WHITE PAPER

NOSQL TECHNOLOGY FOR

TODAY’S DATA

Abstract
Over the past decade, NoSQL has become a mainstream technology
that is used by many enterprises to re-imagine how they use data.
NoSQL is a broad area encompassing multiple underlying database
technologies, each with its specific use case. This white paper
provides an overview of NoSQL and examines how different NoSQL
databases can help companies meet rapidly changing business
needs to stay competitive.
 The evolution of data How is next-gen data Relational databases are unable to handle

– the 3 Vs breaking relational database unstructured data like audio/video files

and social media posts. Moreover, the
Data has come a long way. If one compares models?
storage of large volumes of data in the
the volume of data that could be stored The 1970s saw the growth of relational RDBMS becomes cost prohibitive and
on 80-byte punch cards used in the databases that, until now, have dominated performing any aggregation or analytics
1950s to that on 1TB microSD cards that the IT landscape. These databases can activities on the data is very difficult.
were rolled out in February 2019, the store data in structured formats with RDBMS are designed to scale up and not
increase in storage density is nearly 12 tight controls for data integrity. They also scale out. This means the only way for
billion times. Even the variety of data that leverage a common language (SQL) to an RDBMS to handle larger workloads is
can be stored has changed from highly
access and manipulate data. This allows to deploy more CPU, RAM and hard disk
structured corporate data to videos, chats
enterprises to build massive systems on a single machine, which in turn limits
and pictures. According to a study by
of record that provide a trusted source scalability and increases cost. However,
Marr in 2018[1], 1.2 trillion photographs
of truth. With ACID compliance and this is not the way modern cloud native
were generated in 2017. In recent years,
distributed transactions, even mission- software works.
the velocity of data too has exploded
critical applications can guarantee data
with mobile devices generating nearly 12 The rigid data model at the center of any
integrity. Once an INSERT or UPDATE
exabytes of data (Kemp, 2018)[2] at speeds RDBMS guarantees integrity. But change is
transaction is committed, you are
of 350 GB/second. These 3 “Vs” – Volume, expensive. As agile development becomes
guaranteed that data will be available for
Variety and Velocity – have come to the new paradigm, developers need
access anytime in the future.
define the evolution of data in the last few database solutions that allow the structure
decades. While the capabilities of relational to evolve easily.
databases are still quite valuable, new
Storing, retrieving and managing this Another important trend that has
staggering volume of data has caused an trends over the last decade are creating
emerged over the last two decades
evolutionary leap in technologies under shifts in how companies manage data.
is object-oriented programming. The
the ‘Big Data’ umbrella. This originated For instance, engagement with the end
highly normalized structure of RDBMS
within next-gen companies such as user has become a key business driver.
has resulted in an ‘impedance mismatch’
Google, Facebook, Twitter, and Yahoo Thus, even while ensuring the integrity of
between the application layer and
that have been at the center of this data the core financial transactions is critical,
the database layer, calling for unusual
explosion. Data analytics has also become there are several use cases where the
constructs like object-relational mapping
a crucial tool with Hadoop/MapReduce need for short response times outweigh
(ORM) frameworks.
based architecture that effectively crunch the need for data integrity. In these
large data sets. Today, real-time systems cases, a traditional relational database
need solutions to handle relatively smaller management system (RDBMS) cannot
data volumes even as they grapple with all meet expectations.
the challenges presented by the 3 Vs.

External Document © 2020 Infosys Limited

What is NoSQL and why it key value store, document, and graph Some examples of these databases are
databases are the most popular ones. Each Cassandra and HBase. Cassandra was
is dominating the database
type has its own specific use case and a used by Facebook to handle their huge
market?
different underlying technology. data volumes that also needed to be
Several technologies have evolved to searchable. Some examples of successful
address the above challenges. As these Columnar or Wide Column implementations of wide-column
technologies were compared with the
Stores databases are:
traditional SQL-based RDBMS databases,
This database type is optimized to store • Spotify uses Apache Cassandra to
they were grouped under the umbrella
large volumes of data in columns that can support 40,000 requests per second for
term of ‘NoSQL’ (which stands for ‘Not
be queried efficiently. It is best-suited for persistent data store[3]
Only SQL’) or ‘non-relational’. There are a
number of different database types that ingesting large data volumes and creating • Ebay uses Cassandra to store 250 TB of
are clubbed under this term. Columnar, visualizations and OLAP-like queries. data with 6 billion writes per day[4]

External Document © 2020 Infosys Limited

 Key Value Stores
Here, data is stored as a ‘key’, which is it refers to frequently such as user session • LinkedIn uses Couchbase to cache over
the primary attribute of the ‘value’. These information, master lookup information, 8 million real-time metrics (over 12TB of
databases are usually implemented currency rates, etc. Some examples of key data). Over 16 million entries are loaded
as memory-only data stores to cache value databases are: into Couchbase every 5 minutes[6]
frequently accessed data. They are
• Microsoft uses Redis to power the MSN
effective in scenarios that require the
portal that gets 2 billion hits with a
application to maintain a list of items that
latency of less than 10ms on peak days[5]

External Document © 2020 Infosys Limited

Graph Databases
These databases store data in graph-like
structures and use nodes and edges to
represent the data and its connections.
These are best-suited for specific use
cases that require graph traversal and
route optimization-like functionalities.
For instance, it can be challenging for
companies to determine the inter-
relationships between users, transactions,
locations, etc. A graph database supports
such use cases for activities like fraud
detection, compliance with legal
restrictions on financial transactions,
etc. Neo4J is a prominent example along
with OrientDB that has multi-modal
capabilities. Some examples of successful
implementations of graph databases are:

• ICIJ used Neo4J to uncover relationships

between people and their bank
accounts, helping them ‘follow the
money’ for investigative journalism
related to offshore tax fraud[7]

• Walmart uses Neo4J to improve the

speed and effectiveness of their online
recommendation platform[8]

Document Databases
This is perhaps the most popular type of
NoSQL database. Document databases
organize information in JSON or XML
documents that can be accessed
in multiple ways including SQL-like
languages. Modern applications represent
business entities as objects that are
best represented in a serialized form in
documents. Here, the resulting impedance
mismatch between the application and
the database is much lower, enabling
faster access and simpler programming.
Document databases can efficiently store
all kinds of entities like user profiles,
transaction details, product catalogues,
asset information, etc., in a highly scalable
and performant manner. Some examples
of successful document database
implementations are:

• Craigslist uses MongoDB to store over 5

billion documents amounting to 10TB

External Document © 2020 Infosys Limited

 of data with horizontal scaling through stores over 20 billion documents on some databases types being subsumed.
commodity hardware[9] Couchbase[10] For instance, databases like MongoDB can
work as key value and graph databases
• Amadeus, the world’s largest travel Since NoSQL is an evolving technology,
even though the primary database type
booking engine, manages more than the differentiation between its databases
will offer richer functionalities in that
8 million queries per second and occasionally blurs. Over time, we may see
specific area.

External Document © 2020 Infosys Limited

NoSQL as an alternative to language) must use NoSQL databases reduced batch processing time to 30
because the corresponding capability minutes
RDBMS
in RDBMS is typically offered through
• Infosys helped a Europe-based credit
Enterprises should consider using NoSQL expensive add-ons
rating agency implement a graph
as an alternative to their existing databases
• Use cases like offline mobile apps, database that uncovered hidden
in the following scenarios:
operational data stores which offload relationships within complex holding
• New applications that are expected to read traffic from mainframes, and patterns, allowing them to determine
store an evolving data structure, i.e., customer 360 solutions lead to the potential ownership patterns
the data structure is expected to evolve automatic selection of a NoSQL database
• For a financial asset management
as new business lines are added or as
company, Infosys built an operational
functionality is expanded. An RDBMS will
Success stories data store (ODS) using MongoDB
usually have a pre-defined structure that
Infosys has helped several financial services to unify data from multiple legacy
is difficult to change
companies implement NoSQL databases to platforms. The new database reduced
• Applications that are expected to be the reporting latency, improved asset
meet their growing business needs. Some
impacted by M&A should consider class utilization and improved risk
examples are:
moving to NoSQL. The flexibility offered monitoring
by NoSQL databases makes changing • A retirement financial services leader
was struggling with an Oracle-based • Infosys deployed a flexible and scalable
data structures relatively easy compared
system that ran for 20 hours to perform MongoDB database to help a global
to RDBMS
a monthly batch. Infosys migrated the financial services provider improve client
• Applications that deal with streaming satisfaction through a user-friendly web
system to a scoring solution based
data (social media feeds, IoT) will have interface for post-trade analysis
on Apache Spark and Cassandra that
no alternative but to employ NoSQL
databases because RDBMS technology
is not built to handle the volume and
speed associated with streaming data

• Applications employing RDBMS

databases that are facing significant
performance issues due to high user
count or large data load can benefit
significantly by deploying the right
NoSQL database at the back-end

• Databases where data volume is

expected to grow exponentially with
a starting base of few GBs to multiple
terabytes within a few months will need
the horizontal scalability offered by
NoSQL databases

• Any new agile development using

modern languages like Java and
JavaScript should consider NoSQL
given the ease of development, low
‘impedance mismatch’ and high
developer productivity

• Applications that need high availability,

replication, failover, use of commodity
hardware, and are polyglot (multi-

External Document © 2020 Infosys Limited

Conclusion
As the nature of data evolves, new technologies are emerging to support data and unlock greater value from the business. The three factors
that will drive adoption of NoSQL are response time, data volume and overall cost goals. Enterprises looking to develop modern applications
should carefully identify the challenges of their existing systems, the needs of the business and how NoSQL databases can bridge this gap.
Infosys possesses deep expertise in NoSQL implementation along with a strong partner ecosystem, allowing us to offer end-to-end solutions
that meet client expectations.

About the author

Vageesh Patwardhan
Vageesh Patwardhan is a Principal at Infosys. He leads the Open Source database practice and is a hands-on NoSQL
Architect/Administrator. Vageesh has 22 years of experience in IT ranging from Mainframes to the latest NoSQL databases
across multiple verticals.

References
[1] Forbes. [Online]. Available: [Link]
blowing-stats-everyone-should-read/#7bb4ede360ba.

[2] Wearesocial. [Online]. Available: [Link]

know.

[3] DataStax Inc., [Online]. Available: [Link]

[4] DataStax, “Ebay Case Study,” [Online]. Available: [Link]

[5] Redis Labs Inc, [Online]. Available: [Link]

[6] Couchbase Inc., “High Performance Applications with Distributed Caching” [Online]. Available: [Link]
performance-wp?x=s9hNYZ

[7] Neo4J Inc, “International Consortium of Investigative Journalists Case Study,” [Online]. Available: [Link]

[8] Neo4J Inc., “Walmart Case Study,” [Online]. Available: [Link]

[9] MongoDB inc., “MongoDB Case Study: Craigslist,” [Online]. Available: [Link]
study-craigslist.

[10] Couchbase Inc., “Amadeus Powering the travel industry,” [Online]. Available: [Link]

For more information, contact askus@[Link]

© 2020 Infosys Limited, Bengaluru, India. All Rights Reserved. Infosys believes the information in this document is accurate as of its publication date; such information is subject to change without notice. Infosys
acknowledges the proprietary rights of other companies to the trademarks, product names and such other intellectual property rights mentioned in this document. Except as expressly permitted, neither this
documentation nor any part of it may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, printing, photocopying, recording or otherwise, without the
prior permission of Infosys Limited and/ or any named intellectual property rights holders under this document.

[Link] | NYSE: INFY Stay Connected