Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Distributed Evaluation of Graph Queries using Recursive Relational Algebra

Sarah Chlyah, Pierre Genevès, Nabil Layaïda

TL;DR

This work tackles scalable evaluation of recursive graph queries by extending μ-RA to Dist-$\mu$-RA, enabling distributed execution through fixpoint splitting and stable-column partitioning. It introduces two Spark-based strategies, $\mathcal{P}_{\texttt{gld}}$ and $\mathcal{P}_{\texttt{plw}}$, and a cost-driven physical plan generator, backed by a modular architecture (Query2Mu, MuRewriter, CostEstimator, PhysicalPlanGenerator) with PostgreSQL and Spark backends. Empirical results on Yago, Uniprot, and synthetic graphs show Dist-$\mu$-RA outperforms GraphX and BigDatalog on most UCRPQ classes, especially for large intermediate results, while preserving correctness via formal fixpoint properties and data-partitioning guarantees. The approach reduces inter-node communication, broadens the practical applicability of recursive graph querying in distributed environments, and provides a principled cost model to guide plan selection. Overall, Dist-$\mu$-RA delivers significant efficiency gains for recursive graph queries and lays groundwork for broader integration with traditional RDBMS ecosystems.

Abstract

We present a system called Dist-$μ$-RA for the distributed evaluation of recursive graph queries. Dist-$μ$-RA builds on the recursive relational algebra and extends it with evaluation plans suited for the distributed setting. The goal is to offer expressivity for high-level queries while providing efficiency at scale and reducing communication costs. Experimental results on both real and synthetic graphs show the effectiveness of the proposed approach compared to existing systems.

Distributed Evaluation of Graph Queries using Recursive Relational Algebra

TL;DR

This work tackles scalable evaluation of recursive graph queries by extending μ-RA to Dist--RA, enabling distributed execution through fixpoint splitting and stable-column partitioning. It introduces two Spark-based strategies, and , and a cost-driven physical plan generator, backed by a modular architecture (Query2Mu, MuRewriter, CostEstimator, PhysicalPlanGenerator) with PostgreSQL and Spark backends. Empirical results on Yago, Uniprot, and synthetic graphs show Dist--RA outperforms GraphX and BigDatalog on most UCRPQ classes, especially for large intermediate results, while preserving correctness via formal fixpoint properties and data-partitioning guarantees. The approach reduces inter-node communication, broadens the practical applicability of recursive graph querying in distributed environments, and provides a principled cost model to guide plan selection. Overall, Dist--RA delivers significant efficiency gains for recursive graph queries and lays groundwork for broader integration with traditional RDBMS ecosystems.

Abstract

We present a system called Dist--RA for the distributed evaluation of recursive graph queries. Dist--RA builds on the recursive relational algebra and extends it with evaluation plans suited for the distributed setting. The goal is to offer expressivity for high-level queries while providing efficiency at scale and reducing communication costs. Experimental results on both real and synthetic graphs show the effectiveness of the proposed approach compared to existing systems.
Paper Structure (33 sections, 5 equations, 15 figures, 1 algorithm)

This paper contains 33 sections, 5 equations, 15 figures, 1 algorithm.

Table of Contents

  1. Introduction
  2. Preliminaries
  3. $\mu$-RA syntax
  4. Semantics and properties of the fixpoint
  5. Fixpoint parallelization and distribution
  6. Fixpoint parallelization principle
  7. Data partitioning
  8. Fixpoint distributed evaluation on Spark
  9. Global Loop on the Driver ($\mathcal{P}_\texttt{gld}$)
  10. Parallel Local loops on the Workers ($\mathcal{P}_\texttt{plw}$)
  11. Physical plan generation and selection
  12. Dist-$\mu$-RA architecture
  13. Example
  14. Experiments
  15. Experimental setup
  16. ...and 18 more sections

Figures (15)

  • Figure 1: Grammar of $\mu$-RA mura-sigmod20.
  • Figure 2: Graph example.
  • Figure 3: Distributed execution of $\mathcal{P}_\texttt{gld}$ (top) & $\mathcal{P}_\texttt{plw}$ (bottom).
  • Figure 4: Distributed execution of $\mathcal{P}_\texttt{plw}$ plans.
  • Figure 5: Comparison between $\mathcal{P}_\texttt{plw}^\texttt{pg}$ and $\mathcal{P}_\texttt{plw}^\texttt{s}$.
  • ...and 10 more figures

Theorems & Definitions (2)

  • Example 1
  • Example 2