heavydb

Why we included this project

Analytic databases usually force a trade-off between keeping summaries of your data handy and waiting out long scans on the full tables; HeavyDB, the open-source engine that grew out of MapD and OmniSciDB, sidesteps it by being a columnar, SQL-based store that compiles each query on the fly with LLVM and spreads the work across GPU and CPU cores. Multi-billion-row tables can be scanned in milliseconds, with no pre-aggregation, no downsampling, and no index maintenance, which suits data science and BI teams who want to interrogate raw data at full granularity instead of managing rolled-up cubes. Python users can pull results into notebooks through its Arrow-compatible connector, and the GPU rendering module visualizes large geospatial or time-series results straight from SQL. If you have NVIDIA GPUs, this is one of the few open databases built around that kind of acceleration.