Internal Architecture

TigerGraph is the world's first Native Parallel Graph (NPG) system. Unlike other graph databases that layer graph logic on top of relational or document stores, TigerGraph is built from the ground up for graph analytics.

1. Core Philosophy

TigerGraph's NPG architecture is built on two pillars:

• Local Storage: Data is stored in a format optimized for graph traversals.
• Parallel Computation: The engine is designed to execute graph algorithms across multiple CPU cores and cluster nodes simultaneously.

2. Key Components

The system consists of several specialized services that coordinate via a message-passing design:

Service	Name	Description
GSE	Graph Storage Engine	Manages the physical storage and retrieval of vertices and edges.
GPE	Graph Processing Engine	The heart of the system; executes GSQL queries and parallel algorithms.
RESTPP	RESTful Server	The API gateway that handles incoming REST requests and coordinates tasks.
Kafka	Message Queue	Manages the stream of data updates and internal communication.
GSQL	GSQL Server	Handles the parsing, compilation, and execution of GSQL scripts.

3. Interaction Layers

Users can interact with the architecture through multiple interfaces:

• GSQL Client: Command-line interface for schema and query management.
• GraphStudio: A browser-based GUI for visual modeling and exploration.
• REST API: Production-grade integration for applications.
• gAdmin: The administrative tool for system health and configuration.

4. Performance Advantages

• Fast Loading: 50–150 GB per hour, per machine.
• Real-Time Updates: Supports streaming 2B+ daily events into 100B+ vertex graphs.
• Massive Traversal: Capable of traversing hundreds of millions of edges per second per node.