Introducing GraphScope#

What is GraphScope#

GraphScope is a comprehensive distributed graph computing platform that offers a user-friendly Python interface for performing various graph operations on a cluster of computers. By integrating key Alibaba technologies such as GRAPE, MaxGraph, Graph-Learn (GL), and Vineyard, GraphScope simplifies multi-stage processing of large-scale graph data on compute clusters. These technologies enable analytics, interactive, and graph neural networks (GNN) computation, while Vineyard provides efficient in-memory data transfers.

Why use GraphScope#

GraphScope has several advantages:

  • Performance: GraphScope significantly outperforms other state-of-the-art systems, offering up to 2X or higher performance improvements. Moreover, it includes a collection of optimized built-in graph algorithms for immediate use.

  • Compatibility: It supports Gremlin and Cypher(coming soon) for querying graph, and NetworkX compatible APIs for graph analytic.

  • PyData Integration: GraphScope offers a user-friendly Python interface, allowing seamless integration with PyData frameworks and management of complex workflows involving multiple stages beyond graph processing.

  • Cloud Native: GraphScope is designed for easy deployment on Kubernetes, offering excellent elasticity and scalability.

What are the Use Cases#

GraphScope is versatile and can be used for:

  • Offline Graph Analytical Jobs: Iterative graph computations like pagerank, centrality, community detection, and graph clustering. Utilize built-in algorithms or custom ones for large graphs.

  • Online Graph BI Analysis: Perform interactive graph analysis with complex Gremlin or Cypher queries, focusing on low latency.

  • High QPS Graph Queries: Process high-rate graph queries with exceptional throughput.

  • Graph Neural Network Training and Inference: Support large-scale GNNs and integrate seamlessly with PyG/TensorFlow.

  • One-stop Graph Processing: Manage complex workflows involving multiple stages of graph analytics, queries, GNNs, and beyond.

What are the Limitations#

  • GraphScope is not a graph database: GraphScope focuses on analyzing and computing large-scale graphs using distributed computing, rather than storing and managing data in a graph format with ACID and transactions. See Positioning of GraphScope for more information.

  • Security limitations: GraphScope currently lacks essential security features, such as authentication and encryption.

What are the Next#

You may want to