GES

Benchmark-leading graph database service architecture for high-throughput interactive workloads.

GES is a production graph database service architecture developed at Huawei for high-throughput interactive graph workloads.

Problem

Interactive graph services need to serve complex graph queries with high concurrency, predictable latency, and production-grade manageability. Traditional graph engines often optimize for either offline graph analytics or isolated benchmark kernels, but production services must combine query throughput, maintainability, and extensibility.

Core idea

The newer GES line uses a composable service architecture and factorized execution to improve high-concurrency query processing. The goal is to make graph query execution fast without turning the system into a benchmark-only prototype.

My role

Research and system architecture for graph database infrastructure and the GES line of work, including execution design and benchmark-facing system optimization.

Evidence

Impact

GES reached #1 on the LDBC SNB Interactive declarative track with reported throughput over 3,000× the previous #2 result, and later reached #1 on the imperative track with Graph Engine Service 3.1.0 at SF300.