Database & Graph Research Systems @ CUHK

Earlier research systems spanning distributed SQL execution, graph analytics, and compressed subgraph processing.

These earlier research systems form the database and graph-systems foundation behind my later work on production data infrastructure and AI-native systems.

SeccoSQL

Distributed SQL execution with communication separated from computation. SeccoSQL explores a distributed query-processing model that exposes communication and local computation as separate operators, enabling a larger optimization space for complex SQL and graph-style queries.

DISC

Distributed graph analytics through relational query processing. DISC decomposes local subgraph counting into relational queries so that distributed execution and relational optimization techniques can be reused for graph workloads.

Crystal

Compressed execution for distributed subgraph matching. Crystal computes compressed results directly to reduce intermediate and output materialization costs for very large graph-pattern workloads.

Research Threads

Distributed query execution
Secco · SIGMOD 2022

Separates communication from local computation in distributed query plans to expose a larger optimization space.

System connection: Provides the distributed query-execution foundation for this systems line.

Distributed graph analytics
DISC · VLDB 2020

Expresses local subgraph counting as relational queries so distributed query-processing techniques can be reused.

System connection: Connects graph analytics to relational execution in this systems line.

Compressed graph processing
Crystal · VLDB 2018

Computes compressed subgraph-matching results directly to avoid large intermediate and output materialization.

System connection: Introduces compressed execution to control intermediate and output explosion.