Real time applications NoSQL Databases vs.
Relational Databases – A Performance Review
Vinay Kumar Anamalagundam
X17148146
Data Analytics, School of Computing,
National College of Ireland
Abstract— Relational databases support ACID properties that
guarantee strong consistency of data. In contemporary
warehousing methodologies, relational databases have almost
failed to handle the spreading complexity and vast amount of
data generated from social networks. As an alternative, NoSQL
has led to a new trend. Instead of ACID, NoSQL systems adopt
BASE characteristic’s and with it gives slight consistency
restrictions than relational databases. Data warehouses which are
implemented using relational databases started showing
limitations on real time applications. In this paper, we mainly
review performance of NoSQL databases, Cassandra and
MongoDB on Twitter and E-commerce respectively. And, this
paper discusses, key features of NoSQL databases based on data
warehousing over the Relational databases. NoSQL databases
endorse the feature called 3V i.e. Volume (vast amount), Velocity
(speed), Variety (wide diversity). Various approaches of data
migration from relational databases to NoSQL.
Keywords— ACID, BASE, Relational database, NoSQL,
Cassandra, MongoDB.
I. INTRODUCTION
The main objective discussed in this paper is the performance
comparison between two models ACID vs BASE that supports
Relational and NoSQL databases based on data warehousing
respectively. All database developers know the importance of
acid and base characteristics that gives strength to the
databases. The acronym of ACID as follows: Atomicity,
Consistency, Isolation, Durability.
Atomicity: either transaction succeeds or rolled back. It comes
to the starting position if transaction fails at any point. It
follows all or none principle.
Consistency: it makes the transaction to follow transactional
rules designated by the system. Database must not be
inconsistent state at any time.
Isolation: every transaction is independent on itself. It makes
that to improve performance within database.
Durability: if transaction completed, it cannot be undo. It
could get through power troubles and system failures.
There are many Business organizations, banking sectors,
Military forces that uses RDBMS databases like SQL, MySQL
etc. that support ACID properties because of strong
consistency. (Banothu, 2016)
RDBMS databases are not capable of handling huge amount of
real time data which is unstructured, and it is usually maintain
single server if it fails data cannot be recovered. Data
warehouse needs good logical design. Generally, there are
three approaches for mapping from conceptual to logical
model. They are: (1) ROLAP (2) MOLAP (3) HOLAP but
these approaches are inadequate when dealing with Real time
applications. (Yangui, 2016) Large amount of data which is
generated through applications or services such as social
media, google reviews, weather data. In these sectors, it is
always hard to keep strong consistency together with
availability and scalability. (González-Aparicio, 2018).
Based on CAP theorem, consistency, availability, and partition
tolerance, only two out of three characters can satisfy. This has
begun to a new trend in databases, called NoSQL databases.
NoSQL adopts BASE characteristics for better performance,
scalability and flexibility etc. The acronym of BASE as
follows: Basically available, Soft state, Eventual consistency.
Basically available: it assurance the availability of data. It
responds to the requested data, but it could be failure
sometimes because of inconsistent state.
Soft state: eventual consistency could make the system soft by
changing the state over the time without input.
Eventual consistency: if system receives inputs continuously,
it wont check the consistency of each transaction before going
to another transaction. If system stops receiving input, it
ultimately becomes consistent. (Banothu, 2016)
NoSQL database has very few consistency restrictions over
relational databases. NoSQL is specially designed to carry out
the needs of big data on real time and cloud computing.
NoSQL has increased write performance by allowing new
datatypes to add dynamically to the database. (Storey, 2017)
NoSQL databases are becoming most popular by the number
of organizations acceptance as a solution for growing
complexity and data from the modern technologies. (Yangui,
2016)
II. CLASSICAL DW VERSUS NOSQL BASED DW
“Data Warehousing is a traditional domain of relational
databases, and there are two reasons for that: (1) data
warehouses are mostly used in enterprises with large scale data