As the demand for analyzing and computing complex relational graph data in the big data domain continues to grow, the need for real-time graph computing capabilities in scenarios such as financial risk control and group detection is also on the rise. This requires large-scale graph computation engines to simultaneously provide low latency and high throughput. Currently, stream-batch integrated computing engines are the future direction, but the iterative nature of graph computation differs fundamentally from traditional stream and batch processing. Building a stream-batch-graph integrated computing engine around graph iteration poses a significant challenge. GeaFlow, independently developed by the Ant Group’s graph computing team, is the industry’s first large-scale distributed streaming graph computing engine, supporting real-time graph computing, offline graph computing, and online exploratory analysis on large-scale graphs. In this presentation, we will introduce how GeaFlow splits execution jobs into stream, batch, and graph tasks, orchestrates them in a unified manner, and constructs an efficient stream-batch-graph integrated scheduling and execution framework.

Speakers:

Congbin Qiao: Technical Expert at Ant Group

Congbin Qiao is a technical expert at Ant Group, specializing in the fields of big data, distributed computing, and graph computing. Since joining Ant Group’s graph computing team in 2020, he has been actively involved in the core development of Ant Group’s self-developed streaming graph computing engine, GeaFlow, supporting the real-time computation and analytics for large-scale graph scenarios in business applications.