HugeGraph Computing (OLAP)
The HugeGraph-Computer repository contains two graph computing systems with different deployment and runtime models. Start with Go Vermeer for general graph algorithms; use Computer when you need distributed Java BSP/Pregel computation. Both can connect to HugeGraph, but their configuration and job entry points are not interchangeable.
Default entry: Go Vermeer
With load.type=hugegraph, the Vermeer master queries PD for partition metadata and workers scan HStore Store partitions directly. Workers write results through the Server REST API only when output.type=hugegraph; other input sources or output settings do not require those connections.
Java Computer
Job configuration can be submitted through the Kubernetes Operator or YARN. Workers read graph data through the HugeGraph Server REST API and can write results back according to the output configuration. With HDFS input or output, workers access HDFS directly. The master uses etcd to coordinate BSP jobs.
Vermeer high-performance in-memory graph computing: start once, execute repeatedly, with 15+ OLAP algorithms, seconds-to-minutes execution, deployment, loading, PageRank, and community detection.