[Paper Review] NoSQL Databases: Yearning for Disambiguation

Chaimae Asaad, Karim Bäına+1 more|arXiv (Cornell University)|Mar 9, 2020

Cloud Computing and Resource Management22 references4 citations

TL;DR

This paper provides a comprehensive survey and classification of NoSQL databases by data model, benchmarks key systems, and identifies major conceptual ambiguities—such as CAP theorem interpretation, multimodel vs. polyglot persistence, and distinctions between column-family and columnar databases—offering disambiguated definitions to unify understanding in research and industry.

ABSTRACT

The demanding requirements of the new Big Data intensive era raised the need for flexible storage systems capable of handling huge volumes of unstructured data and of tackling the challenges that traditional databases were facing. NoSQL Databases, in their heterogeneity, are a powerful and diverse set of databases tailored to specific industrial and business needs. However, the lack of theoretical background creates a lack of consensus even among experts about many NoSQL concepts, leading to ambiguity and confusion. In this paper, we present a survey of NoSQL databases and their classification by data model type. We also conduct a benchmark in order to compare different NoSQL databases and distinguish their characteristics. Additionally, we present the major areas of ambiguity and confusion around NoSQL databases and their related concepts, and attempt to disambiguate them.

Motivation & Objective

• Address the lack of theoretical foundation and consensus in NoSQL database research, which leads to widespread confusion among experts and practitioners.
• Clarify fundamental ambiguities in NoSQL terminology and concepts, such as CAP theorem interpretation, the role of graph databases, and distinctions between column-family and columnar databases.
• Provide a systematic classification of NoSQL databases by data model (key-value, document, column-family, graph) to improve conceptual clarity.
• Benchmark representative NoSQL systems to compare their characteristics, consistency models, and performance trade-offs under real-world conditions.
• Promote standardization and unification of NoSQL design methodologies by resolving long-standing misconceptions and fostering shared understanding in both research and industry.

Proposed method

• Conduct a comprehensive survey of NoSQL databases, categorizing them based on their underlying data models and architectural characteristics.
• Perform a comparative benchmarking study of leading NoSQL databases to evaluate their behavior under various consistency and availability conditions.
• Analyze the CAP theorem's application in NoSQL systems, particularly focusing on trade-offs between consistency, availability, and partition tolerance.
• Examine and contrast related concepts such as multimodel databases versus polyglot persistence, clarifying their technical distinctions and use cases.
• Investigate and resolve common misconceptions, including the confusion between column-family and columnar databases, and the misapplication of the CAP theorem.
• Use empirical findings from system behavior under network partitions to assess consistency models and guide recommendations for system selection based on use case.

Experimental results

Research questions

• RQ1 What are the primary sources of conceptual ambiguity in NoSQL databases, and how do they hinder consensus and standardization?
• RQ2 How do different NoSQL database families (key-value, document, column-family, graph) differ in data model, scalability, and consistency guarantees?
• RQ3 To what extent do the CAP theorem and consistency models (e.g., eventual vs. strong consistency) influence the design and selection of NoSQL systems?
• RQ4 How do multimodel databases differ from polyglot persistence architectures, and what are the implications for system design and data modeling?
• RQ5 What are the key performance and reliability trade-offs between leading NoSQL databases such as MongoDB, Cassandra, and Redis under network partition scenarios?

Key findings

• The CAP theorem remains a critical and valid framework for distributed database design, despite common misinterpretations; it highlights the inherent trade-off between consistency and availability during network partitions.
• MongoDB demonstrates stronger consistency guarantees under fault conditions, making it suitable for applications requiring data integrity, while Cassandra excels in high availability and fault tolerance.
• Redis Cluster does not guarantee write persistence during network partitions, as data loss can occur if the master node fails before replication, underscoring the limitations of eventual consistency in some configurations.
• The distinction between column-family databases (e.g., Cassandra) and columnar databases (e.g., data warehouse systems) is frequently misunderstood; the former is optimized for distributed, high-throughput workloads, while the latter is designed for analytical queries on structured data.
• Multimodel databases, which support multiple data models within a single storage engine, differ fundamentally from polyglot persistence, where multiple distinct data stores are used—offering better integration and reduced complexity.
• The absence of a stable theoretical foundation and standardized terminology has led to widespread conceptual confusion, particularly around core concepts like 'schema-less', 'NoSQL', and 'ACID properties', which impedes both research and industrial adoption.