A Comparative Analysis of Document-oriented and Column-oriented Approaches
(Nandar Lwin, University Of Sunderland)
Introduction
NoSQL databases, which offer scalability and flexibility that traditional relational databases (RDBMS) frequently lack,
are changing the data warehousing environment. The usage of document-oriented NoSQL databases (MongoDB) and
column-oriented NoSQL databases (HBase) for data warehousing is the subject of two important research studies that
are examined in this poster.
Paper 1: Implementation of NoSQL-based Data Warehouses
Emphasis: MongoDB and other document-oriented databases. MongoDB enables for flexible and dynamic schemas by organizing data into
documents, which are formats akin to JSON.
Benefits
Data modeling flexibility: MongoDB's ability for dynamic schemas makes it simpler to adjust for shifting needs.
Excellent performance in reading and writing: effective management of write and read operations, particularly for big datasets.
Effectively scales to massive data sets: Easily expands horizontally to meet expanding data requirements.
Implementation: Real-world knowledge gained from creating MongoDB data marts.
The article highlights the applicability of MongoDB for data warehousing by offering insights from actual applications.
Results: By utilizing their adaptability and efficiency for complex data processing jobs, document-oriented NoSQL databases can efficiently
assist data warehousing.
Paper 2: Physical Data Warehouse Design on NoSQL Databases
Emphasis: Databases with columns (HBase).
HBase optimizes for read-intensive tasks that are typical in data analysis by storing data in column families.
Benefits
High efficiency for tasks including analysis: Optimized for read-intensive activities, column-oriented storage is essential for online
analytical processing, or OLAP.
Effective read-only operations: By reading only the columns that are required, this technique minimizes the amount of I/O operations.
Execution: OLAP query performance evaluation using the Star Schema Benchmark based on various physical architectures.
To determine the effects of various data distribution strategies on performance, experiments were carried out.
Results: OLAP query performance is strongly influenced by the physical organization of data into column families, with different distributions
maximizing performance for different types of queries.
Comparison
Flexibility: MongoDB offers more flexible data modeling compared to
HBase.
MongoDB's document-oriented model allows for dynamic schema
changes, making it adaptable to varying data structures.
Performance: HBase excels in read-intensive workloads due to its
columnar storage structure.
HBase's design optimizes read operations by storing related data
together in column families, which is beneficial for analytical queries.
Scalability: Both databases provide robust scalability, but their
effectiveness depends on the specific use case (document-oriented vs.
analytical queries).
MongoDB and HBase both support horizontal scaling, but their
architecture and data organization methods cater to different types
of data processing needs.
Main Papers Reviewed:
1. Oditis, I., Bicevska, Z., Bicevskis, J., & Karnitis, G. (Year). Implementation of
NoSQL-based Data Warehouses. Journal Name.
This paper explores the implementation and benefits of using MongoDB in
data warehouses.
2. Scabora, L.C., Brito, J.J., Ciferri, R.R., & Ciferri, C.D.A. (2016). Physical Data
Warehouse Design on NoSQL Databases: OLAP Query Processing over HBase.
Proceedings of the 18th International Conference on Enterprise Information
Systems (ICEIS 2016).
This paper investigates the physical design and OLAP query performance
in HBase.
Additional References:
Bridgwater, A. (2015). Data Warehousing Guide. Oracle Inc.
Woodey, C. (2014). Overview of NoSQL Databases.
Han, J., Kamber, M., & Pei, J. (2003). Data Mining: Concepts and Techniques.
Morgan Kaufmann Publishers.
Gaspar, A., & Coric, Z. (2017). The Merging Trends of NoSQL and RDBMS
Technologies.
