This is an initial beta release of GRainDB: a new system that extends the RDBMS (DuckDB) to provide graph modeling, querying, and visualization capabilities. In addition, GRainDB modifies the internals of DuckDB to provide a set of fast join capabilities, such as predefined pointer-based joins that use system-level record IDs (RID) and adjacency list-like RID indices, to make DuckDB more efficient on graph workloads.
To understand our internals better, we refer readers to our accepted papers on GRainDB:
- Making RDBMSs Efficient on Graph Workloads Through Predefined Joins. PVLDB 2022.
- GRainDB: A Relational-core Graph-Relational DBMS. CIDR 2022.
Performance Benefits of GRainDB
Our evaluations of GRainDB on LDBC SNB showed that, by integrating predefined joins, GRainDB can outperform vanilla DuckDB, and closed the performance gap with GraphflowDB, a specialized GDBMS.
To run our evaluations, please visit our github repo, and follow our instructions to download the source code and dataset. After compilation, simply run the following command to get our performance benefits.
> ./build/release/benchmark/benchmark_runner "LDBC_LIGHT_OPTIMIZED_Q[0-9][0-9][0-9].*" --out=snb_evaluation.out
Please visit our papers for lots of more detailed evaluation results, and interesting findings!
Demonstration of GRainDB
We provide a specialized demonstration repo, in which users can give a try on our graph modeling and visualization capabilities on top of a RDBMS.
- Our presentation at CIDR 2022.
Please feel free to submit any bug reports as issues and contact me with any questions, feedback or suggestions: firstname.lastname@example.org