Schema-Based Query Optimisation for Graph Databases (2403.01863v2)

Abstract: Recursive graph queries are increasingly popular for extracting information from interconnected data found in various domains such as social networks, life sciences, and business analytics. Graph data often come with schema information that describe how nodes and edges are organized. We propose a type inference mechanism that enriches recursive graph queries with relevant structural information contained in a graph schema. We show that this schema information can be useful in order to improve the performance when evaluating acylic recursive graph queries. Furthermore, we prove that the proposed method is sound and complete, ensuring that the semantics of the query is preserved during the schema-enrichment process.

• The paper introduces a type inference mechanism that rewrites recursive graph queries using schema-based information to preserve semantics and boost performance.
• It implements a three-module system architecture that translates schema-enriched queries into recursive SQL, achieving up to 3.8x faster query runtimes on the YAGO dataset.
• The approach optimizes acyclic recursive queries across various RDBMS platforms and lays the groundwork for further innovations in graph database query processing.

Schema-Based Query Optimization for Graph Databases

The paper "Schema-Based Query Optimisation for Graph Databases" presents a method for enhancing recursive graph queries with schema-based information to optimize their performance. This approach leverages the structural constraints provided by graph schemas to improve the evaluation of acyclic recursive graph queries while maintaining semantic consistency. This essay dissects the theoretical underpinnings, implementation strategies, and experimental evaluation presented in the paper.

Type Inference and Schema Utilization

The core contribution of this paper lies in the use of a type inference mechanism that enriches recursive graph queries using graph schema information. This mechanism involves the use of graph schema triples to rewrite queries into a more optimized form. The paper outlines the formation of basic and general graph schema triples, which serve as the basis for interpreting path expressions in the context of a given schema.

The inference system employs a series of rewrite rules and transformation procedures to simplify and annotate path expressions. Soundness and completeness theorems ensure that the rewritten queries preserve the semantics of the original queries within the confines of schema specifications. The rewritten queries are expressed using the formalism of Union of Conjunctive Queries with Tarski's algebra (UCQT), which allows the incorporation of schema-derived annotations.

System Implementation

The paper details a three-module system architecture that implements the schema-based query optimization approach:

1. Rewriter: Simplifies and rewrites UCQT queries using schema-derived information, producing schema-enriched UCQT queries.
2. Translator: Converts the enriched UCQT queries into recursive SQL queries that are compatible with relational database management systems (RDBMS). This step involves translating UCQT to recursive relational algebra and then to SQL.
3. Backend: Facilitates execution on various RDBMS platforms. The approach uses a relational representation of graph databases, mapping nodes and edges into relational tables, and leverages standard SQL mechanisms to execute recursive queries.

   Figure 1: System architecture.

Performance Evaluation

The paper evaluates the schema-based approach using two datasets: YAGO, a real-world knowledge graph, and LDBC-SNB, a synthetic benchmark for property graphs. Experiments focus on the query runtimes of recursive, acyclic, and cyclic queries across different scale factors and RDBMS platforms.

Key Findings

• YAGO Dataset: The schema-based approach improves query performance significantly, with queries executing 3.8 times faster on average compared to the baseline.

  Figure 2: Query runtime for YAGO dataset.

• LDBC-SNB Dataset: Acyclic recursive queries benefit most from schema-based optimization, particularly as dataset size increases. The results illustrate that while cyclic queries do not gain as much from schema information, the acyclic recursive queries show substantial improvement.

  Figure 3: Runtime based on query shape.

• Cross-RDBMS Evaluation: The system demonstrates consistent performance improvements across multiple RDBMS platforms, highlighting the generalizability of the approach.

  Figure 4: Query runtime on different RDBMS for YAGO.

Conclusion

The schema-based query optimization approach detailed in this paper presents a robust framework for enhancing recursive graph queries by exploiting schema constraints. The proposed methodology not only preserves query semantics but also demonstrates tangible performance improvements across varied datasets and systems. Future work may focus on extending schema capabilities to encompass property constraints and exploring optimizations for cyclic queries, which currently benefit less from schema-based rewriting. This research underscores the potential of schema-driven approaches in optimizing graph database queries and paves the way for further innovations in graph query processing.