,multi-graphs, or hyper-graphs.So, they are
fantastic for friend-of-a-friend questions. Some of them,
interestingly, employ key-value databases as an underlying
technology to improve speed and accomplish scalability.
Some good examples include Neo4j and Hypergraph db.

3. C. CAP Theorem and NoSQL
According to the CAP theorem, a distributed systemcannot
be consistent, available, and partition tolerant at the same
time. When we needed to store more data or expand our
processing power in the past, we had two options: grow
vertically (buy more powerful computers) or further
optimize the existing code base. However, with
advancements in parallel processing and distributed
systems,it is more typical to extend horizontally, orto have
several computers performing the same work in parallel. In
the Apache project, we can already witness a slew of data
manipulation tools like as Spark, Hadoop, Kafka,
Zookeeper, and Storm. However, in order to properly select
the tool ofchoice, a fundamental understanding ofthe CAP
Theorem is required. The CAP Theorem states that a
distributed database systemcan only have two of the three
characteristics: consistency, availability, and partition
tolerance (15).
The classification of NoSQL according to CAP theorem is
as follows (16):
a) Consistency and availability (CA):
Part of the database is unconcerned with partition tolerance
and instead relies on the replication strategy to assure data
consistency and availability. CA-related systems include the
standard relational database, Vertica (column-oriented),
AsterData (Relational), Greenplum (Relational), and others.
b) Consistency and partition tolerance (CP):
Such a database system stores data on remote nodes while
also ensuring data consistency, although support is
insufficient for availability. BigTable (Column-oriented),
Hypertable (Column-oriented), HBase (Column-oriented),
MongoDB (Document), Terrastore (Document), Redis
(Key-value), Scalaris (Key-value), MemcacheDB (Key-
value), Berkeley DB are the primary CP systems (Key-
value).
c) Availability and partition tolerance (AP):
Such systems provide availability and partition tolerance
largely through consistency, as demonstrated by AP's
system: Voldemort (Key-value), Tokyo Cabinet (Key-
value), KAI (Key-value), CouchDB (Document-oriented),
SimpleDB (Document-oriented), Riak (Document-oriented),
Riak (Document-oriented), Riak ( (Document-oriented).
III. ANALYSIS OF POPULAR NOSQL DATABASES
A. MongodB
MongoDB is an open source NOSQL database written in
C++ language.
a) Structure and Data model
MongoDB (17) is a document database, not a relational
database management system. That is, records are saved as
documents (usually JSON) rather than tuples. As a result,
data modelling for MongoDB necessitates a more object-
oriented approach than data modelling for an RDBMS. Both
composition and association relations, for example, can be
described using mapping tables in an RDBMS; it is up to the
application to enforce the exact nature of the connection -
composition or association. However, in the case of a
document database such as MongoDB, a composition
relation must be enforced at the document level using
nested/contained objects.
b) Query operators
Query operations in mongo dB can be used for the following
purposes (18):
 Comparison:
For e.g., “$eq” matches values that are equal to a specified
value; “$gt” matches values that are greater than a specified
value; “$lt” matches values that are less than a specified
value, etc.
 Logical operations:
For e.g., “$and” joins query clauses with a logical AND
returns all documents that match the conditions of both
clauses; “$not” inverts the effect of a query expression and
returns documents that do not match the query expression;
“$nor” joins query clauses with a logical NOR returns all
documents that fail to match both clauses; “$or” joins query
clauses with a logical OR returns all documents that match
the conditions of either clause.
 Element:
For e.g., “$exists” Matches documents that have the
specified field; “$type” selects documents if a field is of the
specified type.
 Evaluation:
For e.g., “$mod” performs a modulo operation on the value
of a field and selects documents with a specified result.

4.  Geospatial:
For e.g., “geoIntersects” selects geometries that intersect with
a GeoJSON geometry. The 2dsphere index supports
$geoIntersects.
 Array
For e.g., “$all” matches arrays that contain all elements
specified in the query.
 Bitwise
For e.g., “$bitsAllClear” matches numeric or binary values in
which a set of bit positions all have a value of 0.
c) Examples of query operations
In MongoDB,you must first create a collection before you can
create a table. The collection works much the same way that
we would build a table when putting data into a relational
database.
To insert the document in the MongoDB collection, you have
to use the insert() method.
d) Common Usage and market
Some of the most common mongo dB usage are as follows
(19):
 Mobility and Scaling:
MongoDB is highly scalable and adaptable, making it ideal
for dealing with a wide range of settings.
 Real time data integration:
Data has a lot of value when it is condensed and aggregated
into a single perspective, and MongoDB plays an important
part in that.
 Product Catalog:
There are many attributes to products which are easily stored
as an object using MongoDB and can be used to understand
the customer better in the digital experience.
Some of the world’s renowned companies that use mongo dB
includes: eBay, MetLife, Shutterfly, Aadhar, EA among
many others.
e) Enhancements in the platform
MongoDB is a highly helpful and easy solution, however for
more complicated projects, there may be certain specialized
needs that necessitate the usage of a different technology. It
increases the complexity of the architecture, yet it may be
helpful to the world, the features, and the simplicity with
which the features are implemented. Also, it could be more
stable. It would be preferable if it were more user-friendly,
like Oracle is. In Oracle, for example, building an index is
straightforward. It's difficult to achieve that with MongoDB.
Performance might be improved. It's quick and reliable, but
you can't put every application you want on MongoDB.
B. Cassandra
Cassandra (20) is an open source database of Facebook.
a) Structure and Data model
Its qualities are as follows: 1) the schema is extremely
flexible and does not require the creation of a database
schema at the outset,and adding or deleting fields is quite
simple. 2) Support range queries, i.e., range queries for
Key; 3) High scalability: a single point of failure does not
affect the entire cluster, and linear extension is supported.
Cassandra is a distributed database system made up of
several database nodes; a write operation is replicated to
other nodes,and a read request is routed to a specific node.
Scalability may be achieved by simply adding nodes to a
Cassandra cluster. Cassandra also has a complex data
structure and a sophisticated query language.
b) Query Operation and example
The Cassandra Update query is used to update the
Cassandra table's data. If no results are given after
changing data, this indicates that the data was properly

5. changed; otherwise, an error will be issued. The 'Set' clause
modifies column values while the 'Where' clause filters data.
Syntax:
Update KeyspaceName.TableName
Set ColumnName1=new Column1Value,
ColumnName2=new Column2Value,
ColumnName3=new Column3Value,
.
.
.
Where ColumnName=ColumnValue
Here is a snapshot of the database before data was updated.
After running the following query:
Update University.Student
Set name='Hayden'
Where rollno=1;
Here is the screenshot that shows the database state after
updating data.
c) Common usage and market
Cassandra was used to power the new Actionable Analytics
product at Ooyala (21), a company that provides an end-to-
end online video platform to corporate companies. All video
engagement and analytics reports delivered to clients are
aggregated acrossmany dimensions and sliceable by the user's
geo (country/region/city), the site/URL where a video is
embedded, various advertising parameters, and a variety of
additional tags.
Cassandra offered the scalability and better write throughput
required to support the new analytics offering without the
requirement to develop their sharding and replication on top
of MySQL. Ooyala manage a big volume of analytics data (>
1B data points per month) with millions of hours of video
supplied each month.
d) Enhancement in the product
Cassandra's secondary index is somewhat troublesome and
might be improved. Cassandra can improve by having a fuller
ecosystem integrator. Companies, for example, required to
deploy extra tools in some maintenance operations to execute
duties that were not provided alongside Cassandra. Another
limitation of the product is that you cannot drop writes like
you do in MongoDB and MySQL, where you may link tables.
Cassandra does not provide joins acrosstables;thus,you must
use another tool for this. You must read all of the data and
store it in memory before adding the joins. That is an area
where they can improve.
IV. CONCLUSION
Both SQL and NoSQL database formats store data, but how
they do so vary. Outline the application to choose the best
database architecture for data structures such schema,
relation, scalability, and data size. Migrating from one to the
other is expensive and time-consuming, thus the differences
should be addressed while designing software. This paper
begins by comparing traditional databases with that of
NoSQL databases. Then it explains the underlying working
of NoSQL database and then finally two different NoSQL
databases are discussed.Each type was database has its own
pros and cons.NoSQL is a recent technology and many of its
dimensions are yet in a phase of development. This paper
provides a comprehensive overview to the businesses of the
kind of database they should use based on their
functionalities.
REFERENCES
[1] Why NoSQL? - NosDB. Alachisoft. (2022). Retrieved 25 May 2022,
from https://www.alachisoft.com/nosdb/why-nosql.html.
[2] Jing Han, HaihongE, Guan Le, & Jian Du. (2011). Survey onNoSQL
database. 2011 6th International Conference on Pervasive Computing
and Applications. https://doi.org/10.1109/icpca.2011.6106531
[3] Why do we neednosql when we can use SQL? Quora. (n.d.). Retrieved
May 25, 2022, from https://www.quora.com/Why-do-we-need-
NoSQL-when-we-can-use-SQL
[4] A quick-start tutorial on relational database design. Relational
Database Design. (n.d.). Retrieved May 25, 2022, from
https://www3.ntu.edu.sg/home/ehchua/programming/sql/Relational_D
atabase_Design.html
[5] Harpreet Kaur, Jaspreet Kaur, Kamaljit Kaur,2013, A Reviewof Non-
Relational Databases, Their Types, Advantages And Disadvantages,
INTERNATIONAL JOURNALOF ENGINEERINGRESEARCH &
TECHNOLOGY (IJERT) Volume 02, Issue 02 (February 2013)
[6] What is the future of database management systems? Quora. (n.d.).
Retrieved May 25, 2022, from https://www.quora.com/What-is-the-
future-of-database-management-systems
[7] Competitions, quizzes, hackathons, scholarships and internships for
students andcorporates. Unstop.(n.d.). RetrievedMay 25, 2022,from
https://unstop.com/blog/advantages-and-disadvantages-of-sql
[8] The 5 features to look for in a nosql database. DataStax. (n.d.).
Retrieved May 25, 2022, from https://www.datastax.com/blog/5-
features-look-nosql-database
[9] What are the characteristics of a nosql database? Quora. (n.d.).
Retrieved May 25, 2022, from https://www.quora.com/What-are-the-
characteristics-of-a-NoSQL-database
[10] The 5 features to look for in a nosql database. DataStax. (n.d.).
Retrieved May 25, 2022, from https://www.datastax.com/blog/5-
features-look-nosql-database
[11] DBMS - Data models. Tutorials Point. (n.d.).RetrievedMay25,2022,
from
https://www.tutorialspoint.com/dbms/dbms_data_models.htm#:~:text
=Data%20Models%20are%20fundamental%20entities,kept%20in%2
0the%20same%20plane.
[12]
[13] Is there an example ofa complex NoSQL - NDB Data Model? Quora.
(n.d.). RetrievedMay25,2022, fromhttps://www.quora.com/Is-there-
an-example-of-a-complex-NoSQL-ndb-data-model
[14] An illustratedproofofthecaptheorem. (n.d.). RetrievedMay25, 2022,
from

6. https://mwhittaker.github.io/blog/an_illustrated_proof_of_the_cap_th
eorem/
[15]
[16] Scribd. (n.d.). Blognahurst com visual guide to Nosql Systems. Scribd.
Retrieved May 25, 2022, from
https://www.scribd.com/document/164913941/Blog-Nahurst-Com-
Visual-Guide-to-Nosql-Systems
[17] The ApplicationData Platform. MongoDB. (n.d.). RetrievedMay 25,
2022, from https://www.mongodb.com/
[18] Query andProjectionOperators - MongoDB Manual.(n.d.). Retrieved
May 25, 2022, from
https://www.mongodb.com/docs/manual/reference/operator/query/
[19] MongoDB Real World Use Cases: Advantages & Top Companies
[2022]. upGradblog. (2022, January5).RetrievedMay25,2022, from
https://www.upgrad.com/blog/mongodb-real-world-use-cases/
[20] Open source nosql database. Apache Cassandra. (n.d.). RetrievedMay
25, 2022, from http://cassandra.apache.org/
[21] Agile Media Solutions. Dalet. (2022, May9). RetrievedMay 25, 2022,
from http://www.ooyala.com/blog?eid=133
.