GES @ Huawei

Introduces a production graph service architecture with composable components and factorized execution for high-concurrency interactive workloads.

System connection: Provides the core service architecture and factorized executor behind the production system.

Separates reads and writes and decouples version data to support concurrent queries and updates on dynamic graphs.

System connection: Provides scalable concurrent storage for dynamic graph workloads.

Compares in-memory dynamic graph storage designs and identifies key storage and concurrency bottlenecks.

System connection: Motivates the storage and concurrency choices in this design space.

Develops high-concurrency incremental graph query processing for evolving graph data.

System connection: Extends the system line toward incremental processing under concurrent updates.

Studies reachability under ordered label and time-dependent constraints, with indexing strategies that balance construction cost and query efficiency.

System connection: Extends the Huawei-era graph research portfolio toward indexed graph reachability queries.

Updates attributed-network embeddings in a streaming style while retaining information from previously observed attributes.

System connection: Represents adjacent work on evolving attributed graphs rather than a component of GES.

Develops a label-free adversarial attack for graph classification using contrastive representations and learned edge perturbations.

System connection: Represents adjacent graph-learning research and is not presented as part of the GES implementation.