[Paper Review] Views over RDF Datasets: A State-of-the-Art and Open Challenges

Lorena Etcheverry, Alejandro Vaisman|arXiv (Cornell University)|Nov 1, 2012

Semantic Web and Ontologies28 references3 citations

TL;DR

This paper proposes a formal definition of views over RDF datasets to address the lack of consensus in the literature, evaluates three leading view mechanisms (including Networked Graphs) against four key use cases—virtual data integration, query answering, data security, and query modularization—and identifies critical open challenges in expressiveness, scalability, inference support, and query optimization, particularly for materialized views and execution plan tuning.

ABSTRACT

Views on RDF datasets have been discussed in several works, nevertheless there is no consensus on their definition nor the requirements they should fulfill. In traditional data management systems, views have proved to be useful in different application scenarios such as data integration, query answering, data security, and query modularization. In this work we have reviewed existent work on views over RDF datasets, and discussed the application of existent view definition mechanisms to four scenarios in which views have proved to be useful in traditional (relational) data management systems. To give a framework for the discussion we provided a definition of views over RDF datasets, an issue over which there is no consensus so far. We finally chose the three proposals closer to this definition, and analyzed them with respect to four selected goals.

Motivation & Objective

• To establish a standardized definition of views over RDF datasets, addressing the current lack of consensus in the literature.
• To evaluate existing RDF view mechanisms against four key application scenarios: virtual data integration, query answering, data security, and query modularization.
• To identify and analyze open challenges in view expressiveness, scalability, support for RDFS inference, and query optimization in RDF systems.
• To assess the feasibility and performance of existing tools (e.g., Networked Graphs) in supporting view-based querying over large-scale RDF datasets.
• To advocate for the inclusion of explicit view definition mechanisms (e.g., CREATE VIEW) in the SPARQL specification to enable advanced query rewriting and optimization.

Proposed method

• Proposes a formal definition of views over RDF datasets based on SPARQL graph patterns, emphasizing reusability, modularity, and semantic consistency.
• Evaluates three representative view mechanisms—Networked Graphs, RDFMatView, and Dritsou et al.’s path-based views—against four use-case scenarios.
• Employs experimental evaluation on datasets of 500K and 2M triples to measure query performance degradation under varying conditions, including RDFS inference.
• Analyzes query execution plans and performance bottlenecks, particularly for UNION and JOIN operations in SPARQL without OPTIONAL clauses.
• Identifies gaps in existing systems regarding view materialization, update, and maintenance, especially in dynamic, web-scale RDF environments.
• Proposes the need for standardized, extensible view mechanisms in SPARQL to support query rewriting, optimization, and security policies.

Experimental results

Research questions

• RQ1 What constitutes a proper definition of a view over RDF datasets, and how can it be formally grounded to ensure consistency and reusability?
• RQ2 How well do existing RDF view mechanisms support key use cases such as virtual data integration, secure data access, and query modularization?
• RQ3 What are the performance implications of using views in large-scale RDF datasets, particularly when RDFS inference is enabled?
• RQ4 To what extent can existing systems support efficient query rewriting and optimization when views are involved?
• RQ5 What are the open challenges in maintaining and updating materialized RDF views in dynamic, distributed data environments?

Key findings

• Query performance degrades by more than 10 times when scaling from 500K to 2M triples in Networked Graphs, indicating scalability limitations.
• The inclusion of RDFS inference support causes an average 4-fold performance degradation in query execution times on 500K-triple datasets.
• UNION queries exhibit significantly worse performance compared to other SPARQL constructs, highlighting a critical bottleneck in view-based query evaluation.
• No existing system provides native support for query optimization over views, such as pushdown filtering or join reordering, despite their potential to reduce execution costs.
• There is no standardized mechanism for defining views in SPARQL, and no support for materialized view maintenance, which is essential for dynamic web data.
• The absence of a CREATE VIEW statement in SPARQL limits the ability to semantically model and reason about views, hindering advanced query rewriting and security policy enforcement.