Source-linked AI summary
NoSQL Database: New Era of Databases for Big data Analytics - Classification, Characteristics and Comparison
A B M Moniruzzaman, Syed Akhter Hossain
TL;DR
Big Data’s scale and variety expose limitations in relational databases, especially for large and unstructured datasets. The paper surveys NoSQL characteristics, classifies its database types, and evaluates them across selected attributes, reporting adoption evidence and practical uses while noting that NoSQL may not guarantee ACID properties.
Problem
Growing data volume, variety, and scale create modeling and horizontal-scalability problems for relational databases used with large and unstructured datasets.
Method
The paper provides an overview of NoSQL characteristics, classifies databases into four categories, and evaluates them using a matrix covering design, integrity, indexing, distribution, and system attributes.
Results
49% of respondents cited rigid schemas as the primary migration driver, while MongoDB accounted for 45% of NoSQL technology mentions in LinkedIn profiles.
Takeaways & Limitations
NoSQL approaches support large-scale processing, embedded retrieval, exploratory semi-structured analytics, and high-volume storage across distributed systems.
Takeaways & Limitations
NoSQL databases did not guarantee ACID properties.
Abstract
from arXiv · showhide
Digital world is growing very fast and become more complex in the volume (terabyte to petabyte), variety (structured and un-structured and hybrid), velocity (high speed in growth) in nature. This refers to as Big Data that is a global phenomenon. This is typically considered to be a data collection that has grown so large it can not be effectively managed or exploited using conventional data management tools: e.g., classic relational database management systems (RDBMS) or conventional search engines. To handle this problem, traditional RDBMS are complemented by specifically designed a rich set of alternative DBMS; such as - NoSQL, NewSQL and Search-based systems. This paper motivation is to provide - classification, characteristics and evaluation of NoSQL databases in Big Data Analytics. This report is intended to help users, especially to the organizations to obtain an independent understanding of the strengths and weaknesses of various NoSQL database approaches to supporting applications that process huge volumes of data.
1. Introduction
NoSQL denotes non-relational database systems designed for very large datasets whose structure does not require a relational model. They support distributed, massively parallel processing, while NewSQL retains relational semantics with scalable architectures.
- NoSQL means “Not Only SQL” and describes non-relational systems that generally do not organize data primarily in tables or use SQL for manipulation.
- NoSQL systems target large-scale storage and massively parallel processing across many commodity servers.
- NoSQL uses non-SQL languages and mechanisms, although some APIs translate SQL queries into native query languages.
- NewSQL databases provide ACID-compliant real-time OLTP and conventional SQL-based OLAP in Big Data environments.
- NewSQL scales relational database performance through features such as column-oriented storage, distributed architectures, in-memory processing, SMP, and MPP.
2. Background
The relational model faces data-modeling and horizontal-scalability problems as data volume, interdependency, and application complexity increase. NoSQL emerged as a horizontally scalable alternative for large-scale analytical and unstructured-data workloads, though it may not guarantee ACID properties.
- Relational databases increasingly encounter data-modeling deficits and horizontal-scalability constraints across several servers and large datasets.
- Data volume has grown exponentially through users, systems, sensors, and concentration on distributed cloud systems.
- Internet applications, Web 2.0, social networks, and standardized data access have increased data interdependency and complexity.
- Big Data Analytics, Business Intelligence, and social networking over petabyte datasets pushed centralized SQL-like databases toward their limits.
- NoSQL databases developed as horizontally scalable, distributed data stores for organizations handling large amounts of unstructured data.
- NoSQL databases may not guarantee ACID properties, distinguishing them from relational database management systems.
3. Characteristics of NoSQL Databases
Distributed database systems face a trade-off among consistency, availability, and partition tolerance. NoSQL commonly relaxes consistency through BASE-style designs and supports several large-scale storage, retrieval, analytics, and processing uses.
- The CAP theorem states that distributed systems cannot fully achieve strong consistency, high availability, and partition tolerance simultaneously.
- NoSQL databases often loosen consistency requirements to improve availability and partitioning, resulting in BASE systems with eventual consistency.
- NoSQL databases support large-scale parallel processing, embedded information retrieval, exploratory analytics on semi-structured data, and large-volume storage.
- They provide relatively inexpensive, highly scalable storage for high-volume historical records and unwieldy semi-structured or unstructured data.
- Their distributed framework supports massive batch processing operations including aggregation, filtering, sorting, and algorithmic computation.
4. Classification of NoSQL Databases
The paper classifies NoSQL databases into four categories—key-value, document, wide-column, and graph—each suited to different data-management tasks. Their designs range from simple exact-key retrieval to semi-structured storage, distributed column processing, and relationship traversal.
- Classification: NoSQL databases are classified into key-value, document, wide-column, and graph stores.The paper presents these as four basic categories suited to different kinds of tasks.
- Key-Value stores: Key-value stores associate identifiers with values in simple tables and usually support exact searches on keys.Values may be strings, lists, or sets, while searches generally cannot target values.
- Document databases: Document databases store semi-structured XML, JSON, or BSON documents whose keys and values are fully searchable.They support irregular or sparse data, including literal documents and denormalized product or customer entities.
- Wide-Column stores: Wide-column stores use distributed, column-oriented structures that accommodate multiple attributes per key.Bigtable-style systems may combine distributed storage with file systems and MapReduce frameworks for large-scale processing.
- Graph databases: Graph databases represent nodes, relationships, and properties as interconnected structures, emphasizing relationship traversal over querying.They are used for social networks, recommendations, and forensic pattern detection.
- Adoption: NoSQL attracted attention because massive websites encountered data-utilization rates that conventional relational databases struggled to handle.The paper also describes NoSQL solutions as handling huge quantities of data across clusters of commodity servers.
5. Comparison of NoSQL Database
The paper compares four NoSQL database categories using a matrix of design, integrity, indexing, distribution, and system attributes. The supplied passages identify the evaluation framework but report no comparison outcomes.
- Evaluation framework: The evaluation compares four NoSQL database categories across design, integrity, indexing, distribution, and system attributes.The comparison is presented as a matrix in Table 1.
6. Adoption of NoSQL Database
NoSQL adoption is associated with growing organizational investment, migration away from relational systems, and strong demand for NoSQL expertise. MongoDB is identified as the most popular database in LinkedIn profile mentions.
- Adoption drivers: 49% of Couchbase survey respondents cited rigid schemas as the primary driver for migrating from relational to NoSQL databases.Lack of scalability and high latency or low performance also ranked highly among migration reasons.
- Organizational investment: Nearly half of more than 1,300 respondents had funded NoSQL projects in the first half of 2012.Among companies with more than 250 developers, nearly 70% planned to fund NoSQL projects during 2012.
- Operational importance: 40% of respondents said NoSQL was very important or critical to daily operations, while another 37% said it was becoming more important.
- Workforce demand: Organizations with massive storage needs are examining NoSQL, while NoSQL database expertise is highly demanded by developing organizations.The cited section presents job trends for five NoSQL databases from Indeed.com.
- Technology popularity: MongoDB accounted for 45% of mentions of NoSQL technologies in LinkedIn profiles, making it the most popular database in that measure.
7. Conclusion
The paper concludes that NoSQL addresses large-scale data requirements through distributed, non-relational approaches. It surveys NoSQL characteristics, four feature-based categories, comparative attributes, and adoption.
- Uses: NoSQL supports large-scale data processing through parallel processing over distributed systems.
- Uses: NoSQL supports embedded information retrieval, exploratory analytics on semi-structured data, and large-volume storage across several data forms.
- Paper scope: The paper examines NoSQL characteristics, classifies databases into four feature-based categories, and compares them using design, integrity, indexing, distribution, and system attributes.
- Paper scope: The study aims to provide an independent understanding of the strengths and weaknesses of NoSQL approaches for applications processing huge data volumes.
Authors
The paper lists two authors associated with Daffodil International University in Dhaka, Bangladesh. Their research and professional backgrounds span computing, cloud computing, big data, distributed systems, and database modeling.
- Authors: A. B. M. Moniruzzaman is a research associate at Daffodil International University and researches cloud computing and Big Data Analytics.
- Authors: Syed Akhter Hossain is a professor and department head at Daffodil International University with research areas including distributed systems and database modeling.