UNDERSTANDING BIG DATA AND NOSQL

M.A.MANALANG

RECENT TRENDS AND MAJOR ISSUES

It is essential to understand technologies related to big data
analytics because the social demand for big data analytics and
big data technicians (engineers, data scientists, etc.) is
increasing significantly.
In addition, NoSQL, which has the characteristics of a BASE
focusing on the fast-processing speed for unstructured data and
large-scale data, is increasingly being adopted in enterprises’
operating environment.
Therefore, it is necessary to understand NoSQL (Not Only SQL)
as it is based on a different concept to existing relational
databases.
Why it is important to understand technologies
related to big data analytics?
What is NoSQL stands for?
LEARNING OBJECTIVES
1.To be able to explain the concept of big data
and related technologies.
2.To be able to explain the concept and
characteristics of NoSQL.
WHY DO WE NEED TO UNDERSTAND BIG DATA TECHNOLOGY?

Although interest in data analysis has increased, the existing system architecture and DBMS face limitations in
terms of processing speed and performance when an existing analysis system processes tens of PBs of
unstructured data generated by multimedia, SNS, sensors, loT, etc, due to the advancement of IT technologies.
As a result, solutions are being developed that are suitable for analyzing the large volume of unstructured data
(variety) generated at high speed (velocity). Technologies related to big data, which was still in its infancy only
a few years ago, have developed rapidly and are now being used directly in our real life.
There are many examples of this. For instance, the US movie Minority Report, which was released in 2002,
portrayed futuristic crime prediction by “precogs’ in the year 2054, yet something akin to this technology was
actually realized as a crime prevention system in San Francisco as early as 2009. (However, the concept of
“big data” had only been introduced at that time for the study of genomes, and prediction based on big data
analytics had not yet been thought of.) Other commonly known examples include Google's flu map, the US
presidential election case, ZARA, the DHL case, and distribution demand forecasting. Furthermore big data
technology was unfamiliar to IT workers until relatively recently to the extent that they only needed to know the
definition of big data. However, with the rapid development of big data technology, terms such as crawler,
Hadoop, MapReduce, R, and NoSQL, as well as the 3V characteristics of big data, have become familiar
technical terms, and IT workers now experience big data systems more frequently in the workplace. Therefore,
IT workers need to understand at least the concepts of the technical terms for each phase of big data
analytics, if not the detailed technical principles, in order to quickly adapt to the changed workplace.
BIG DATA OVERVIEW
A. Definition and characteristics of big data
1. Definition of big data
• Big data generally refers to either data that exceed the ability of database management tools used to capture, store,
and analyze data (McKinsey, 2011), or to next-generation technologies and architectures designed to extract value
from large-scale data at low cost and support the rapid collection, discovery, and analysis of data (IDC, 2011).
2. Characteristics of big data (3v)
• The characteristics of big data can be explained by the three elements (3V) of big data, namely, volume, velocity, and
variety. Each element has the following characteristics.
 3. Structured data vs.
unstructured data
BIG DATA OVERVIEW
Description
Structured data vs. unstructured data
Data type Structured
• Data to be stored in a fixed field.
Semi-structured • Data not stored in a fixed field, but which
contain metadata or schema, such as
XML or HTML.
• XML, CVS, XLS, RDF, etc.
 Unstructured • Data not to be stored in a fixed field. •
Document, picture, video, and audio data,
etc.

3.
BIG DATA OVERVIEW
B. Detailed technology by big data life cycle
ITEM DESCRIPTION DETAILED TECHNOLOGY
Collection - A technology that can collect data from - Crawling (web robot), ETL,
all devices and systems CEP (Complex Event
Processing), etc.
Storage/ - A technology that can store and process - Distributed file system,
pro collected large-scale data using a NoSQL, MapReduce
cessing distributed processing system. processing
Analysis - A method of analysis that can assist - Natural language
companies and the public with using big data in processing, machine
business and daily life. learning, data mining
algorithms, etc.
Visualization - A technology that can visualize analyzed - Visualization such as R,
results effectively. graphs, drawing, etc.

TECHNOLOGIES RELATED TO BIG DATA

A. Collection technology

Various technologies can be used for data collection, such as ETL, web crawling, RSS Feeding, Open API, and CEP
(Complex Event Processing). Among them, the web crawling technology automatically collects various
documents and data generated on the web, and is used to collect data such as SNS, blogs, and news. Web
crawling copies the entire web page after collecting the URLs to be collected, or collects data with a specific tag
only after analyzing the HTML code.

Technology Description Solutions

Collection using Collects data using the SQL function of Oracle, MariaDB, MS SQL,
the DBMS the DBMS. Tibero, etc.
Collection using sensors Collects data when a certain condition is met CQL, Kafka
FTP collection Collects data using port that can transfer files.
HTTP collection Collects data by reading HTML tags Scraper

TECHNOLOGIES RELATED TO BIG DATA

B. Big data storage/processing technology
The Distributed File System (DFS), NoSQL, MapReduce, and other technologies are used to store and process
large amounts of data and unstructured data (that is, big data) generated at high speed in an efficient, cost-
effective manner. Recently, the distributed file system, which is based on the cloud and uses virtualization
technology in the
cloud computing environment, was introduced.
Technology Description Solutions

Distributed File System (DFS) A file system that allows access to files GFS (Google File System), HDFS
on multiple host computers which are (Hadoop Distributed File
shared over a computer network. System), etc.
 NoSQL (Not Only SQL) A new type of data storage/retrieval Hbase, Cassandra,
system that uses a less restrictive Mongodb, CouchBase,
consistency model (BASE Redis, Neo4J, etc.
characteristics) than the traditional
relational database.
Distributed parallel processing A technology that processes a large MapReduce
amount of data in a distributed
parallel computing environment.

TECHNOLOGIES RELATED TO BIG DATA

1. Distributed File System (DFS)

The DFS is a file system architecture for storing and processing large-scale and unstructured data in
a distributed environment. It has the following characteristics.
 It is composed of inexpensive servers.

 Scale-out: Its entire available capacity and performance increase almost linearly each time equipment
is added.
 High availability: Even if some servers fail, the usability of the entire system is not affected very
much.  Optimized for throughput: It is suitable for the batch processing of large-scale data.

TECHNOLOGIES RELATED TO BIG DATA

2.MapReduce
MapReduce is a programming model designed for the parallel
distributed processing of big data using inexpensive machines. This
model can process large amounts of data in parallel using a
program composed of a map procedure and a reduce method.
MapReduce allows the analysis of large-scale data by processing
data that has been distributed and stored in multiple machines.
Basically, MapReduce performs batch-based processing and can
handle large-scale data conveniently. The data of the execution
result is copied, distributed, and stored safely in consideration of the
failure of the physical device.
TECHNOLOGIES RELATED TO BIG DATA
 C. Visualization technology

This technique provides insights by effectively transferring numbers,

statistics, and valuable meanings, by classifying data for the user’s easy
understanding, and by analyzing large-scale data. It allows the exact and
effective transfer of information without actually requiring the user to
look at the data.
BIG DATA VISUALIZATION METHOD
BIG DATA VISUALIZATION METHOD
BIG DATA VISUALIZATION METHOD
TECHNOLOGIES RELATED TO BIG DATA
D. Classification of big data analytics
Big data analytics refers to the process of “discovering meaningful patterns from big data.”
Classification Modeling Contents Applied technique
criteria
The primary purpose is to find Association rule,
Descriptive modeling patterns that describe the given clustering, database,
data. segmentation,
Purpose of use visualization, etc.
Predictive modeling A model is created based on the Classification, Regression,
given data, and is used to predict time series analysis,
new input data. Neural network, SVM
Presence of Supervised data When the target is determined Decision Tree, Neural
target variable network, Case-based
reasoning
Unsupervised data When there is no target. The Association rule
correlation or similarity between discovery, market
data is analyzed with the focus on basket K-means
input variables clustering

TECHNOLOGIES RELATED TO BIG DATA

E. Main methods of big data analysis
Concept Description
• A statistical technique used to predict the possibility of an event’s
Logistic regression occurrence (probability of occurrence) using a linear combination of
independent variables.
Decision tree analysis • A method of quantitative analysis that classifies an interested group into
several subgroups or performs Prediction by drawing a decision tree chart.
Neural network analysis • A method of analysis that handles a problem with
parallel/distributed/probabilistic calculation based on the idea that
digital information is a network of nerve cells, rather than a method of
processing digital information based on a deterministic binary
computational model using the human brain itself as the model.
Text mining • A technology that extracts and processes useful information by applying
natural language processing technology and document processing
technology to unstructured/semi-structured data.
• The core technologies of text mining include document summarization,
document classification, document clustering, and feature extraction.

TECHNOLOGIES RELATED TO BIG DATA

E. Main methods of big data analysis
Concept Description
• A social network refers to the network between the components constituting a given society.
SNA (Social
• An analysis methodology that analyzes and visualizes the relationship between objects -
Network
such as people, groups, organizations, computers and data - and the characteristics and
Analysis)
structure of the network.
Opinion mining • A technology that quickly analyzes the information the user wants and intelligently infers
meaningful information from a large number of unstructured reviews such as SNS and
replies.
• Itis used effectively for corporate marketing policies or public opinion analysis by extracting
the hot topics of the social network service and analyzing the flow in real time.
Natural • An artificial intelligence technology that understands, creates, and analyzes human
Language language using computers.
Processing • The process of understanding natural language by analyzing human language mechanically
(NLP) to convert it into a form that can be understood by computers.
• Or various technologies that express such a form in language that humans can understand.
• Natural language is processed in the following order: preprocessing »morpheme analysis—
syntax analysis semantic analysis —-dialog analysis.

TECHNOLOGIES RELATED TO BIG DATA

F. Data scientist
A data scientist is an expert who can collect, organize, investigate, analyze, and visualize data. The data
scientist provides information necessary for corporate/organizational decision-making by collecting, analyzing, and
discovering the value of data, using various platform foundations and analysis infrastructures related to big data.
Capabilities of the data scientist
 Point of view Capability Details
Understanding the business of the company concerned
Business
and expressing it as a business model.
Data management Exploring and integrating the internal/external data of a
corporation, and manipulating structured/unstructured
Management data.
Data analysis Predictive analysis based on data mining/statistics, and
analysis based on cognitive psychology, R, and
visualization techniques.
Change management Establishment of the data strategy, communication skills.

TECHNOLOGIES RELATED TO BIG DATA

Capabilities of the data scientist
Point of view Capability Details
• Experience and training are needed to
Understanding
understand statistical analysis tools.
statistical analysis tools
• R, SAS, SPSS
 Programming language • Overall knowledge and experience in
programming are needed.
• Understanding of various languages such as C, Java,
Ruby and Perl.
RDBMS technology • Ability to design keys, indexes, queries,
normalization, and constraints based on SQL.
Technology – Distributed computing • Hadoop (MapReduce, HDFS), NoSQL (Cassandra,
capability BigTable, MongoDB)
Mathematical • Understanding of matrix operation and
knowledge numerical analysis.

DEFINITION AND CHARACTERISTICS OF NOSQL

A NoSQL database provides other means of processing than the tabular relations used in relational databases.
NoSQL is a non-relational distributed data repository that can be expanded horizontally, such as data replication and
distributed storage on multiple servers focusing on the write speed for processing unstructured and ultra-high-
capacity data.
 Characteristics Description
Processing of large
- Provides a loose data structure that allows data to be processed at the petabyte level.
scale data
Use of flexible schemas - Saves data relatively freely without a predefined schema.
- Saves (data) in a simplified form such as the key value, graph, and document
structure.
Inexpensive cluster - Supports scale-out, data replication, and distributed storage using multiple servers
configuration composed of Pc-level commercial hardware.
Simple CLI (Call - No query language like SQL in existing relational databases is
Level Interface( provided. - A simple interface is provided by calling a simple API or
HTPP.
High availability - NoSQL loads data by automatically dividing data items into the cluster environment.

DEFINITION AND CHARACTERISTICS OF NOSQL (CONT.)

Characteristics Description
- While the relational DBMS focuses on ensuring logical structure and ACID, NoSQL
Allow as much
makes the application process some of the integrity works instead of assigning
integrity is needed
them all to the DBMS.
 Schema-less - The methods of saving data are largely divided into column, value, document, and
chart, using a function that allows data storage and access using the key values,
without the fixed data schema for data modeling.
Elasticity - NoSQL has a structure that allows expansion of the system’s scale and performance
and distribution of the I/O load more easily, so that large-scale data can be
created, updated, and queried, while not causing downtime for any clients and
application systems that access the system, even if the system fails partially.
Query - NoSQL provides query language, related processing technology, and API that can
efficiently search and process data according to the characteristics of data even
in a system composed of tens or thousand of servers.
Caching - NoSQL has a structure in which memory-based caching technology is very important,
and which can provide a high-performance response speed even for large-scale
queries and be consistently applied to development and operation.

DEFINITION AND CHARACTERISTICS OF NOSQL (CONT.)

Characteristics Description

High scalability - Partitioning allows a gradual node increase.

High availability - There is no single point of failure, and data are available even
though a certain node is down because they are replicated.
High performance - The result should be quickly returned based on memory
instead of disk, which can be achieved by using the non-
blocking write and low complexity algorithm.
Atomicity - Each write operation needs to be atomic.
Consistency - Strong consistency is not needed, but the resulting consistency
is sufficient (Read-Your-Writes).

DEFINITION AND CHARACTERISTICS OF NOSQL (CONT.)

Characteristics Description

Persistence - Data should be kept on a disk, not just in a volatile memory only.
Deployment - When a node is added or deleted, data should be automatically
loaded without the need for data distribution or manual
mediation, and there should be no constraints, such as
distributed file system or shared storage, or any need for
special hardware. Hardware should be operable in
heterogeneous hardware.
 Modeling - Data of various types such as key-value pairs, hierarchical data,
flexibility and graphs should be modeled conveniently.
Query flexibility - Multiple GET that obtains a set of values for the provided key
from a query, and queries that obtain data based on a specific
range of keys, are needed.

BASE ATTRIBUTES OF NOSQL

Description of NoSQL’s BASE properties
Characteristics Description

- Emphasis is placed on availability and the use of optimistic

Basically locking and queue.
Available - Availability is ensured even with multiple failures and copies
are stored in multiple storages.
 Soft-State - Node status is determined by the information transmitted from
outside.
- Updates between distributed nodes are updated when data
reach the node.
Eventually - The property of maintaining consistency optimally even though
Consistent consistency is lost temporarily.

BASE ATTRIBUTES OF NOSQL

Comparison of BASE attributes and ACID attributes
Attribute BASE ACID

Application field - NoSQL - RDBMS

Scope - Characteristics of the entire - Limited to transactions
system
Consistency - Weak consistency - Strong consistency
aspect
 Main points - Focused on “availability” - Focused on “commit”
System aspect - Focused on “performance” - Focused on “strict data
management”
Efficiency - “Query design” is important - “Table design” is important

BASE ATTRIBUTES OF NOSQL

NoSQL databases can be divided as follows from the viewpoint of the data model that stores data.
Types of NoSQL
Type Description
- The most basic NoSQL database that provides simple and fast Get, Put and
Key-value based Delete functions based on the key-value.
- Dynamo, Redis, MemcacheDB, etc.
Column family based - An NoSQL database that stores data in rows in the column family, which
corresponds to the table in the relational database.
- Cassandra, Hbase, SimpleDB, etc.
 Document based - An NoSQL database that stores documents such as XML, JSON, BSON, etc.
in the value part of the key-value database, such as NoSQL DB, MongoDB,
CouchDB, etc.
Graph based - An NoSQL database that expresses the entry attribute of the relational
database as a node and the relationship as the edge between nodes, such
as Neo4J, Flock DB, etc.

BASE ATTRIBUTES OF NOSQL

CHARACTERISTICS OF THE NoSQL DATA MODEL
Characteristics of NoSQL data modeling compared to relational DB data modeling
BASE ATTRIBUTES OF NOSQL
CHARACTERISTICS OF THE NoSQL DATA MODEL
Characteristics of NoSQL data modeling compared to relational DB data modeling
BASE ATTRIBUTES OF NOSQL- CHARACTERISTICS OF THE NOSQL DATA MODEL 1.
Definition of the CAP theorem
 This theory asserts that as it is impossible for a distributed data store to simultaneously satisfy all of
data consistency, availability, and partition tolerance, only two should be strategically selected.
2. Composition and management strategy of the CAP theorem
BASE ATTRIBUTES OF NOSQL- CHARACTERISTICS OF THE NOSQL DATA MODEL
LINKS TO VIDEO SOURCES

 Big Data In 5 Minutes | What Is Big Data?| Introduction To Big Data |Big Data Explained |
 Simplilearn  https://www.youtube.com/watch?v=bAyrObl7TYE
 Hadoop In 5 Minutes | What Is Hadoop? | Introduction To Hadoop | Hadoop Explained |
Simplilearn  https://www.youtube.com/watch?v=aReuLtY0YMI
 SQL vs NoSQL | Difference Between SQL And NoSQL | SQL And NoSQL Tutorial | SQL Training |
Simplilearn  https://www.youtube.com/watch?v=jh14LlMHyds&t=3s