Graph Algorithms Overview

TigerGraph's Graph Data Science (GDS) Library is a collection of over 50 ready-to-use GSQL queries, each implementing a standard graph or machine learning algorithm.

1. Why Run Algorithms In-Database?

Unlike traditional tools that require exporting data to a separate ML environment, TigerGraph runs algorithms directly where the data resides:

• Massive Parallelism: Leverages all CPU cores to compute metrics across billions of edges.
• Real-Time Updates: Algorithms can run on the most recent data without stale exports.
• No Data Movement: Eliminates the latency and security risks of ETL processes.

2. Library Structure

The library is open-source and extensible. You can find the raw GSQL code on GitHub or install them directly via GraphStudio.

Maturity Classifications

• Production: Optimized for speed and resource efficiency. Rigorously tested.
• Beta: Full-featured and well-tested but may have minor optimizations pending.
• Alpha: Basic functionality for experimental use.

3. How to Use Algorithms

There are three main ways to deploy an algorithm:

1. GraphStudio: Choose from a library of pre-loaded algorithms and install them with one click.
2. Packaged Templates: Use the TigerGraph CLI to install templated versions that compile on-the-fly.
3. Custom GSQL: Download the .gsql file, modify it for your specific schema, and install it as a standard query.

4. Key Categories

• Centrality: Which nodes are the "hubs" or influencers?
• Community: Which nodes form clusters or hidden rings?
• Pathfinding: What is the most efficient way to get from A to B?
• Similarity: Which nodes are most "like" each other based on their connections?

[!TIP] All algorithms in the GDS library are written in standard GSQL. You can open any algorithm to learn how complex traversals and accumulators are implemented.