Relational Database Management Systems (RDBMS) have been the most prominent form of database in the world for several decades. While relational databases are often applied within high-frequency/low-volume transactional applications such as website backends, the poor performance of relational databases on low-frequency/high-volume queries often precludes their application to big data analysis fields like graph analytics. This work explores the construction of an RDBMS solution that uses the GraphBLAS API to execute Structured Query Language (SQL) in an effort to improve performance on high-volume queries. Tables are redefined to be collections of sparse scalars, vectors, matrices, and more generally sparse tensors. The explicit values (nonzeros) in these sparse tensors define the rows and NULL values within the tables. A prototype database called TenSQL was constructed and evaluated against several SQL implementations including PostgreSQL. Preliminary results comparing the performance on queries common in graph analysis applications offer performance improvements as high as 1,400x over PostgreSQL for moderately sized datasets when returning results in a columnar format.
TenSQL was created by Sandia National Laboratories, with assistance provided by the University of Utah.
TenSQL has only been tested with Python 3.9. Python 3.10 is too new for the version of numpy supported by pygraphblas.
To install from PyPI:
pip install tensql
To install from source:
git clone 'https://github.com/sandialabs/TenSQL.git'
cd TenSQL
pip install .
To run the tests, you must first clone the sourcecode from github, and then build the extensions and install testing dependencies.
git clone 'https://github.com/sandialabs/TenSQL.git'
cd TenSQL
python setup.py build_ext --inplace
pip install -e ".[test]"
The tests can then be run either with the run_tests.py script which outputs
code coverage information:
python3 run_tests.py
Or via python's built-in unittest module
python3 -m unittest -v tensql.test
Specific tests can be run via the unittest module:
python3 -m unittest -v tensql.test.test_queries.xAy.TestQuery_xAy
Note: Certain tests for memory leaks can take about a minute to execute.
To run the benchmarks, you must first clone the sourcecode from github, and then build the extensions and install testing dependencies.
git clone 'https://github.com/sandialabs/TenSQL.git'
cd TenSQL
python setup.py build_ext --inplace
pip install -e ".[test,benchmark]"
You must also install PostgreSQL 15 to run the postgres tests.
Once installed, you can run the benchmarks via slurm with:
bash download_benchmark_data.sh
bash benchmark_twohop.sh
bash benchmark_ingest_and_named_edges.sh
Alternatively, you can run single benchmarks (without slurm) like this:
bash download_benchmark_data.sh
bash single_twohop.sh "`pwd`/tmp" "`pwd`/results" all
bash single_ingest_and_named_edges.sh "`pwd`/tmp" "`pwd`/results" all
Note: You will likely need to tune the settings in postgresql.conf if your
system has less memory than our benchmarking system.
Note: Running some benchmarks requires a very large amount of memory (hundreds of gigabytes).
TenSQL was described in the paper "An SQL Database Built on GraphBLAS", which was accepeted by the IEEE High Performance Extreme Computing Virtual Conference in September 2023. It has not yet been published in IEEE Xplore.
Roose, J. P., Vaidya, M., Sadayappan, P., & Rajamanickam, S. (2023). TenSQL: An SQL Database Built on GraphBLAS.
IEEE High Performance Extreme Computing Virtual Conference, forthcoming.